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 B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,1,balanced,0.11984533071517944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,1,balanced,0.13699199755986533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,1,balanced,0.0874826709429423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,1,balanced,0.2436586618423462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,1,balanced,0.19610132773717245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,1,balanced,0.4554986556371053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,1,balanced,0.33033599456151325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,1,balanced,0.14071999986966452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,1,balanced,0.4394133488337199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,1,balanced,0.3373066584269206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,1,balanced,0.23990400632222494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,1,balanced,0.44192532698313397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,1,balanced,0.33376534779866535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,1,balanced,0.4445013205210368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,1,balanced,0.3301173249880473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,1,balanced,0.243231991926829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,1,balanced,0.44971199830373126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,1,balanced,0.33701332410176593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,1,balanced,0.031397332747777305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,1,balanced,0.31019200881322223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,1,balanced,0.4479466676712036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,1,balanced,0.24167466163635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,1,balanced,0.31383999188741046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,1,balanced,0.45054932435353595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,1,balanced,0.03201066702604294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,1,balanced,0.33132266998291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,1,balanced,0.4492479960123698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,1,balanced,0.3200266758600871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,1,power_law_1.01,0.12507519721984864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,1,power_law_1.01,0.0301503986120224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,1,power_law_1.01,0.13662079572677613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,1,power_law_1.01,0.08673920035362244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,1,balanced,0.464629332224528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,1,balanced,0.03162666658560435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,1,balanced,0.24618667364120483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,1,balanced,0.29639999071757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,1,balanced,0.4660319884618123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,1,balanced,0.3025706609090169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,1,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,1,power_law_1.01,0.11345920562744141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,1,power_law_1.01,0.029267200827598573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,1,balanced,0.46984533468882245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,1,power_law_1.01,0.1803007960319519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,1,power_law_1.01,0.16366080045700074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,1,balanced,0.3116213281949361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,1,balanced,0.2493013342221578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,1,power_law_1.01,0.15493760108947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,1,balanced,0.0356480007370313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,1,balanced,0.4957066774368286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,1,balanced,0.2821919918060303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,1,power_law_1.01,0.23756799697875977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,1,power_law_1.01,0.030329599976539612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,1,power_law_1.01,0.2403775930404663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,1,power_law_1.01,0.20265600681304932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,1,balanced,0.04155199974775314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,1,balanced,0.25438400109608966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,1,power_law_1.01,0.22809600830078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,1,balanced,0.4991413354873657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,1,balanced,0.24920000632603964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,1,balanced,0.04634666442871094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,1,balanced,0.24202134211858115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,1,power_law_1.01,0.033670398592948916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,1,power_law_1.01,0.4436160087585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,1,power_law_1.01,0.24524800777435302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,1,power_law_1.01,0.3352191925048828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,1,balanced,0.5077439943949381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,1,power_law_1.01,0.2512063980102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,1,balanced,0.47995734214782715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,1,balanced,0.047925333182017006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,1,balanced,0.24931732813517252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,1,power_law_1.01,0.43695359230041503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,1,power_law_1.01,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,1,power_law_1.01,0.33244800567626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,1,power_law_1.01,0.24775679111480714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,1,balanced,0.5397866566975912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,1,balanced,0.3834880193074544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,1,balanced,0.0640533318122228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,1,power_law_1.01,0.2544447898864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,1,power_law_1.01,0.44768638610839845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,1,power_law_1.01,0.039212799072265624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,1,balanced,0.5281440019607544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,1,balanced,0.2527359922726949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,1,power_law_1.01,0.3245568037033081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,1,balanced,0.6272000074386597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,1,balanced,0.053930665055910744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,1,power_law_1.01,0.2611583948135376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,1,power_law_1.01,0.447814416885376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,1,balanced,0.05894933144251505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,1,balanced,0.6766613324483236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,1,power_law_1.01,0.2652224063873291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,1,power_law_1.01,0.04431999921798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,1,power_law_1.01,0.33070080280303954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,1,balanced,0.6172106663386027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,1,balanced,0.2540586590766907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,1,power_law_1.01,0.2800256013870239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,1,balanced,0.072202667593956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,1,power_law_1.01,0.4585408210754395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,1,balanced,0.9676640033721924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,1,power_law_1.01,0.32945919036865234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,1,power_law_1.01,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,1,power_law_1.01,0.28714239597320557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,1,balanced,0.8819733460744222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,1,balanced,0.2682186762491862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,1,balanced,0.07839466631412506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,1,power_law_1.01,0.29687678813934326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,1,power_law_1.01,0.4583295822143555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,1,power_law_1.01,0.3109568119049072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,1,power_law_1.01,0.06408960223197938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,1,balanced,1.2726240158081055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,1,power_law_1.01,0.31651840209960935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,1,balanced,0.07445333401362102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,1,balanced,0.2695840001106262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,1,balanced,0.9466773668924967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,1,power_law_1.01,0.48423042297363283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,1,power_law_1.01,0.32126080989837646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,1,power_law_1.01,0.321779203414917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,1,power_law_1.01,0.051744002103805545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,1,balanced,0.10802132884661357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,1,power_law_1.01,0.3516992092132568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,1,power_law_1.01,0.4741312026977539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,1,balanced,1.8513654073079426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,1,balanced,0.2805759906768799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,1,power_law_1.01,0.4043136119842529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,1,power_law_1.01,0.3186624050140381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,1,balanced,0.08659733335177104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,1,balanced,1.3350666364034016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,1,power_law_1.01,0.05686399936676025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,1,power_law_1.01,0.49843201637268064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,1,power_law_1.01,0.46973438262939454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,1,balanced,0.29019733270009357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,1,power_law_1.01,0.33240959644317625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,1,balanced,0.09992532928784688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,1,power_law_1.01,0.5895487785339355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,1,power_law_1.01,0.502668809890747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,1,power_law_1.01,0.06944640278816223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,1,balanced,2.4459786415100098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,1,power_law_1.01,0.7000383853912353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,1,balanced,0.2914560039838155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,1,balanced,1.6298133532206218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,1,power_law_1.01,0.3350271940231323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,1,power_law_1.01,0.5672575950622558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,1,balanced,0.1418773333231608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,1,power_law_1.01,0.958944034576416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,1,power_law_1.01,0.07567359805107117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,1,power_law_1.01,0.3668607950210571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,1,balanced,0.30766934156417847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,1,power_law_1.01,0.5955071926116944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,1,power_law_1.01,1.1854463577270509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,1,balanced,0.16547200083732605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,1,power_law_1.01,0.07285119891166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,1,power_law_1.01,1.6815231323242188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,1,power_law_1.01,0.6517375946044922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,1,balanced,2.4173332850138345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,1,power_law_1.01,0.38078720569610597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,1,balanced,3.605152130126953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,1,balanced,0.3758560021718343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,1,balanced,0.23400533199310303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,1,power_law_1.01,2.2087039947509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,1,power_law_1.01,0.10542720556259155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,1,power_law_1.01,0.6501952171325683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,1,power_law_1.01,0.387007999420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,1,power_law_1.01,3.150092887878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,1,power_law_1.01,0.7335680007934571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,1,balanced,0.3886186679204305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,1,power_law_1.01,0.08324480056762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,1,balanced,0.29712533950805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,1,power_law_1.01,4.103359985351562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,1,power_law_1.01,0.7864640235900879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,1,power_law_1.01,0.4149759769439697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,1,balanced,3.2377920150756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,1,power_law_1.01,0.09786239862442017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,1,balanced,0.5406826734542847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,1,power_law_1.01,0.9553728103637695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,1,power_law_1.01,5.059187316894532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,1,balanced,4.792794545491536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,1,balanced,0.4226986567179362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,1,power_law_1.01,0.13879679441452025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,1,power_law_1.01,0.4932544231414795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,1,power_law_1.01,1.1722496032714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.01,0.16268160343170165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,1,power_law_1.01,7.945273590087891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.01,0.563321590423584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,1,power_law_1.01,1.5469951629638672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,1,balanced,0.603440006573995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,1,balanced,0.5349440177281698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.01,0.23231360912322999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,1,balanced,4.803104082743327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.01,0.7019775867462158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,1,power_law_1.01,1.8591295242309571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.01,0.29468159675598143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,1,balanced,0.8547093073527018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,1,power_law_1.01,16.398316955566408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.01,0.8180095672607421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,1,power_law_1.01,2.6582847595214845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.01,0.4208960056304932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,1,balanced,5.945413589477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,1,balanced,0.788703997929891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.01,0.5330304145812989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.01,1.1499903678894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,1,power_law_1.01,3.6978431701660157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,1,balanced,1.0672213236490886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.01,0.7868415832519531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.01,1.4369983673095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,1,power_law_1.01,5.330643081665039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.01,1.026860809326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,1,balanced,6.5393492380778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,1,balanced,1.0285813013712566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.01,2.0155008316040037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.01,1.5179583549499511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,1,power_law_1.01,7.037465667724609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.01,2.5826047897338866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,1,balanced,1.5768532752990723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.01,2.0169536590576174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.01,3.7941761016845703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.01,2.5077632904052733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,1,power_law_1.01,8.660671997070313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,1,balanced,1.5200053850809734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,1,balanced,9.47869873046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.01,4.953606414794922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,1,power_law_1.01,3.9575809478759765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,1,balanced,2.08023993174235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,1,balanced,8.215183893839518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,1,power_law_1.01,13.232608032226562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.01,6.114169692993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,1,power_law_1.01,7.849037170410156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,1,balanced,2.014645258585612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,1,balanced,3.05297056833903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,1,power_law_1.01,9.612032318115235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,1,power_law_1.01,26.0893310546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,1,power_law_1.01,18.989523315429686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,1,balanced,12.99188232421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,1,balanced,2.509941260019938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,1,balanced,4.0117387771606445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,1,balanced,19.0806884765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,1,balanced,4.9678347905476885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,1,balanced,3.964122772216797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,1,balanced,25.939066569010418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,1,balanced,7.849568049112956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,1,balanced,7.858879725138347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,1,balanced,16.195770263671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,1,power_law_1.2,0.13672319650650025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,1,power_law_1.2,0.1730944037437439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,1,power_law_1.2,0.21713919639587403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,1,power_law_1.2,0.437113618850708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,1,power_law_1.2,0.435968017578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,1,power_law_1.2,0.4501632213592529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,1,power_law_1.2,0.4517375946044922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,1,power_law_1.2,0.46085758209228517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,1,power_law_1.2,0.4759488105773926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,1,power_law_1.2,0.4801472187042236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,1,power_law_1.2,0.49097599983215334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,1,power_law_1.2,0.49918718338012696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,1,power_law_1.2,0.5099584102630615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,1,power_law_1.2,0.5223296165466309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,1,power_law_1.2,0.6085696220397949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,1,power_law_1.2,0.6347455978393555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,1,power_law_1.2,0.08551040291786194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,1,power_law_1.2,0.030636799335479737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,1,power_law_1.2,0.12343679666519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,1,power_law_1.2,0.6589951992034913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,1,power_law_1.2,0.1602560043334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,1,power_law_1.2,0.029446399211883544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,1,power_law_1.2,0.1130687952041626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,1,power_law_1.2,0.7097983837127686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,1,power_law_1.2,0.8067904472351074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,1,power_law_1.2,0.20102400779724122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,1,power_law_1.2,0.03079040050506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,1,power_law_1.2,0.1331264019012451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,1,power_law_1.2,0.9695167541503906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,1,balanced,0.12571199735005698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,1,power_law_1.2,0.03372159898281098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,1,power_law_1.2,0.2264319896697998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,1,balanced,0.20982933044433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,1,power_law_1.2,0.33255040645599365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,1,power_law_1.2,1.151916790008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,1,balanced,0.21452800432840982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,1,power_law_1.2,0.034815999865531924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,1,power_law_1.2,0.23498880863189697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,1,power_law_1.2,0.3259455919265747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,1,balanced,0.3787200053532918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,1,power_law_1.2,1.5810175895690919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,1,balanced,0.39292800426483154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,1,balanced,0.7243200143178304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,1,power_law_1.2,0.3150592088699341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,1,power_law_1.2,0.039027199149131775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,1,power_law_1.2,0.24403200149536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,1,power_law_1.2,1.891084861755371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,1,balanced,0.38991467157999676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,1,balanced,0.7274026870727539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,1,power_law_1.2,0.3253567934036255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,1,balanced,0.3922133445739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,1,power_law_1.2,0.04490880072116852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,1,power_law_1.2,0.2532032012939453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,1,balanced,0.7260746955871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,1,power_law_1.2,2.665843200683594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,1,power_law_1.2,0.3212863922119141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,1,balanced,0.39399464925130206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,1,balanced,0.7333172957102457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,1,power_law_1.2,0.04588159918785095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,1,power_law_1.2,0.24897921085357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,1,power_law_1.2,0.3110975980758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,1,balanced,0.3964373270670573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,1,power_law_1.2,3.728403091430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,1,balanced,0.7326666514078776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,1,power_law_1.2,0.32027521133422854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,1,power_law_1.2,0.0636031985282898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,1,balanced,0.39800000190734863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,1,power_law_1.2,0.2556864023208618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,1,balanced,0.7378933429718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,1,power_law_1.2,0.3123647928237915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,1,balanced,0.3956960042317708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,1,power_law_1.2,0.26093440055847167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,1,power_law_1.2,0.05187199711799621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,1,balanced,0.7330079873402914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,1,power_law_1.2,5.393791961669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,1,power_law_1.2,0.3283263921737671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,1,balanced,0.39850131670633954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,1,power_law_1.2,0.26080639362335206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,1,balanced,0.7368319829305013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,1,power_law_1.2,0.05681279897689819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,1,power_law_1.2,0.32875521183013917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,1,balanced,0.404090682665507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,1,power_law_1.2,0.27917439937591554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,1,balanced,0.7463839848836263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,1,power_law_1.2,0.33164799213409424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,1,balanced,0.4123680194218953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,1,power_law_1.2,0.06924160122871399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,1,power_law_1.2,6.935308837890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,1,balanced,0.1830400029818217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,1,balanced,0.7474186420440674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,1,power_law_1.2,0.2887808084487915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,1,balanced,0.4131466547648112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,1,power_law_1.2,0.34628479480743407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,1,power_law_1.2,0.07591040134429931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,1,balanced,0.75436798731486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,1,balanced,0.44422932465871173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,1,power_law_1.2,0.2919872045516968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,1,power_law_1.2,0.373088002204895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,1,power_law_1.2,0.07402880191802978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,1,balanced,0.32313599189122516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,1,balanced,0.44394667943318683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,1,power_law_1.2,8.887091064453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,1,balanced,0.7979146639506022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,1,power_law_1.2,0.39631359577178954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,1,power_law_1.2,0.31833600997924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,1,power_law_1.2,0.10497920513153076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,1,balanced,0.4492479960123698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,1,balanced,0.7850186824798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,1,power_law_1.2,0.48624000549316404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,1,balanced,0.5230506658554077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,1,power_law_1.2,0.08312320113182067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,1,power_law_1.2,0.3190975904464722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,1,balanced,0.4673173427581787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,1,balanced,0.7967840035756429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.2,0.5424895763397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,1,power_law_1.2,0.3481856107711792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,1,power_law_1.2,0.09845759868621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,1,balanced,0.5846720139185587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,1,power_law_1.2,13.240345764160157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,1,balanced,0.544922669728597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,1,balanced,0.897653341293335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,1,balanced,0.04805333415667216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.2,0.684115219116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,1,power_law_1.2,0.40213761329650877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,1,power_law_1.2,0.13904000520706178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,1,balanced,0.5895040035247803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,1,balanced,1.013749361038208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,1,balanced,0.4908906618754069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.2,0.8045824050903321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,1,power_law_1.2,0.478220796585083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.2,0.16224000453948975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,1,balanced,0.8001493612925211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,1,balanced,0.998528003692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,1,power_law_1.2,0.5865791797637939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.2,0.2313472032546997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.2,1.1153663635253905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,1,balanced,0.5273973147074381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,1,balanced,0.0483893354733785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,1,balanced,0.8981013298034668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,1,power_law_1.2,0.7032320022583007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,1,balanced,1.3899626731872559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.2,0.2936320066452026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,1,power_law_1.2,26.10406494140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,1,balanced,0.04841066896915436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.2,1.4177215576171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,1,balanced,0.51747198899587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,1,balanced,1.2317279974619548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.2,0.4206399917602539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,1,balanced,1.473493258158366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,1,balanced,0.522208015124003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.2,2.0524864196777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.2,0.5333439826965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,1,balanced,0.04986133178075155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,1,balanced,1.6394826571146648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,1,power_law_1.2,0.9603391647338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,1,balanced,0.46402132511138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.2,0.7857984066009521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,1,balanced,2.051487922668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.2,2.559833526611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,1,balanced,0.47176531950632733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.2,1.0297087669372558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,1,power_law_1.2,1.1744895935058595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,1,balanced,0.050383999943733215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.2,3.7311168670654298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,1,balanced,2.3764373461405435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,1,balanced,2.6986347834269204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,1,balanced,0.4957706530888875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.2,1.5172032356262206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,1,balanced,0.48027201493581134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,1,balanced,0.062165334820747375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,1,power_law_1.2,1.671072006225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.2,2.0170495986938475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.2,4.877286529541015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,1,balanced,3.952437400817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,1,balanced,3.208319981892904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,1,balanced,0.45584531625111896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,1,balanced,0.07669866581757863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.2,2.509503936767578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,1,balanced,0.44741864999135333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.2,6.057484817504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,1,power_law_1.2,2.169388771057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,1,balanced,0.423253337542216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,1,balanced,5.332741419474284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,1,balanced,0.06863999863465627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,1,power_law_1.2,3.959539031982422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,1,balanced,0.37171733379364014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,1,balanced,4.712602615356445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,1,balanced,0.3530720074971517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,1,power_law_1.2,3.139507293701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,1,power_law_1.2,9.578253173828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,1,balanced,0.09098666906356812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,1,balanced,0.7292533715566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,1,power_law_1.2,7.902642822265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,1,balanced,0.08135466774304707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,1,balanced,7.886629104614258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,1,power_law_1.2,4.102508926391602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,1,balanced,0.10539199908574422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,1,balanced,0.6030506690343221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,1,balanced,6.270528157552083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,1,balanced,0.12119999527931213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,1,balanced,0.8482826550801595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,1,power_law_1.2,19.035801696777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,1,balanced,0.13909332950909933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,1,power_law_1.2,5.048889541625977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,1,balanced,0.10267733534177144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,1,balanced,1.0825386842091878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,1,balanced,0.1462559998035431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,1,balanced,10.828404744466146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,1,power_law_1.2,7.910546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,1,balanced,0.127402663230896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,1,balanced,0.16405866543451944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,1,balanced,7.788901646931966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,1,balanced,1.56005859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,1,balanced,0.2129759987195333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,1,balanced,0.27566399176915485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,1,power_law_1.2,16.031269836425782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,1,balanced,2.045482635498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,1,balanced,0.3915626605351766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,1,balanced,0.5199946562449137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,1,balanced,13.48910903930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,1,balanced,2.9721546173095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,1,balanced,0.7452426751454672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,1,balanced,12.445306142171225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,1,balanced,0.9879786968231201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,1,balanced,3.913461367289225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,1,balanced,1.4579680760701497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,1,balanced,21.28822962443034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,1,balanced,5.804037094116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,1,balanced,1.9186399777730305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,1,balanced,2.872096061706543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,1,balanced,7.698917388916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,1,balanced,3.7999839782714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,1,balanced,24.38983408610026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,1,balanced,9.641007741292318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,1,balanced,4.734933217366536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,1,balanced,42.338277180989586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,1,balanced,15.415743509928385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,1,balanced,7.674277623494466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,1,balanced,30.968228658040363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,1,balanced,15.631956736246744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,1,power_law_1.01,0.20819199085235596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,1,power_law_1.01,0.2798912048339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,1,power_law_1.01,0.3860352039337158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,1,power_law_1.01,0.7185023784637451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,1,power_law_1.01,0.7114943981170654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,1,power_law_1.01,0.7304895877838135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,1,power_law_1.01,0.7198016166687011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,1,power_law_1.01,0.7293568134307862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,1,power_law_1.01,0.7569791793823242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,1,power_law_1.01,0.12443519830703735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,1,power_law_1.01,0.7679296016693116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,1,power_law_1.01,0.16768640279769897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,1,power_law_1.01,0.7919616222381591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,1,power_law_1.01,0.04681600034236908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,1,power_law_1.01,0.17973120212554933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,1,power_law_1.01,0.8336832046508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,1,power_law_1.01,0.23582079410552978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,1,power_law_1.01,0.8271167755126954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,1,power_law_1.01,0.24518399238586425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,1,power_law_1.01,0.3833600044250488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,1,power_law_1.01,0.8420031547546387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,1,power_law_1.01,0.04632959961891174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,1,power_law_1.01,0.3696000099182129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,1,power_law_1.01,0.2853951930999756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,1,power_law_1.01,0.8998335838317871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,1,power_law_1.01,0.40586237907409667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,1,power_law_1.01,0.5330687999725342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,1,power_law_1.01,0.4209727764129639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,1,power_law_1.01,0.046323201060295104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,1,power_law_1.01,0.899392032623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,1,power_law_1.01,0.5103040218353272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,1,power_law_1.01,0.40171518325805666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,1,power_law_1.01,0.9973631858825683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,1,power_law_1.01,0.4239999771118164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,1,power_law_1.01,0.5068096160888672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,1,power_law_1.01,0.04795520007610321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,1,power_law_1.01,1.172000026702881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,1,power_law_1.01,0.41950078010559083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,1,power_law_1.01,0.5136703968048095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,1,power_law_1.01,1.2991744041442872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,1,power_law_1.01,0.049081599712371825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,1,power_law_1.01,0.42665600776672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,1,power_law_1.01,0.5040256023406983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,1,power_law_1.01,1.608799934387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,1,power_law_1.01,0.4663871765136719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,1,power_law_1.01,0.060710400342941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,1,power_law_1.01,0.4534463882446289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,1,power_law_1.01,0.49974398612976073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,1,power_law_1.01,1.8936832427978516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,1,power_law_1.01,0.07504000067710877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,1,power_law_1.01,0.46004481315612794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,1,power_law_1.01,0.48275198936462405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,1,power_law_1.01,2.5278335571289063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,1,power_law_1.01,0.06716799736022949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,1,power_law_1.01,0.49981441497802737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,1,power_law_1.01,0.47927680015563967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,1,power_law_1.01,0.0898751974105835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,1,power_law_1.01,3.1131839752197266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,1,power_law_1.01,0.495359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,1,power_law_1.01,0.4934783935546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,1,power_law_1.01,0.4987008094787598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,1,power_law_1.01,0.551852798461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,1,power_law_1.01,0.0798143982887268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,1,power_law_1.2,0.20992000102996827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,1,power_law_1.01,4.51357421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,1,power_law_1.01,0.6383039951324463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,1,power_law_1.01,0.5199679851531982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,1,power_law_1.01,0.10451200008392333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,1,power_law_1.2,0.2636415958404541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,1,power_law_1.01,0.7272704124450684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,1,power_law_1.01,0.5986176013946534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,1,power_law_1.01,0.12088960409164429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,1,power_law_1.01,6.158406448364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,1,power_law_1.2,0.4220287799835205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,1,power_law_1.01,0.5996352195739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,1,power_law_1.01,0.8884480476379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,1,power_law_1.01,0.13541120290756226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,1,power_law_1.2,0.7392896175384521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,1,power_law_1.01,0.640825605392456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,1,power_law_1.01,1.1008000373840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,1,power_law_1.2,0.7290751934051514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,1,power_law_1.01,0.10079360008239746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,1,power_law_1.01,8.870438385009766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,1,power_law_1.01,0.7693503856658935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,1,power_law_1.01,1.4583552360534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,1,power_law_1.2,0.7359871864318848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,1,power_law_1.01,0.14391039609909057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.01,0.8951935768127441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,1,power_law_1.2,0.7511231899261475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,1,power_law_1.01,1.8477567672729491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.01,1.148921585083008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,1,power_law_1.01,0.1246783971786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,1,power_law_1.2,0.7632512092590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,1,power_law_1.01,11.668045043945312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,1,power_law_1.01,2.562227249145508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,1,power_law_1.2,0.7846144199371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.01,1.3377087593078614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,1,power_law_1.01,0.16234879493713378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,1,power_law_1.01,3.5434944152832033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,1,power_law_1.2,0.7972928047180176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.01,1.8298688888549806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,1,power_law_1.2,0.8003071784973145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,1,power_law_1.01,0.21262080669403077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,1,power_law_1.01,14.303935241699218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,1,power_law_1.01,5.058137512207031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,1,power_law_1.2,0.811308765411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.01,2.2918912887573244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,1,power_law_1.2,0.12428159713745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.01,0.27460479736328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,1,power_law_1.2,0.816703987121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,1,power_law_1.2,0.8261183738708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.01,0.39035520553588865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,1,power_law_1.01,6.571180725097657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.01,3.260153579711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,1,power_law_1.2,0.173471999168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,1,power_law_1.2,0.9009280204772949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.01,0.5193024158477784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,1,power_law_1.2,0.9096447944641113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,1,power_law_1.2,0.21101438999176025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.01,4.172467041015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,1,power_law_1.01,21.777529907226562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.01,0.7419519901275635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,1,power_law_1.01,8.087942504882813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,1,power_law_1.2,1.0192319869995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,1,power_law_1.2,0.35735039710998534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.01,0.9870207786560059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,1,power_law_1.2,1.1197759628295898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.01,6.114956665039062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,1,power_law_1.2,0.3832511901855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,1,power_law_1.2,1.3351103782653808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.01,1.4539199829101563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,1,power_law_1.2,0.3895807981491089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,1,power_law_1.2,1.604915237426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,1,power_law_1.01,12.69823989868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,1,power_law_1.2,0.39880321025848386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.01,1.9144128799438476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.01,7.952819061279297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,1,power_law_1.2,1.8946752548217773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,1,power_law_1.2,0.40332798957824706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,1,power_law_1.2,2.549260711669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.01,2.867852783203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,1,power_law_1.01,44.145388793945315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,1,power_law_1.2,0.41437439918518065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,1,power_law_1.2,3.0510847091674806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.01,9.924249267578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,1,power_law_1.2,0.4151872158050537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.01,3.7958400726318358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,1,power_law_1.01,24.64718017578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,1,power_law_1.2,4.340460968017578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,1,power_law_1.2,0.41665282249450686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.01,4.720851135253906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,1,power_law_1.2,0.4483071804046631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,1,power_law_1.2,0.04591360092163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,1,power_law_1.2,6.140147018432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,1,power_law_1.2,0.18200960159301757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,1,power_law_1.01,15.560960388183593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,1,power_law_1.2,0.4594240188598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,1,power_law_1.2,0.04650239944458008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,1,power_law_1.2,0.2407167911529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,1,power_law_1.2,8.859935760498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,1,power_law_1.01,7.5513664245605465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,1,power_law_1.2,0.4410880088806152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,1,power_law_1.2,0.04658559858798981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,1,power_law_1.2,0.3037568092346191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,1,power_law_1.2,0.487827205657959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,1,power_law_1.2,0.04739840030670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,1,power_law_1.2,0.5330560207366943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,1,power_law_1.2,11.629344177246093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,1,power_law_1.2,0.04878079891204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,1,power_law_1.2,0.4928959846496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,1,power_law_1.2,0.5210944175720215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,1,power_law_1.2,0.060729598999023436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,1,power_law_1.01,15.121856689453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,1,power_law_1.01,30.819793701171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,1,power_law_1.2,0.5190400123596192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,1,power_law_1.2,0.5523263931274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,1,power_law_1.2,0.07475200295448303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,1,power_law_1.2,0.5165440082550049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,1,power_law_1.2,14.182707214355469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,1,power_law_1.2,0.0667967975139618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,1,power_law_1.2,0.6219903945922851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,1,power_law_1.2,0.5083712100982666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,1,power_law_1.2,0.08951039910316468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,1,power_law_1.2,0.49672961235046387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,1,power_law_1.2,0.7438975811004639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,1,power_law_1.2,0.07947520017623902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,1,power_law_1.2,0.5022848129272461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,1,power_law_1.2,0.8964544296264648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,1,power_law_1.2,0.10362240076065063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,1,power_law_1.2,0.497983980178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,1,power_law_1.2,22.06759033203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,1,power_law_1.2,0.11882239580154419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,1,power_law_1.2,1.1187135696411132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,1,power_law_1.2,0.5171840190887451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,1,power_law_1.2,0.5268671989440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,1,power_law_1.2,0.13660800457000732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,1,power_law_1.2,1.4918208122253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,1,power_law_1.2,0.1010815978050232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,1,power_law_1.2,0.5076735973358154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,1,power_law_1.2,1.8218368530273437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,1,power_law_1.2,0.1411903977394104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,1,power_law_1.2,0.5424767971038819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,1,power_law_1.2,0.12501120567321777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,1,power_law_1.2,2.6045055389404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,1,power_law_1.2,0.6075200080871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,1,power_law_1.2,42.82405090332031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,1,power_law_1.2,0.16314879655838013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,1,power_law_1.2,0.6043583869934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,1,power_law_1.2,3.592019271850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,1,power_law_1.2,0.21164801120758056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,1,power_law_1.2,0.7790847778320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.2,0.2734143972396851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,1,power_law_1.2,5.054867172241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.2,0.9099776268005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.2,0.3877376079559326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.2,1.0967040061950684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,1,power_law_1.2,6.578701019287109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.2,1.3446080207824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.2,0.5189888000488281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,1,power_law_1.2,8.099852752685546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.2,1.8355072021484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.2,0.7430784225463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.2,2.256275177001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.2,0.9877951622009278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.2,3.1668991088867187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,1,power_law_1.2,12.740902709960938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.2,1.4530688285827638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.2,4.184236907958985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.2,1.9142400741577148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.2,6.031955337524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.2,2.8701055526733397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,1,power_law_1.2,24.586790466308592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.2,7.998188781738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.2,3.7932735443115235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.2,4.722380828857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.2,9.938534545898438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,1,power_law_1.2,7.550950622558593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,1,power_law_1.2,15.675173950195312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,1,power_law_1.2,15.08362274169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,1,power_law_1.2,39.838436889648435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,1,balanced,0.06381333371003468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,1,balanced,0.10336533188819885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,1,balanced,0.14387200276056925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,1,balanced,0.09529067079226176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,1,power_law_1.01,0.0931007981300354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,1,power_law_1.01,0.1449280023574829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,1,balanced,0.14590400457382202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,1,balanced,0.06451733410358429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,1,balanced,0.13300800323486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,1,balanced,0.24150399367014566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,1,balanced,0.2432533303896586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,1,balanced,0.09455999732017517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,1,balanced,0.19217065970102945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,1,balanced,0.43653865655263263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,1,balanced,0.44566933314005536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,1,power_law_1.01,0.1421504020690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,1,power_law_1.01,0.24405760765075685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,1,balanced,0.14573333660761514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,1,balanced,0.844101349512736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,1,balanced,0.31842132409413654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,1,balanced,0.8403093020121256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,1,balanced,0.2549333373705546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,1,balanced,1.6257279713948567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,1,power_law_1.01,0.1602112054824829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,1,power_law_1.01,0.26715519428253176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,1,balanced,0.5636373360951742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,1,balanced,0.4811520179112752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,1,balanced,1.6201440493265789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,1,balanced,1.6325119336446126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,1,balanced,0.49240533510843915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,1,balanced,1.0452106793721516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,1,power_law_1.01,0.3380928039550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,1,power_law_1.01,0.6284416198730469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,1,balanced,3.166895866394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,1,balanced,1.6346133550008137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,1,balanced,0.5088213284810384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,1,balanced,1.0527466932932537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,1,power_law_1.01,0.5530816078186035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,1,balanced,1.6416959762573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,1,power_law_1.01,1.0368639945983886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,1,balanced,3.1813332239786782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,1,balanced,0.5206986665725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,1,power_law_1.01,0.09873279929161072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,1,balanced,1.0583146413167317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,1,balanced,1.644602616628011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,1,power_law_1.01,0.761356782913208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,1,balanced,0.5335893233617147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,1,power_law_1.01,1.461235237121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,1,power_law_1.01,0.049420800805091855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,1,balanced,3.1873013178507485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,1,balanced,1.655962626139323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,1,balanced,1.0609760284423828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,1,balanced,0.5598666667938232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,1,power_law_1.01,1.0922880172729492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,1,power_law_1.01,0.130022394657135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,1,power_law_1.01,2.1059968948364256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,1,balanced,1.6618506113688152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,1,balanced,1.0668799877166748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,1,balanced,3.188927968343099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,1,balanced,0.5861013333002726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,1,power_law_1.01,0.07689599990844727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,1,power_law_1.01,1.2090368270874023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,1,balanced,1.6771732966105144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,1,balanced,1.0736640294392903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,1,power_law_1.01,0.13377280235290528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,1,power_law_1.01,2.267475128173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,1,balanced,0.6113706827163696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,1,balanced,1.6848427454630535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,1,balanced,3.1988000869750977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,1,power_law_1.01,0.0726527988910675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,1,balanced,1.0834506352742512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,1,balanced,0.6647359927495321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,1,power_law_1.01,1.2325119972229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,1,power_law_1.01,2.3753215789794924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,1,power_law_1.01,0.2590912103652954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,1,balanced,1.6952640215555828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,1,balanced,0.5937653382619222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,1,balanced,1.0850133101145427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,1,power_law_1.01,1.2582847595214843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,1,balanced,3.2051467895507812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,1,power_law_1.01,0.12861440181732178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,1,power_law_1.01,0.37954559326171877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,1,power_law_1.01,2.37969913482666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,1,balanced,1.7164799372355144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,1,balanced,0.6186399857203165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,1,power_law_1.01,1.2875840187072753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,1,balanced,1.098912000656128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,1,power_law_1.01,0.5077695846557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,1,power_law_1.01,0.25045759677886964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,1,balanced,3.2235358556111655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,1,power_law_1.01,2.5197439193725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,1,balanced,1.7421919504801433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,1,balanced,0.6670560042063395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,1,power_law_1.01,1.3567872047424316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,1,power_law_1.01,0.7276288032531738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,1,balanced,1.1125120321909587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,1,power_law_1.01,0.35579519271850585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,1,power_law_1.01,2.628108787536621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,1,balanced,1.7910614013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,1,balanced,3.2382240295410156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,1,power_law_1.01,1.385862445831299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,1,balanced,0.6501386562983195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,1,power_law_1.01,0.7681471824645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,1,balanced,1.1218506495157878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,1,power_law_1.01,0.532147216796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,1,power_law_1.01,2.7212352752685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,1,power_law_1.01,1.4148736000061035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,1,balanced,1.8361013730367024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,1,power_law_1.01,0.8206912040710449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,1,balanced,3.256890614827474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,1,balanced,0.7050879796346029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,1,balanced,1.1437119642893474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,1,power_law_1.01,1.4515968322753907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,1,power_law_1.01,0.6198272228240966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,1,power_law_1.01,2.7404544830322264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,1,power_law_1.01,0.8225152015686035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,1,balanced,1.9327680269877117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,1,balanced,3.2631092071533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,1,power_law_1.01,1.4923775672912598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,1,balanced,1.181605339050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,1,power_law_1.01,0.5797632217407227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,1,power_law_1.01,2.810700798034668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,1,power_law_1.01,0.8612735748291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,1,balanced,0.7099200089772543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,1,power_law_1.01,1.5686976432800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,1,power_law_1.01,0.6405375957489013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,1,power_law_1.01,0.8797696113586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,1,balanced,3.2837438583374023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,1,power_law_1.01,2.8893823623657227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,1,balanced,2.0528586705525718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,1,balanced,1.2202719847361247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,1,power_law_1.01,1.7611967086791993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,1,power_law_1.01,0.6975743770599365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,1,power_law_1.01,0.9296256065368652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,1,balanced,0.7730773289998373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,1,power_law_1.01,3.068070411682129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,1,balanced,3.3242400487264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,1,power_law_1.01,1.7749696731567384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,1,power_law_1.01,0.7303808212280274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,1,power_law_1.01,0.946560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,1,balanced,1.3128106594085693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,1,power_law_1.01,3.361004638671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,1,power_law_1.01,2.018374443054199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,1,power_law_1.01,0.7647232055664063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,1,balanced,2.209850629170736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,1,balanced,3.382362683614095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,1,power_law_1.01,0.9499327659606933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,1,power_law_1.01,2.0302080154418944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,1,power_law_1.01,3.3469825744628907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,1,balanced,1.7671465873718262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,1,power_law_1.01,0.8297408103942872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,1,power_law_1.01,0.9820672035217285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,1,power_law_1.01,0.78504958152771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,1,power_law_1.01,2.3439807891845703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,1,balanced,1.4220800399780273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,1,power_law_1.01,3.7610111236572266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,1,balanced,3.4897867838541665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,1,power_law_1.01,1.0347776412963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,1,power_law_1.01,0.8019200325012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,1,power_law_1.01,2.6718271255493162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,1,balanced,2.3849120140075684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,1,power_law_1.01,3.5535873413085937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,1,power_law_1.01,1.1050751686096192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,1,power_law_1.01,0.8873344421386719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,1,balanced,2.0402026176452637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,1,power_law_1.01,3.3952190399169924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.01,1.171833610534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,1,balanced,3.5857601165771484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,1,power_law_1.01,4.025388717651367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,1,power_law_1.01,0.8696191787719727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,1,balanced,1.6031252543131511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.01,1.3434944152832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,1,power_law_1.01,3.9888702392578126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,1,power_law_1.01,4.836729431152344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.01,0.9419008255004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.01,1.4284992218017578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,1,power_law_1.01,5.142233657836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,1,balanced,3.6970240275065103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.01,0.974886417388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,1,power_law_1.01,5.595904159545898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,1,balanced,3.0947306950887046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.01,1.6822271347045898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.01,1.0516608238220215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,1,balanced,1.2773973147074382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,1,balanced,1.784656047821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,1,power_law_1.01,6.474944305419922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,1,power_law_1.01,6.307206344604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.01,1.9988800048828126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.01,2.6187328338623046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,1,balanced,3.858346621195475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.01,2.5721920013427733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,1,power_law_1.01,8.121389007568359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,1,power_law_1.01,7.637452697753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.01,3.0394367218017577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.01,3.1753471374511717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.01,1.7810304641723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,1,power_law_1.01,10.132230377197265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,1,power_law_1.01,11.895148468017577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.01,4.357632064819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,1,balanced,3.20908260345459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.01,2.0893375396728517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,1,balanced,2.497018655141195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,1,balanced,4.31335989634196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,1,balanced,1.3973066012064617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,1,power_law_1.01,12.036268615722657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.01,2.6869504928588865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.01,5.5625663757324215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.01,3.188319969177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,1,power_law_1.01,21.628057861328124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,1,power_law_1.01,6.697350311279297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,1,power_law_1.01,18.116064453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,1,power_law_1.01,3.7504257202148437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,1,power_law_1.01,10.029023742675781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,1,balanced,4.485653241475423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,1,power_law_1.01,5.605318450927735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,1,balanced,2.842677434285482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,1,balanced,5.145327885945638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,1,power_law_1.01,31.63341369628906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,1,power_law_1.01,10.335622406005859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,1,power_law_1.01,19.355673217773436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,1,balanced,2.4371946652730307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,1,balanced,4.871935844421387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,1,balanced,4.039701461791992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,1,balanced,5.421285629272461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,1,balanced,2.552997271219889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,1,balanced,7.387370427449544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,1,balanced,5.10148270924886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,1,balanced,7.999615987141927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,1,balanced,7.237701416015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,1,balanced,3.6234560012817383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,1,balanced,6.450554529825847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,1,balanced,11.333627065022787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,1,balanced,10.476544062296549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,1,balanced,4.912624041239421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,1,balanced,9.65341313680013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,1,balanced,15.952170054117838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,1,balanced,9.68838373819987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,1,balanced,19.847423553466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,1,balanced,18.676923116048176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,1,balanced,28.453941345214844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,1,power_law_1.2,0.14647040367126465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,1,power_law_1.2,0.0946943998336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,1,power_law_1.2,0.23808639049530028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,1,power_law_1.2,0.1444991946220398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,1,power_law_1.2,0.2833791971206665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,1,power_law_1.2,0.16499199867248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,1,power_law_1.2,0.5749567985534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,1,power_law_1.2,0.31498239040374754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,1,power_law_1.2,0.8712448120117188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,1,power_law_1.2,0.4893184185028076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,1,power_law_1.2,1.3066816329956055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,1,power_law_1.2,0.6576704025268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,1,power_law_1.2,2.023628807067871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,1,power_law_1.2,0.09886080026626587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,1,power_law_1.2,1.0546560287475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,1,power_law_1.2,0.04855040013790131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,1,power_law_1.2,2.327353668212891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,1,power_law_1.2,1.1479552268981934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,1,power_law_1.2,0.07445120215415954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,1,power_law_1.2,2.2992576599121093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,1,power_law_1.2,0.12760319709777831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,1,power_law_1.2,1.2231871604919433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,1,power_law_1.2,0.0726207971572876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,1,power_law_1.2,1.2324607849121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,1,power_law_1.2,2.431865692138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,1,power_law_1.2,0.13842560052871705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,1,power_law_1.2,0.1360640048980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,1,power_law_1.2,1.3111680030822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,1,power_law_1.2,2.522316741943359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,1,power_law_1.2,0.24184319972991944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,1,power_law_1.2,0.2261120080947876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,1,power_law_1.2,1.3489919662475587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,1,power_law_1.2,2.594278335571289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,1,power_law_1.2,0.33914880752563475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,1,power_law_1.2,0.33841919898986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,1,power_law_1.2,0.4924736022949219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,1,power_law_1.2,0.4665408134460449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,1,power_law_1.2,2.7051519393920898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,1,power_law_1.2,1.3915648460388184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,1,power_law_1.2,0.577561616897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,1,power_law_1.2,0.693235206604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,1,power_law_1.2,2.757785606384277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,1,power_law_1.2,0.6215551853179931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,1,power_law_1.2,0.7674431800842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,1,power_law_1.2,1.4723711967468263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,1,power_law_1.2,2.916249656677246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,1,power_law_1.2,0.6224768161773682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,1,power_law_1.2,0.8099967956542968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,1,power_law_1.2,1.5343104362487794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,1,power_law_1.2,3.010527992248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,1,power_law_1.2,0.6910783767700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,1,power_law_1.2,0.8248640060424804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,1,power_law_1.2,3.243577575683594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,1,power_law_1.2,1.5679743766784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,1,power_law_1.2,0.6927167892456054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,1,power_law_1.2,0.8755200386047364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,1,power_law_1.2,1.6915264129638672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,1,power_law_1.2,3.463411331176758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,1,power_law_1.2,0.7260992050170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,1,power_law_1.2,0.8789183616638183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,1,power_law_1.2,1.841926383972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,1,power_law_1.2,3.606719970703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,1,power_law_1.2,0.8433792114257812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,1,power_law_1.2,1.8919744491577148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,1,power_law_1.2,0.7723264217376709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,1,power_law_1.2,0.909932804107666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,1,power_law_1.2,4.038150405883789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,1,power_law_1.2,0.8447936058044434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,1,power_law_1.2,0.9384767532348632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,1,power_law_1.2,2.1271615982055665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,1,power_law_1.2,3.8823616027832033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,1,power_law_1.2,0.8833984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,1,power_law_1.2,2.1591680526733397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,1,power_law_1.2,4.421830368041992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,1,power_law_1.2,0.9048704147338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,1,power_law_1.2,0.9661824226379394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,1,power_law_1.2,2.465990447998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.2,0.9781503677368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,1,power_law_1.2,5.139814376831055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,1,power_law_1.2,1.0291263580322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,1,power_law_1.2,2.78143367767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.2,1.0112895965576172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,1,power_law_1.2,5.998054504394531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,1,power_law_1.2,1.060147190093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.2,1.1051456451416015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,1,power_law_1.2,3.4822593688964845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.2,2.6997888565063475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,1,power_law_1.2,1.150873565673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,1,power_law_1.2,6.843135833740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,1,power_law_1.2,4.094265747070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.2,3.2786624908447264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.2,1.2497535705566407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,1,power_law_1.2,5.290412902832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.2,1.9308736801147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,1,power_law_1.2,8.547769927978516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.2,2.1369535446166994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,1,power_law_1.2,6.551443481445313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.2,1.421337604522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,1,power_law_1.2,10.879776000976562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.2,2.703104019165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.2,1.5154879570007325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,1,power_law_1.2,7.755999755859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.2,3.304172897338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.2,1.8008832931518555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,1,power_law_1.2,12.690195465087891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,1,power_law_1.2,3.8438785552978514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.2,2.086227226257324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,1,power_law_1.2,12.106060791015626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.2,2.694611167907715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,1,power_law_1.2,19.24610595703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,1,power_law_1.2,5.731046295166015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.2,3.2608768463134767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,1,power_law_1.2,21.880889892578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,1,power_law_1.2,10.449343872070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.2,4.433491134643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,1,power_law_1.2,32.04033203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.2,5.696480178833008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,1,power_law_1.2,6.855136108398438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,1,power_law_1.2,10.649549102783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,1,power_law_1.2,19.500210571289063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,1,balanced,0.03976000100374222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,1,balanced,0.05007466673851013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,1,balanced,0.07478933533032735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,1,balanced,0.11922666430473328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,1,balanced,0.20547199249267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,1,balanced,0.2047626574834188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,1,balanced,0.20756800969441733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,1,balanced,0.20829866329828897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,1,balanced,0.20972800254821777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,1,balanced,0.20955200990041098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,1,balanced,0.21388266483942667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,1,balanced,0.21423999468485513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,1,balanced,0.21785066525141397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,1,balanced,0.2199839949607849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,1,balanced,0.22453866402308145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,1,balanced,0.23037334283192953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,1,balanced,0.2377600073814392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,1,balanced,0.03771200031042099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,1,balanced,0.05036266644795736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,1,balanced,0.2520586649576823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,1,balanced,0.039647998909155525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,1,balanced,0.01911466692884763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,1,balanced,0.05213866631189982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,1,balanced,0.27184534072875977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,1,balanced,0.05530133346716563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,1,balanced,0.02223466585079829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,1,balanced,0.0582826683918635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,1,balanced,0.08457600076993306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,1,balanced,0.3094613353411357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,1,balanced,0.025221332907676697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,1,balanced,0.1288746694723765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,1,balanced,0.08672533432642619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,1,balanced,0.02956266701221466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,1,balanced,0.13019200166066489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,1,balanced,0.12118400136629741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,1,balanced,0.34401599566141766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,1,balanced,0.04702933132648468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,1,balanced,0.1329919993877411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,1,balanced,0.1237600048383077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,1,balanced,0.04877866804599762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,1,balanced,0.13261866569519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,1,balanced,0.42474134763081867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,1,balanced,0.049813335140546165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,1,balanced,0.12902399897575378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,1,balanced,0.13457066814104715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,1,balanced,0.052101333936055504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,1,balanced,0.12686933080355325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,1,balanced,0.1374613344669342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,1,balanced,0.0537066658337911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,1,balanced,0.1283253331979116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,1,balanced,0.47467732429504395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,1,balanced,0.13983999689420065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,1,balanced,0.05600533386071523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,1,balanced,0.13168000181516012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,1,balanced,0.14453333616256714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,1,balanced,0.060266668597857155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,1,balanced,0.13246400157610574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,1,balanced,0.14507200320561728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,1,balanced,0.05782933533191681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,1,balanced,0.1388320028781891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,1,balanced,0.6599413156509399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,1,balanced,0.14867200454076132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,1,balanced,0.137882669766744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,1,balanced,0.05890133480230967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,1,balanced,0.15876266360282898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,1,balanced,0.14432533582051596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,1,balanced,0.06397866706053416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,1,balanced,0.16403200229008993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,1,balanced,0.15286933382352194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,1,balanced,0.06222933530807495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,1,balanced,0.17360534270604452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,1,balanced,0.15453867117563883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,1,balanced,0.800981362660726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,1,balanced,0.0642133355140686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,1,balanced,0.19416000445683798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,1,balanced,0.16617600123087564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,1,balanced,0.06820266445477803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,1,balanced,0.22037333250045776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,1,balanced,0.18889600038528442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,1,balanced,0.07420800129572551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,1,balanced,0.2680799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,1,balanced,0.21519466241200766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,1,balanced,0.080335999528567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,1,balanced,0.31381332874298096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,1,balanced,1.1553813616434734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,1,balanced,0.2585600018501282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,1,balanced,0.1350826621055603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,1,balanced,0.30449599027633667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,1,balanced,0.4161333243052165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,1,balanced,0.1564906636873881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,1,balanced,0.4326773484547933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,1,balanced,0.14247999588648477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,1,balanced,0.5113706588745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,1,power_law_1.01,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,1,balanced,0.5227466821670532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,1,balanced,0.1448746621608734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,1,power_law_1.01,0.039878401160240176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,1,power_law_1.01,0.06335359811782837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,1,balanced,1.503178596496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,1,balanced,0.7299573421478271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,1,power_law_1.01,0.08789119720458985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,1,power_law_1.01,0.11086080074310303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,1,balanced,0.762224038441976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,1,balanced,0.2558773358662923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,1,power_law_1.01,0.1729856014251709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,1,power_law_1.01,0.1803712010383606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,1,power_law_1.01,0.18442879915237426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,1,balanced,0.9223466714223226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,1,power_law_1.01,0.19466880559921265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,1,balanced,0.9603253205617269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,1,power_law_1.01,0.1958080053329468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,1,balanced,0.254095991452535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,1,power_law_1.01,0.19850879907608032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,1,power_law_1.01,0.20691840648651122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,1,power_law_1.01,0.20736639499664306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,1,balanced,1.8653333981831868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,1,power_law_1.01,0.21592319011688232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,1,balanced,1.344549338022868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,1,power_law_1.01,0.22442879676818847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,1,power_law_1.01,0.2298815965652466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,1,balanced,1.411087989807129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,1,balanced,0.3626026709874471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,1,power_law_1.01,0.23495039939880372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,1,power_law_1.01,0.2557440042495728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,1,power_law_1.01,0.2823359966278076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,1,power_law_1.01,0.32592000961303713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,1,power_law_1.01,0.3605247974395752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,1,balanced,1.7798293431599934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,1,power_law_1.01,0.441158390045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,1,balanced,1.8527092933654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,1,balanced,0.4734613498051961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,1,power_law_1.01,0.5308671951293945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,1,power_law_1.01,0.7061376094818115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,1,balanced,2.935333251953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,1,power_law_1.01,0.874835205078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,1,power_law_1.01,1.200972843170166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,1,power_law_1.01,1.5277119636535645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,1,balanced,2.215338706970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,1,power_law_1.01,1.8667583465576172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,1,balanced,2.3079093297322593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,1,power_law_1.01,2.878022384643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,1,balanced,0.5837813218434652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,1,power_law_1.01,0.03441919982433319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,1,power_law_1.01,0.04681600034236908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,1,power_law_1.01,0.03699840009212494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,1,power_law_1.01,5.606515121459961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,1,power_law_1.01,0.04642559885978699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,1,power_law_1.01,0.04746879935264588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,1,power_law_1.01,0.06469119787216186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,1,power_law_1.01,0.05456640124320984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,1,power_law_1.01,0.07984640002250672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,1,power_law_1.01,0.06769279837608337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,1,power_law_1.01,0.11219199895858764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,1,power_law_1.01,0.11989760398864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,1,power_law_1.01,0.08071680068969726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,1,power_law_1.01,0.12183040380477905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,1,power_law_1.01,0.1084671974182129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,1,power_law_1.01,0.1261888027191162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,1,power_law_1.01,0.019385600090026857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,1,power_law_1.01,0.12764159440994263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,1,power_law_1.01,0.11568000316619872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,1,power_law_1.01,0.1360640048980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,1,power_law_1.01,0.11813119649887086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,1,power_law_1.01,0.019782400131225585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,1,power_law_1.01,0.13952640295028687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,1,power_law_1.01,0.12101759910583496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,1,balanced,3.4130560557047525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,1,power_law_1.01,0.023507200181484222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,1,power_law_1.01,0.14261759519577027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,1,power_law_1.01,0.12216320037841796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,1,power_law_1.01,0.03149439990520477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,1,power_law_1.01,0.14673279523849486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,1,balanced,0.9181813398996989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,1,balanced,3.648655891418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,1,power_law_1.01,0.12744319438934326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,1,power_law_1.01,0.041555199027061465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,1,power_law_1.01,0.15684479475021362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,1,power_law_1.01,0.13093119859695435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,1,power_law_1.01,0.16484479904174804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,1,power_law_1.01,0.0542527973651886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,1,power_law_1.01,0.13582079410552977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,1,power_law_1.01,0.17644799947738649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,1,power_law_1.01,0.14143999814987182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,1,power_law_1.01,0.20300800800323487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,1,power_law_1.01,0.057011198997497556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,1,power_law_1.01,0.14926719665527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,1,power_law_1.01,0.22530560493469237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,1,power_law_1.01,0.06041600108146668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,1,power_law_1.01,0.15691519975662233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,1,power_law_1.01,0.2758271932601929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,1,power_law_1.01,0.17002880573272705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,1,power_law_1.01,0.061324799060821535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,1,power_law_1.01,0.32410240173339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,1,balanced,5.585482915242513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,1,power_law_1.01,0.20179200172424316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,1,power_law_1.01,0.06406400203704835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,1,power_law_1.01,0.43662080764770506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,1,power_law_1.01,0.2169856071472168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,1,power_law_1.01,0.06769919991493226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,1,power_law_1.01,0.5458687782287598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,1,power_law_1.01,0.2748863935470581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,1,power_law_1.01,0.7499328136444092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,1,power_law_1.01,0.3243391990661621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,1,power_law_1.01,0.06706560254096985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,1,power_law_1.01,0.9576512336730957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,1,power_law_1.01,0.44075517654418944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,1,power_law_1.01,1.3774144172668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,1,power_law_1.01,0.5551743984222413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,1,power_law_1.01,0.07138559818267823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,1,power_law_1.01,1.796268844604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,1,power_law_1.01,0.7793087959289551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,1,power_law_1.01,0.07192320227622986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,1,power_law_1.01,2.218239974975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,1,power_law_1.01,0.07338879704475403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,1,power_law_1.01,1.0000255584716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,1,power_law_1.01,0.0742143988609314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,1,power_law_1.01,3.3923198699951174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,1,power_law_1.01,1.4378432273864745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,1,power_law_1.01,0.0785535991191864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,1,power_law_1.01,1.8680128097534179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,1,power_law_1.01,6.624205017089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,1,power_law_1.01,0.08421760201454162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,1,power_law_1.01,2.3003583908081056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,1,power_law_1.01,0.09265279769897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,1,power_law_1.01,3.616435241699219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,1,power_law_1.01,0.18302079439163207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,1,balanced,6.665029525756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,1,power_law_1.01,0.21363840103149415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,1,power_law_1.01,7.106598663330078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,1,power_law_1.01,0.16569600105285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,1,balanced,1.8102933565775554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,1,power_law_1.01,0.1920639991760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,1,balanced,7.1025441487630205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,1,power_law_1.01,0.2454848051071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,1,power_law_1.01,0.2994688034057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,1,power_law_1.01,0.4114367961883545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,1,power_law_1.01,0.5166016101837159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,1,power_law_1.01,0.625657606124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,1,power_law_1.01,0.9622079849243164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,1,power_law_1.01,1.8649984359741212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,1,power_law_1.2,0.03518719971179962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,1,power_law_1.2,0.04038400053977966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,1,power_law_1.2,0.0554751992225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,1,power_law_1.2,0.07508479952812194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,1,power_law_1.2,0.10061440467834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,1,power_law_1.2,0.1768895983695984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,1,power_law_1.2,0.18574719429016112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,1,power_law_1.2,0.18709759712219237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,1,power_law_1.2,0.19466240406036378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,1,power_law_1.2,0.20048000812530517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,2,balanced,0.08463467160860698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,1,power_law_1.2,0.20182399749755858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,1,power_law_1.2,0.20929279327392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,2,balanced,0.13849066694577536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,1,power_law_1.2,0.21342079639434813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,2,balanced,0.2537386616071065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,1,power_law_1.2,0.21966080665588378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,2,balanced,0.25100799401601154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,1,power_law_1.2,0.22989439964294434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,2,balanced,0.24753600358963013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,1,power_law_1.2,0.2363136053085327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,2,balanced,0.2439039945602417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,1,power_law_1.2,0.23854079246520996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,2,balanced,0.24701333045959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,1,power_law_1.2,0.2638015985488892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,2,balanced,0.24602667490641275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,1,power_law_1.2,0.2881792068481445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,2,balanced,0.2463573416074117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,1,power_law_1.2,0.3337408065795898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,2,balanced,0.062218666076660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,2,balanced,0.24631466468175253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,1,power_law_1.2,0.03397760093212128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,1,power_law_1.2,0.36727681159973147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,2,balanced,0.08727999528249104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,2,balanced,0.25801066557566327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,2,balanced,0.08805333574612935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,1,power_law_1.2,0.45450878143310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,1,power_law_1.2,0.036985599994659425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,2,balanced,0.13571733236312866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,2,balanced,0.25545599063237506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,1,power_law_1.2,0.5359424114227295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,2,balanced,0.14046399792035422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,2,balanced,0.12144000331560771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,2,balanced,0.25938665866851807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,1,power_law_1.2,0.04471679925918579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,1,power_law_1.2,0.7093183994293213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,2,balanced,0.13913066188494363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,2,balanced,0.19058666626612344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,2,balanced,0.14112533132235208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,1,power_law_1.2,0.8849599838256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,2,balanced,0.2644960085550944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,1,power_law_1.2,0.06044160127639771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,2,balanced,0.14355732997258505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,2,balanced,0.18735466400782266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,1,power_law_1.2,1.2134271621704102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,2,balanced,0.28521599372227985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,2,balanced,0.14421332875887552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,1,power_law_1.2,0.07384960055351257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,1,power_law_1.2,0.04708479940891266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,2,balanced,0.19422399997711182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,1,power_law_1.2,1.552019214630127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,2,balanced,0.1455893317858378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,2,balanced,0.2911093235015869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,2,balanced,0.1828426718711853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,1,power_law_1.2,1.8855680465698241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,1,power_law_1.2,0.11240320205688477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,2,balanced,0.14696533481280008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,2,balanced,0.19800533850987753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,2,balanced,0.2956586678822835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,2,balanced,0.14900267124176025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,1,power_law_1.2,0.047993600368499756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,1,power_law_1.2,2.8933759689331056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,1,power_law_1.2,0.12061439752578736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,2,balanced,0.15150933464368185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,2,balanced,0.19834667444229126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,1,power_law_1.2,0.05389440059661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,2,balanced,0.3458026647567749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,1,power_law_1.2,0.12300159931182861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,2,balanced,0.18736000855763754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,2,balanced,0.1509866714477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,1,power_law_1.2,0.066348797082901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,1,power_law_1.2,5.6395263671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,2,balanced,0.3646719853083293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,2,balanced,0.19058134158452353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,2,balanced,0.15478932857513428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,1,power_law_1.2,0.07779200077056884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,1,power_law_1.2,0.12798080444335938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,2,balanced,0.18491200606028238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,2,balanced,0.17563199996948242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,1,power_law_1.2,0.10819200277328492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,2,balanced,0.4542986551920573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,2,balanced,0.1934773325920105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,1,power_law_1.2,0.1309440016746521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,2,balanced,0.1824586590131124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,1,power_law_1.2,0.1164736032485962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,2,balanced,0.5367679993311564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,2,balanced,0.1804693341255188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,1,power_law_1.2,0.12037119865417481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,2,balanced,0.18760534127553305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,1,power_law_1.2,0.1369472026824951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,2,balanced,0.1801813244819641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,2,balanced,0.03350933392842611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,1,power_law_1.2,0.12282880544662475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,2,balanced,0.21954667568206787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,2,balanced,0.7328320344289144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,1,power_law_1.2,0.14097919464111328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,2,balanced,0.15239999691645303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,1,power_law_1.2,0.12435840368270874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,2,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,2,balanced,0.23225067059199014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,1,power_law_1.2,0.14673279523849486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,1,power_law_1.2,0.1293503999710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,1,power_law_1.2,0.018265600502490997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,2,balanced,0.9064640204111735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,2,balanced,0.03139200061559677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,1,power_law_1.2,0.13273600339889527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,2,balanced,0.29334932565689087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,1,power_law_1.2,0.14913920164108277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,2,balanced,0.15466133753458658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,1,power_law_1.2,0.1390015959739685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,2,balanced,0.033471999069054924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,1,power_law_1.2,0.020287999510765077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,1,power_law_1.2,0.16047999858856202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,2,balanced,0.34569064776102704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,1,power_law_1.2,0.14244480133056642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,2,balanced,1.3125440279642742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,1,power_law_1.2,0.16560640335083007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,2,balanced,0.03435733417669932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,1,power_law_1.2,0.15278719663619994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,2,balanced,0.14434132973353067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,1,power_law_1.2,0.024403199553489685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,2,balanced,0.4767306645711263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,1,power_law_1.2,0.15950080156326293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,1,power_law_1.2,0.17838720083236695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,2,balanced,0.039503999054431915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,2,balanced,0.27577600876490277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,1,power_law_1.2,0.02975359857082367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,1,power_law_1.2,0.17317119836807252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,2,balanced,1.6692479451497395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,2,balanced,0.045663997530937195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,2,balanced,0.5913600126902262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,2,balanced,0.22313600778579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,1,power_law_1.2,0.03761920034885406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,1,power_law_1.2,0.2040640115737915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,1,power_law_1.2,0.20611839294433593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,2,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,2,balanced,0.2958986759185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,2,balanced,0.8542346954345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,1,power_law_1.2,0.2220479965209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,1,power_law_1.2,0.055103999376296994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,2,balanced,0.36931200822194415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,2,balanced,2.399242719014486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,2,balanced,0.06397333244482677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,1,power_law_1.2,0.27509119510650637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,1,power_law_1.2,0.22838399410247803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,1,power_law_1.2,0.05847679972648621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,1,power_law_1.2,0.330297589302063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,2,balanced,0.5201226472854614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,2,balanced,0.05258133510748545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,2,balanced,1.119322697321574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,1,power_law_1.2,0.44099202156066897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,1,power_law_1.2,0.06257920265197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,1,power_law_1.2,0.27911040782928465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,2,balanced,0.05874133110046387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,1,power_law_1.2,0.5598336219787597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,1,power_law_1.2,0.06231039762496948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,1,power_law_1.2,0.3312959909439087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,2,balanced,0.6694453557332357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,2,balanced,0.07044800122578938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,2,balanced,3.3134988149007163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,1,power_law_1.2,0.7796224117279053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,1,power_law_1.2,0.44008960723876955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,2,balanced,1.6485279401143391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,1,power_law_1.2,0.06529279947280883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,2,balanced,0.07770666480064392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,2,balanced,0.9673706690470377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,1,power_law_1.2,0.9978560447692871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,1,power_law_1.2,1.4335871696472169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,2,balanced,0.07428266604741414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,1,power_law_1.2,0.06825600266456604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,1,power_law_1.2,0.552396821975708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,2,balanced,0.10915199915568034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,1,power_law_1.2,1.8621696472167968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,2,balanced,4.14632002512614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,2,balanced,1.267802635828654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,1,power_law_1.2,0.06758400201797485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,2,balanced,2.134069283803304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,2,balanced,0.08520000179608662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,1,power_law_1.2,2.2925439834594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,1,power_law_1.2,0.7569536209106446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,2,balanced,0.09955199559529622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,1,power_law_1.2,0.06785280108451844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,1,power_law_1.2,3.623980712890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,2,balanced,0.1402186652024587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,1,power_law_1.2,0.07351679801940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,1,power_law_1.2,0.9596927642822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,2,balanced,0.16671999295552573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,2,balanced,1.8672480583190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,1,power_law_1.2,0.07205119729042053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,1,power_law_1.2,7.122220611572265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,2,balanced,0.233189324537913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,2,balanced,2.653301397959391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,1,power_law_1.2,0.07338879704475403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,1,power_law_1.2,1.3817343711853027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,2,balanced,0.2958666682243347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,1,power_law_1.2,0.07991679906845092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,2,balanced,6.57966423034668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,2,balanced,2.464992046356201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,2,balanced,0.42230931917826336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,1,power_law_1.2,0.08526080250740051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,1,power_law_1.2,1.7980159759521483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,1,power_law_1.2,0.09418879747390747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,2,balanced,0.5306026538213094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,1,power_law_1.2,0.186080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,1,power_law_1.2,2.236454391479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,1,power_law_1.2,0.21377279758453369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,2,balanced,4.229898770650228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,2,balanced,0.7863840262095133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,1,power_law_1.2,0.1693120002746582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,2,balanced,3.0504961013793945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,1,power_law_1.2,3.392287826538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,1,power_law_1.2,0.19528959989547728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,2,balanced,1.0255146821339924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,1,power_law_1.2,0.24990720748901368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,1,power_law_1.2,0.3025343894958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,1,power_law_1.2,6.611904144287109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,1,power_law_1.2,0.41495041847229003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,2,balanced,1.5185653368632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,1,power_law_1.2,0.5228415966033936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,2,balanced,13.073674519856771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,1,power_law_1.2,0.6392640113830567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,2,balanced,4.843269348144531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,1,power_law_1.2,0.9696703910827636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,2,balanced,2.0113706588745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,2,balanced,8.200447718302408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,1,power_law_1.2,1.8633535385131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,2,balanced,2.4935146967569985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,2,balanced,9.60974375406901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,2,balanced,3.949909210205078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,2,balanced,7.89686902364095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,2,power_law_1.01,0.10305919647216796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,2,power_law_1.01,0.13665920495986938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,2,power_law_1.01,0.18995840549468995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,2,power_law_1.01,0.24944000244140624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,2,power_law_1.01,0.24583039283752442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,2,power_law_1.01,0.24369280338287352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,2,power_law_1.01,0.25614080429077146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,2,power_law_1.01,0.2689663887023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,2,power_law_1.01,0.2506688117980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,2,power_law_1.01,0.2710848093032837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,2,power_law_1.01,0.2731519937515259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,2,power_law_1.01,0.2780224084854126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,2,power_law_1.01,0.2825151920318604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,2,power_law_1.01,0.3133120059967041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,2,power_law_1.01,0.3333823919296265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,2,power_law_1.01,0.3611135959625244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,2,power_law_1.01,0.39694719314575194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,2,power_law_1.01,0.4214655876159668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,2,power_law_1.01,0.5281919956207275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,2,power_law_1.01,0.6148928165435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,2,power_law_1.01,0.7649407863616944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,2,power_law_1.01,1.029587173461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,2,power_law_1.01,1.372326374053955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,2,power_law_1.01,1.8328512191772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,2,power_law_1.01,2.442963218688965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,2,power_law_1.01,3.625689697265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,2,power_law_1.01,4.807040023803711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,2,power_law_1.01,5.691731262207031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,2,power_law_1.01,9.849478149414063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,2,power_law_1.01,21.468838500976563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,2,power_law_1.01,0.08105599880218506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,2,power_law_1.01,0.12145919799804687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,2,power_law_1.01,0.08160640001296997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,2,power_law_1.01,0.03022719919681549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,2,power_law_1.01,0.09290239810943604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,2,power_law_1.2,0.11365760564804077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,2,power_law_1.2,0.07489280104637146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,2,power_law_1.2,0.11967359781265259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,2,power_law_1.01,0.1379647970199585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,2,power_law_1.01,0.13512320518493653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,2,power_law_1.01,0.029791998863220214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,2,power_law_1.2,0.030393600463867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,2,power_law_1.01,0.14020479917526246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,2,power_law_1.01,0.1547968029975891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,2,power_law_1.2,0.12679680585861205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,2,power_law_1.2,0.12761600017547609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,2,power_law_1.2,0.08983680009841918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,2,power_law_1.01,0.14620800018310548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,2,power_law_1.01,0.029267200827598573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,2,power_law_1.2,0.02969599962234497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,2,power_law_1.01,0.1933184027671814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,2,power_law_1.2,0.15029759407043458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,2,power_law_1.2,0.10268800258636475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,2,power_law_1.2,0.12328959703445434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,2,power_law_1.01,0.1500223994255066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,2,power_law_1.2,0.029228800535202028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,2,power_law_1.01,0.19560960531234742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,2,power_law_1.2,0.032620799541473386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,2,power_law_1.2,0.24407680034637452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,2,power_law_1.01,0.03327359855175018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,2,power_law_1.2,0.13755520582199096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,2,power_law_1.01,0.14907519817352294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,2,power_law_1.2,0.1875391960144043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,2,power_law_1.2,0.03341439962387085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,2,power_law_1.01,0.19771519899368287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,2,power_law_1.01,0.16126079559326173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,2,power_law_1.2,0.2426624059677124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,2,power_law_1.2,0.03865599930286408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,2,power_law_1.2,0.1388543963432312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,2,power_law_1.2,0.19470720291137694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,2,power_law_1.01,0.0335999995470047
TRTLLM,1.2.0rc5,NVIDIA B200,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 B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,2,power_law_1.01,0.18524800539016723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,2,power_law_1.01,0.16366080045700074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,2,power_law_1.2,0.14466559886932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,2,power_law_1.2,0.19067519903182983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,2,power_law_1.2,0.04445439875125885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,2,power_law_1.2,0.24149758815765382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,2,power_law_1.01,0.15713920593261718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,2,power_law_1.01,0.1874176025390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,2,power_law_1.2,0.06208639740943909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,2,power_law_1.01,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,2,power_law_1.2,0.1801151990890503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,2,power_law_1.2,0.1491520047187805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,2,power_law_1.2,0.26282880306243894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,2,power_law_1.01,0.16031999588012696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,2,power_law_1.01,0.1902143955230713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,2,power_law_1.2,0.050329601764678954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,2,power_law_1.2,0.18762240409851075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,2,power_law_1.2,0.05620480179786682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,2,power_law_1.2,0.15018880367279053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,2,power_law_1.2,0.2639552116394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,2,power_law_1.01,0.04374400079250336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,2,power_law_1.01,0.1636288046836853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,2,power_law_1.01,0.18577280044555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,2,power_law_1.2,0.0699455976486206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,2,power_law_1.2,0.1897663950920105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,2,power_law_1.2,0.15575040578842164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,2,power_law_1.2,0.27497599124908445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,2,power_law_1.01,0.04506880044937134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,2,power_law_1.01,0.18584959506988524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,2,power_law_1.01,0.18150399923324584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,2,power_law_1.2,0.07505919933319091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,2,power_law_1.2,0.18525439500808716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,2,power_law_1.2,0.07370240092277527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,2,power_law_1.2,0.1603775978088379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,2,power_law_1.2,0.10438400506973267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,2,power_law_1.2,0.2685120105743408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,2,power_law_1.2,0.18253439664840698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,2,power_law_1.01,0.06244480013847351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,2,power_law_1.01,0.18696320056915283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,2,power_law_1.2,0.15943039655685426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,2,power_law_1.01,0.20101759433746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,2,power_law_1.2,0.08354560136795045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,2,power_law_1.2,0.2006335973739624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,2,power_law_1.2,0.27234559059143065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,2,power_law_1.2,0.09784960150718688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,2,power_law_1.2,0.15598080158233643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,2,power_law_1.2,0.20631680488586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,2,power_law_1.01,0.050944000482559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,2,power_law_1.01,0.2115391969680786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,2,power_law_1.2,0.1377344012260437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,2,power_law_1.01,0.21063039302825928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,2,power_law_1.2,0.1931712031364441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,2,power_law_1.2,0.2803263902664185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.2,0.16244479417800903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,2,power_law_1.2,0.1641535997390747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,2,power_law_1.01,0.05685120224952698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,2,power_law_1.2,0.22815999984741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,2,power_law_1.01,0.24355840682983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.2,0.2302720069885254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,2,power_law_1.01,0.20470399856567384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,2,power_law_1.2,0.18565759658813477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,2,power_law_1.2,0.28142719268798827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.2,0.29217278957366943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,2,power_law_1.2,0.23597440719604493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,2,power_law_1.01,0.07107840180397033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,2,power_law_1.01,0.2711679935455322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,2,power_law_1.01,0.22163839340209962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.2,0.4180863857269287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,2,power_law_1.2,0.26670079231262206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,2,power_law_1.2,0.20188798904418945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,2,power_law_1.2,0.3243200063705444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.2,0.5298816204071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,2,power_law_1.2,0.33157761096954347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,2,power_law_1.01,0.0751039981842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,2,power_law_1.01,0.30643200874328613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,2,power_law_1.01,0.2446592092514038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.2,0.7848703861236572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,2,power_law_1.2,0.21935999393463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,2,power_law_1.2,0.32826240062713624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.2,0.3449343919754028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.2,1.0242752075195312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,2,power_law_1.01,0.25287680625915526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,2,power_law_1.01,0.44219517707824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.2,0.5029568195343017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,2,power_law_1.01,0.07374079823493958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,2,power_law_1.2,0.24099841117858886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.2,1.5157055854797363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,2,power_law_1.2,0.37624320983886717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.2,0.625216007232666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.2,2.0102720260620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,2,power_law_1.2,0.282092809677124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,2,power_law_1.01,0.31841919422149656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,2,power_law_1.01,0.48263039588928225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,2,power_law_1.01,0.10635520219802856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,2,power_law_1.2,0.3857088088989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.2,0.8648063659667968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,2,power_law_1.2,2.492755126953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,2,power_law_1.2,0.33584640026092527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.01,0.3691648006439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,2,power_law_1.01,0.7058688163757324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,2,power_law_1.01,0.0840831995010376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,2,power_law_1.2,0.46058239936828616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,2,power_law_1.2,3.9494911193847657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.2,1.0858816146850585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,2,power_law_1.01,0.9497344017028808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,2,power_law_1.2,0.44549760818481443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.01,0.5091648101806641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,2,power_law_1.01,0.09880319833755494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,2,power_law_1.01,1.2129856109619142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,2,power_law_1.01,0.1388479948043823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,2,power_law_1.2,0.4690559864044189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,2,power_law_1.2,0.5539135932922363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,2,power_law_1.2,8.499622344970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.01,0.5495999813079834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.2,1.3962688446044922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.01,0.163264000415802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,2,power_law_1.01,1.6770559310913087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,2,power_law_1.2,0.6598207950592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,2,power_law_1.2,0.6855423927307129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.01,0.8392895698547364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.2,2.2833471298217773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.01,0.2306368112564087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,2,power_law_1.2,0.9488448143005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,2,power_law_1.01,2.6159040451049806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,2,power_law_1.2,0.739404821395874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.01,0.9810175895690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.2,2.813651275634766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.01,0.2924544095993042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.01,1.4895999908447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,2,power_law_1.2,1.4769536018371583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,2,power_law_1.2,0.9634048461914062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,2,power_law_1.01,3.1934272766113283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.2,4.418815994262696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.01,1.9122943878173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.01,0.41861758232116697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,2,power_law_1.2,1.8208383560180663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,2,power_law_1.2,1.5431615829467773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,2,power_law_1.01,4.075193786621094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.01,2.411334419250488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.01,0.5301631927490235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,2,power_law_1.2,4.906867218017578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,2,power_law_1.2,2.5267520904541017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,2,power_law_1.2,2.203116798400879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,2,power_law_1.01,6.135948944091797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.01,3.405401611328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.01,0.7831295967102051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,2,power_law_1.2,3.1777856826782225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,2,power_law_1.2,2.8818368911743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,2,power_law_1.2,8.487935638427734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.01,1.0223872184753418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,2,power_law_1.01,4.769055938720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,2,power_law_1.2,4.351839828491211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,2,power_law_1.2,4.706732940673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,2,power_law_1.01,12.947386169433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.01,1.5182271957397462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,2,power_law_1.2,15.819891357421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,2,power_law_1.01,6.3267265319824215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,2,power_law_1.2,5.673977661132812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,2,power_law_1.2,7.134156799316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.01,2.0090047836303713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,2,power_law_1.2,7.211788940429687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,2,power_law_1.01,2.4988544464111326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,2,power_law_1.01,15.612850952148438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,2,power_law_1.2,13.33623046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,2,power_law_1.2,11.367967987060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,2,power_law_1.01,3.9500545501708983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,2,power_law_1.01,7.844608306884766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,2,power_law_1.2,20.784031677246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,1,balanced,0.082997332016627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,1,balanced,0.13668266932169595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,1,balanced,0.24280534187952676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,1,balanced,0.23770666122436523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,1,balanced,0.24076267083485922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,1,balanced,0.2413439949353536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,1,balanced,0.243231991926829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,1,balanced,0.24445333083470663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,1,balanced,0.24443199237187704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,1,balanced,0.24553066492080688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,1,balanced,0.2507466673851013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,1,balanced,0.2501973311106364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,1,balanced,0.252895991007487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,1,balanced,0.25986133019129437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,1,balanced,0.28866666555404663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,1,balanced,0.292192002137502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,1,balanced,0.2927520076433818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,1,balanced,0.354751984278361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,1,balanced,0.3690933386484782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,1,balanced,0.4522240161895752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,1,balanced,0.5240960121154785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,1,balanced,0.7183252970377604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,1,balanced,0.878111998240153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,1,balanced,1.2794453303019206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,1,balanced,1.6557440757751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,1,balanced,2.4399147033691406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,1,balanced,3.2862507502237954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,1,balanced,4.071530659993489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,1,balanced,6.48690668741862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,1,balanced,0.05885866781075796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,1,balanced,0.0499893327554067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,1,balanced,0.08390933275222778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,1,balanced,0.13180266817410788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,1,balanced,0.1376479963461558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,1,balanced,0.07156800230344136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,1,balanced,0.13672000169754028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,1,balanced,0.0999893347422282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,1,balanced,0.13979732990264893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,1,balanced,0.14120533068974814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,1,balanced,0.10415466626485188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,1,balanced,0.14226667086283365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,1,balanced,13.151776631673178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,1,balanced,0.1434986690680186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,1,balanced,0.10365866621335347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,1,balanced,0.14474667112032572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,1,balanced,0.10362133383750916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,1,balanced,0.14803199966748556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,1,balanced,0.15235199530919394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,1,balanced,0.10616000493367513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,1,balanced,0.15128533045450845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,1,balanced,0.1053600013256073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,1,balanced,0.15818132956822714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,1,balanced,0.10602133472760518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,1,balanced,0.17636799812316895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,1,balanced,0.10803733269373576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,1,balanced,0.1766080061594645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,1,balanced,0.10929600397745769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,1,balanced,0.1821333368619283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,1,balanced,0.2224000096321106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,1,balanced,0.11275733510653178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,1,balanced,0.23190933465957642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,1,balanced,0.11352533102035522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,1,balanced,0.11860799789428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,1,balanced,0.3097333312034607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,1,balanced,0.12794666488965353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,1,balanced,0.3435200055440267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,1,balanced,0.130922665198644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,1,balanced,0.47461867332458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,1,balanced,0.13944533467292786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,1,balanced,0.1677280068397522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,1,balanced,0.5905866622924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,1,balanced,0.18686934312184653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,1,balanced,0.023402666052182514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,1,balanced,0.23206400871276855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,1,balanced,0.8580906391143799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,1,power_law_1.01,0.057894402742385866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,1,balanced,0.02316266546646754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,1,power_law_1.01,0.07946239709854126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,1,balanced,0.2826240062713623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,1,balanced,0.024847999215126038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,1,power_law_1.01,0.07067520022392274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,1,balanced,0.0272533322374026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,1,balanced,0.38126933574676514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,1,power_law_1.01,0.11089279651641845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,1,balanced,1.092133363087972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,1,power_law_1.01,0.08061439990997314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,1,balanced,0.031370667119820915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,1,balanced,0.4837973515192668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,1,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,1,power_law_1.01,0.13544319868087767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,1,power_law_1.01,0.144595205783844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,1,power_law_1.01,0.05061119794845581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,1,balanced,0.03874133278926214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,1,power_law_1.01,0.13379199504852296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,1,balanced,0.6738986968994141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,1,balanced,0.043280000487963356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,1,power_law_1.01,0.23804159164428712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,1,power_law_1.01,0.13808640241622924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,1,power_law_1.01,0.06650879979133606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,1,balanced,1.6034453709920247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,1,power_law_1.01,0.023647999763488768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,1,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,1,power_law_1.01,0.13763200044631957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,1,power_law_1.01,0.07043200135231018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,1,power_law_1.01,0.23875200748443604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,1,balanced,0.040394666294256844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,1,power_law_1.01,0.13924479484558105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,1,power_law_1.01,0.09789440035820007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,1,power_law_1.01,0.022188800573349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,1,balanced,0.8738186359405518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,1,power_law_1.01,0.1013375997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,1,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,1,power_law_1.01,0.14470399618148805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,1,power_law_1.01,0.23912959098815917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,1,power_law_1.01,0.10423040390014648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,1,balanced,0.04359999795754751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,1,power_law_1.01,0.15011839866638182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,1,power_law_1.01,0.24186239242553711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,1,power_law_1.01,0.023571200668811798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,1,power_law_1.01,0.10409599542617798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,1,balanced,2.1104532877604165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,1,balanced,0.04595733185609182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,1,power_law_1.01,0.14461439847946167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,1,power_law_1.01,0.10479999780654907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,1,power_law_1.01,0.249619197845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,1,power_law_1.01,0.025894400477409363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,1,power_law_1.01,0.10536960363388062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,1,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,1,power_law_1.01,0.14967039823532105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,1,balanced,1.2416853109995525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,1,power_law_1.01,0.24682879447937012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,1,power_law_1.01,0.1068160057067871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,1,balanced,0.10097066561381023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,1,power_law_1.01,0.15816320180892945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,1,power_law_1.01,0.030694401264190672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,1,power_law_1.01,0.10981760025024415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,1,power_law_1.01,0.25639679431915285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,1,balanced,0.11760532855987549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,1,power_law_1.01,0.1696128010749817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,1,power_law_1.01,0.11274880170822144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,1,power_law_1.01,0.04246399998664856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,1,power_law_1.01,0.2630975961685181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,1,power_law_1.01,0.18153599500656128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,1,power_law_1.01,0.11701120138168335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,1,balanced,0.1474293371041616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,1,balanced,2.62009604771932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,1,power_law_1.01,0.11980160474777221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,1,power_law_1.01,0.26777598857879636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,1,power_law_1.01,0.1896512031555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,1,power_law_1.01,0.03706879913806915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,1,balanced,0.08424533406893413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,1,power_law_1.01,0.13496320247650145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,1,balanced,1.6409494082132976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,1,power_law_1.01,0.2697792053222656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,1,power_law_1.01,0.1964735984802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,1,power_law_1.01,0.1384511947631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,1,power_law_1.01,0.04151679873466492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,1,balanced,0.10081066687901814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,1,power_law_1.01,0.23087360858917236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,1,power_law_1.01,0.14886399507522582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,1,power_law_1.01,0.2901439905166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,1,power_law_1.01,0.03635840117931366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,1,balanced,0.13945066928863525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,1,power_law_1.01,0.17111680507659913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,1,power_law_1.01,0.26618878841400145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,1,power_law_1.01,0.303654408454895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,1,power_law_1.01,0.03948160111904144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.01,0.19794559478759766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,1,balanced,0.17305066188176474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,1,power_law_1.01,0.3289344072341919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.01,0.24741120338439943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,1,balanced,2.044010639190674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,1,power_law_1.01,0.39246718883514403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,1,power_law_1.01,0.31313281059265136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,1,power_law_1.01,0.044710400700569156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.01,0.29909119606018064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,1,balanced,0.23965867360432944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,1,balanced,4.119850794474284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,1,power_law_1.01,0.041868799924850465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,1,power_law_1.01,0.3329983949661255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.01,0.3986815929412842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,1,power_law_1.01,0.04496639966964722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,1,balanced,0.300981342792511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.01,0.5022336006164551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,1,power_law_1.01,0.37518720626831054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,1,power_law_1.01,0.5331520080566406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,1,power_law_1.01,0.050944000482559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.01,0.6988671779632568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,1,power_law_1.01,0.42356481552124026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.01,0.8945152282714843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,1,balanced,0.43757331371307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,1,power_law_1.01,0.10026880502700805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,1,power_law_1.01,0.5309631824493408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,1,power_law_1.01,0.6478079795837403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.01,1.3194047927856445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,1,power_law_1.01,0.6044159889221191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,1,power_law_1.01,0.11615359783172607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,1,balanced,3.2008533477783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.01,1.7068096160888673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,1,power_law_1.01,0.9109696388244629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,1,power_law_1.01,0.8165311813354492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,1,power_law_1.01,0.14650239944458007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.01,2.141414451599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,1,balanced,0.564629316329956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,1,power_law_1.01,0.0835968017578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,1,power_law_1.01,0.9944000244140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,1,power_law_1.01,1.1475520133972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,1,power_law_1.01,3.404313659667969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.01,0.09812480211257935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,1,power_law_1.01,1.401689624786377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,1,power_law_1.01,1.6691904067993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,1,power_law_1.01,6.610221099853516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,1,balanced,0.8318186601003011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.01,0.13696639537811278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,1,power_law_1.01,1.7743616104125977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,1,power_law_1.01,2.1656639099121096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,1,balanced,8.21504020690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.01,0.17100800275802613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,1,power_law_1.01,2.617689514160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.01,0.23916800022125245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,1,power_law_1.01,2.6565120697021483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,1,balanced,1.0933653513590496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,1,power_law_1.01,3.435424041748047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.01,0.2987967967987061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,1,power_law_1.01,4.200876617431641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.01,0.43683199882507323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,1,power_law_1.01,4.239929580688477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.01,0.5652927875518798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,1,balanced,6.571055730183919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,1,power_law_1.01,8.270553588867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,1,power_law_1.01,6.6733245849609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.01,0.8300928115844727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,1,balanced,1.3578826586405437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.01,1.0944448471069337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,1,power_law_1.01,13.337612915039063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.01,1.35863037109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,1,power_law_1.01,2.1581184387207033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,1,balanced,2.148522694905599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,1,power_law_1.01,4.260428619384766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,1,balanced,4.257178624471028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,1,power_law_1.2,0.0771776020526886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,1,power_law_1.2,0.11055359840393067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,1,power_law_1.2,0.13980799913406372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,1,power_law_1.2,0.23384320735931396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,1,power_law_1.2,0.23878400325775145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,1,power_law_1.2,0.23968639373779296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,1,power_law_1.2,0.24320640563964843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,1,power_law_1.2,0.24503679275512696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,1,power_law_1.2,0.2520319938659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,1,power_law_1.2,0.25207679271697997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,1,power_law_1.2,0.2619647979736328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,1,power_law_1.2,0.27037439346313474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,1,power_law_1.2,0.2717184066772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,1,power_law_1.2,0.2896384000778198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,1,power_law_1.2,0.3053440093994141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,1,power_law_1.2,0.31775360107421874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,1,power_law_1.2,0.3233855962753296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,1,power_law_1.2,0.39470720291137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,1,power_law_1.2,0.42490878105163576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,1,power_law_1.2,0.5284416198730468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,1,power_law_1.2,0.6105792045593261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,1,power_law_1.2,0.8142271995544433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,1,power_law_1.2,0.9857664108276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,1,power_law_1.2,1.4067328453063965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,1,power_law_1.2,1.7949504852294922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,1,power_law_1.2,2.6140480041503906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,1,power_law_1.2,3.448409652709961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,1,power_law_1.2,4.210540771484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,1,power_law_1.2,6.661695861816407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,1,power_law_1.2,13.278233337402344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,2,balanced,0.1306666632493337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,1,power_law_1.2,0.0226623997092247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,1,power_law_1.2,0.05237119793891907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,1,power_law_1.2,0.05729920268058777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,2,balanced,0.21149333318074545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,2,balanced,0.3861546516418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,2,balanced,0.07566933333873749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,2,balanced,0.08162666857242584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,1,power_law_1.2,0.07024000287055969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,2,balanced,0.38513068358103436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,1,power_law_1.2,0.022259199619293214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,1,power_law_1.2,0.06354560256004334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,2,balanced,0.38945066928863525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,1,power_law_1.2,0.08098559975624084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,1,power_law_1.2,0.023321600258350374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,2,balanced,0.12771733601888022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,1,power_law_1.2,0.06910079717636108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,2,balanced,0.09876267115275066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,2,balanced,0.3851360082626343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,2,balanced,0.2137440045674642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,2,balanced,0.14827199776967367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,1,power_law_1.2,0.026060798764228822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,1,power_law_1.2,0.09889280200004577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,1,power_law_1.2,0.13392640352249147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,2,balanced,0.3892800013224284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,2,balanced,0.2172586719195048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,2,balanced,0.15291733543078104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,1,power_law_1.2,0.030508801341056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,2,balanced,0.3891040086746216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,1,power_law_1.2,0.10287359952926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,2,balanced,0.2155146598815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,2,balanced,0.1529759963353475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,1,power_law_1.2,0.042387199401855466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,1,power_law_1.2,0.13697279691696168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,2,balanced,0.38980265458424884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,2,balanced,0.21492266654968262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,2,balanced,0.15413866440455118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,1,power_law_1.2,0.10408960580825806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,2,balanced,0.15686399737993875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,2,balanced,0.3917119900385539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,2,balanced,0.21727466583251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,1,power_law_1.2,0.03733760118484497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,1,power_law_1.2,0.10380159616470337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,1,power_law_1.2,0.13863680362701417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,2,balanced,0.2203893264134725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,2,balanced,0.39822399616241455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,2,balanced,0.15482133626937866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,1,power_law_1.2,0.041075199842453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,1,power_law_1.2,0.10492160320281982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,2,balanced,0.04811733464399973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,2,balanced,0.21836799383163452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,2,balanced,0.1555359959602356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,2,balanced,0.3956426779429118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,1,power_law_1.2,0.1403839945793152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,1,power_law_1.2,0.036108800768852235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,2,balanced,0.22002132733662924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,1,power_law_1.2,0.10565760135650634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,2,balanced,0.157669335603714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,2,balanced,0.3973013162612915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,1,power_law_1.2,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,2,balanced,0.0496373325586319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,1,power_law_1.2,0.14314240217208862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,2,balanced,0.22606400648752847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,2,balanced,0.15818666418393454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,1,power_law_1.2,0.10631680488586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,2,balanced,0.4385226567586263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,1,power_law_1.2,0.04428800046443939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,2,balanced,0.22513065735499063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,2,balanced,0.16055466731389365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,1,power_law_1.2,0.146451199054718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,2,balanced,0.04946133494377136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,1,power_law_1.2,0.10831359624862671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,2,balanced,0.1595146656036377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,2,balanced,0.22618667284647623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,1,power_law_1.2,0.04238080084323883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,2,balanced,0.4291146596272786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,1,power_law_1.2,0.14947839975357055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,2,balanced,0.168122669061025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,2,balanced,0.24447466929753622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,1,power_law_1.2,0.04460160136222839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,2,balanced,0.0592853327592214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,1,power_law_1.2,0.11266560554504394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,2,balanced,0.43616533279418945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,1,power_law_1.2,0.050547200441360476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,2,balanced,0.17876799901326498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,2,balanced,0.2561546762784322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,1,power_law_1.2,0.11520639657974244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,2,balanced,0.44491732120513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,1,power_law_1.2,0.14510719776153563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,2,balanced,0.09308266639709473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,2,balanced,0.25763734181722003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,2,balanced,0.18077333768208823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,1,power_law_1.2,0.09999359846115112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,2,balanced,0.5386879841486613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,1,power_law_1.2,0.12132480144500732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,2,balanced,0.1841813325881958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,2,balanced,0.2690933346748352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,2,balanced,0.09463999668757121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,1,power_law_1.2,0.15128320455551147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,1,power_law_1.2,0.11640959978103638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,2,balanced,0.5306560198465983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,1,power_law_1.2,0.13269120454788208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,2,balanced,0.22580800453821817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,2,balanced,0.3210986653963725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,1,power_law_1.2,0.14621440172195435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,2,balanced,0.07558399935563405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,2,balanced,0.23915199438730875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,1,power_law_1.2,0.15786880254745483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,1,power_law_1.2,0.13717119693756102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,2,balanced,0.3370506763458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,2,balanced,0.7307039896647135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,1,power_law_1.2,0.08284800052642823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,2,balanced,0.29418667157491046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,2,balanced,0.42715732256571454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,1,power_law_1.2,0.14555519819259644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,2,balanced,0.09545066952705383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.2,0.0983295977115631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,1,power_law_1.2,0.16594560146331788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,2,balanced,0.7973653475443522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,2,balanced,0.36537599563598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.2,0.13716479539871215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,2,balanced,0.5018186569213867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,1,power_law_1.2,0.17065600156784058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,2,balanced,0.09105066458384196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,1,power_law_1.2,0.18250880241394044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.2,0.17134079933166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,2,balanced,0.4873973528544108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.2,0.20039680004119872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,2,balanced,1.0852959950764973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,2,balanced,0.6858346462249756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,2,balanced,0.08146666487058003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.2,0.237939190864563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,1,power_law_1.2,0.19052159786224365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.2,0.2472127914428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,2,balanced,0.6254613399505615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.2,0.29916160106658934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,2,balanced,0.10644800464312236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,2,balanced,0.8903253078460693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,1,power_law_1.2,0.19684480428695678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,2,balanced,1.4691839218139648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.2,0.3022336006164551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.2,0.43573760986328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,2,balanced,0.8627306620279948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,2,balanced,0.18705066045125326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.2,0.4022655963897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,1,power_law_1.2,0.23112959861755372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.2,0.5641471862792968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,2,balanced,1.2630133628845215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.2,0.5009151935577393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,2,balanced,0.1856373349825541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.2,0.832089614868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,2,balanced,2.051263968149821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,1,power_law_1.2,0.26416640281677245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,2,balanced,1.1320052941640217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.2,0.7016191959381104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.2,1.0946623802185058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,2,balanced,1.684234619140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,2,balanced,0.22564266125361124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,1,power_law_1.2,0.3303999900817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.2,0.903007984161377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.2,1.356608009338379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,2,balanced,0.14564800262451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,1,power_law_1.2,0.3911616086959839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,2,balanced,2.7729279200236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.2,1.3272383689880372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,2,balanced,1.6444692611694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,1,power_law_1.2,2.1502784729003905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,2,balanced,0.12692800164222717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,2,balanced,2.454554716746012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,1,power_law_1.2,0.5236735820770264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.2,1.7306175231933594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,2,balanced,0.163482666015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,1,power_law_1.2,4.2620094299316404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,1,power_law_1.2,0.6512256145477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.2,2.1623231887817385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,2,balanced,0.21549866596857706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,1,power_law_1.2,0.9063360214233398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,2,balanced,4.341552098592122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,1,power_law_1.2,3.442630386352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,2,balanced,2.179994742075602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,2,balanced,0.277839998404185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,1,power_law_1.2,1.1600064277648925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,2,balanced,3.2684958775838218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,2,balanced,0.3937600056330363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,1,power_law_1.2,1.669945526123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,1,power_law_1.2,6.597465515136719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,2,balanced,0.5221493244171143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,1,power_law_1.2,2.1617088317871094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,2,balanced,5.474954605102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,2,balanced,2.680522600809733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,1,power_law_1.2,2.6564352035522463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,2,balanced,0.7448480129241943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,2,balanced,4.06984011332194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,1,power_law_1.2,4.217580795288086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,2,balanced,0.9907733599344889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,2,balanced,6.78988774617513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,1,power_law_1.2,8.26604766845703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,2,balanced,1.4565439224243164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,2,balanced,4.51478385925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,2,balanced,6.432175954182942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,2,balanced,1.919424057006836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,2,balanced,10.760971069335938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,2,balanced,2.8755200703938804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,2,balanced,3.8029066721598306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,2,balanced,9.098639806111654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,2,balanced,12.691370646158854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,2,balanced,4.73419189453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,2,balanced,21.378885904947918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,2,power_law_1.01,0.20451838970184327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,2,power_law_1.01,0.17564159631729126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,2,power_law_1.01,0.2866624116897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,2,power_law_1.01,0.3685823917388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,2,power_law_1.01,0.38675200939178467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,2,balanced,7.57473627726237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,2,power_law_1.01,0.38663039207458494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,2,power_law_1.01,0.39165439605712893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,2,power_law_1.01,0.4054975986480713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,2,power_law_1.01,0.42712960243225095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,2,power_law_1.01,0.4321152210235596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,2,power_law_1.01,0.43366398811340334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,2,power_law_1.01,0.4767807960510254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,2,power_law_1.01,0.4745151996612549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,2,power_law_1.01,0.46866559982299805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,2,balanced,15.339823404947916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,2,power_law_1.01,0.5021376132965087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,2,power_law_1.01,0.5447999954223632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,2,power_law_1.01,0.5424767971038819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,2,power_law_1.01,0.6225791931152344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,2,power_law_1.01,0.7811200141906738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,2,power_law_1.01,0.9032768249511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,2,power_law_1.01,1.1713408470153808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,2,power_law_1.01,1.833977508544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,2,power_law_1.01,2.4463424682617188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,2,power_law_1.01,3.276768112182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,2,power_law_1.01,4.71038703918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,2,power_law_1.01,8.089836883544923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,2,power_law_1.01,9.52015380859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,2,power_law_1.01,11.597901153564454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,2,power_law_1.01,16.4496826171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,2,power_law_1.01,36.18338623046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,2,power_law_1.01,0.12704000473022461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,2,power_law_1.01,0.09263359904289245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,2,power_law_1.01,0.1153216004371643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,2,power_law_1.01,0.10351999998092651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,2,power_law_1.01,0.16649600267410278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,2,power_law_1.01,0.10593279600143432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,2,power_law_1.01,0.20703999996185302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,2,power_law_1.01,0.15236480236053468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,2,power_law_1.01,0.21370880603790282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,2,power_law_1.01,0.14651520252227784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,2,power_law_1.01,0.2178431987762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,2,power_law_1.01,0.1481279969215393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,2,power_law_1.01,0.23598079681396483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,2,power_law_1.01,0.15392639636993408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,2,power_law_1.01,0.23855359554290773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,2,power_law_1.01,0.15459200143814086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,2,power_law_1.01,0.2460479974746704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,2,power_law_1.01,0.15556479692459108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,2,power_law_1.01,0.245849609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,2,power_law_1.01,0.047295999526977536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,2,power_law_1.01,0.157478404045105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,2,power_law_1.01,0.25049600601196287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,2,power_law_1.01,0.1615679979324341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,2,power_law_1.01,0.04852479994297028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,2,power_law_1.01,0.24447360038757324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,2,power_law_1.01,0.16803200244903566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,2,power_law_1.01,0.04642559885978699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,2,power_law_1.01,0.26912000179290774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,2,power_law_1.01,0.17471359968185424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,2,power_law_1.01,0.28176639080047605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,2,power_law_1.01,0.05802239775657654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,2,power_law_1.01,0.3359872102737427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,2,power_law_1.01,0.18001279830932618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,2,power_law_1.01,0.09141759872436524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,2,power_law_1.2,0.18557440042495726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,2,power_law_1.01,0.33966081142425536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,2,power_law_1.01,0.189792001247406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,2,power_law_1.01,0.09104639887809754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,2,power_law_1.01,0.33585920333862307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,2,power_law_1.01,0.21426560878753662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,2,power_law_1.2,0.19136639833450317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,2,power_law_1.01,0.22801918983459474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,2,power_law_1.01,0.07315840125083924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,2,power_law_1.01,0.3837055921554565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,2,power_law_1.2,0.2531455993652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,2,power_law_1.01,0.47317118644714357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,2,power_law_1.01,0.2836607933044434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,2,power_law_1.01,0.0939199984073639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,2,power_law_1.2,0.3829184055328369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,2,power_law_1.01,0.6306879997253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.01,0.31959679126739504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,2,power_law_1.01,0.08799999952316284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,2,power_law_1.2,0.3878015995025635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,2,power_law_1.01,0.7577856063842774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,2,power_law_1.01,0.07955840229988098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.01,0.438047981262207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,2,power_law_1.2,0.3694591999053955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,2,power_law_1.2,0.3973439931869507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,2,power_law_1.01,0.10317440032958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,2,power_law_1.01,0.9791296005249024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.01,0.5194496154785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,2,power_law_1.2,0.3955264091491699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,2,power_law_1.01,0.18613120317459106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.01,0.7217152118682861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,2,power_law_1.01,1.247257614135742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,2,power_law_1.2,0.42074880599975584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,2,power_law_1.01,0.18547199964523314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,2,power_law_1.2,0.43227519989013674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.01,0.9644864082336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,2,power_law_1.01,1.8817472457885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,2,power_law_1.01,0.22442240715026857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,2,power_law_1.2,0.4253695964813232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.01,1.5051136016845703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,2,power_law_1.2,0.4651519775390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,2,power_law_1.01,0.1449728012084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,2,power_law_1.01,2.5520320892333985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.01,1.6549951553344726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,2,power_law_1.2,0.46289920806884766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,2,power_law_1.01,0.12401280403137208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,2,power_law_1.2,0.4438271999359131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,2,power_law_1.01,3.9344383239746095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,2,power_law_1.2,0.12520960569381714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,2,power_law_1.01,0.16316159963607788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.01,2.4040447235107423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,2,power_law_1.2,0.5112192153930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,2,power_law_1.01,0.21246719360351562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,2,power_law_1.2,0.5641151905059815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,2,power_law_1.2,0.0969215989112854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.01,3.456537628173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,2,power_law_1.2,0.13112959861755372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,2,power_law_1.01,5.271244812011719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.01,0.2741760015487671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,2,power_law_1.2,0.5705023765563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,2,power_law_1.2,0.1324671983718872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.01,0.38933119773864744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,2,power_law_1.2,0.7138368129730225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,2,power_law_1.2,0.09320319890975952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,2,power_law_1.01,5.559430313110352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,2,power_law_1.01,6.97760009765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,2,power_law_1.2,0.2143872022628784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.01,0.5187071800231934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,2,power_law_1.2,0.7803520202636719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,2,power_law_1.2,0.1063423991203308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,2,power_law_1.2,0.21246719360351562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.01,0.7434239864349366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,2,power_law_1.2,1.0417728424072266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,2,power_law_1.2,0.14766720533370972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,2,power_law_1.01,6.514438629150391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,2,power_law_1.2,0.21724801063537597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.01,0.9882431983947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,2,power_law_1.2,0.1525760054588318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,2,power_law_1.2,1.3462783813476562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,2,power_law_1.01,10.368287658691406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,2,power_law_1.2,0.22451200485229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,2,power_law_1.2,0.15348479747772217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.01,1.4560064315795898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,2,power_law_1.2,1.7623935699462892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,2,power_law_1.2,0.15454720258712767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,2,power_law_1.2,0.22830719947814943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,2,power_law_1.2,0.04627200067043304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.01,1.9184768676757813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,2,power_law_1.2,2.4873983383178713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,2,power_law_1.2,0.15391360521316527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,2,power_law_1.2,0.23479039669036866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,2,power_law_1.01,16.76305236816406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,2,power_law_1.2,0.1567744016647339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,2,power_law_1.2,0.24007039070129393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.01,2.8718847274780273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,2,power_law_1.2,3.723641586303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,2,power_law_1.2,0.04718720018863678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,2,power_law_1.01,20.18824920654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,2,power_law_1.2,0.15741440057754516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,2,power_law_1.2,0.2441472053527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.01,3.799603271484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,2,power_law_1.2,6.040864181518555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,2,power_law_1.2,0.1614400029182434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,2,power_law_1.2,0.046540799736976626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,2,power_law_1.2,0.2555775880813599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,2,power_law_1.2,0.1674496054649353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,2,power_law_1.2,0.26791040897369384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,2,power_law_1.2,0.057036799192428586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,2,power_law_1.01,4.730355072021484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,2,power_law_1.2,0.17141120433807372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,2,power_law_1.2,7.364454650878907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,2,power_law_1.2,0.28124799728393557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,2,power_law_1.2,0.17745920419692993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,2,power_law_1.2,0.09199360013008118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,2,power_law_1.2,0.30951039791107177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,2,power_law_1.2,0.20220799446105958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,2,power_law_1.01,7.5676414489746096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,2,power_law_1.2,0.3571968078613281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,2,power_law_1.2,0.09317119717597962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,2,power_law_1.2,9.698291015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,2,power_law_1.2,0.211680006980896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,2,power_law_1.2,0.32595200538635255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,2,power_law_1.2,0.07399680018424988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,2,power_law_1.2,0.23463039398193358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,2,power_law_1.2,0.28241279125213625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,2,power_law_1.2,0.3837183952331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,2,power_law_1.2,11.546176147460937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,2,power_law_1.2,0.09394559860229493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,2,power_law_1.01,15.168365478515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.2,0.34522879123687744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,2,power_law_1.2,0.47781758308410643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,2,power_law_1.2,0.08966400027275086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.2,0.44732160568237306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.2,0.5325823783874511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,2,power_law_1.2,0.6207168102264404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,2,power_law_1.2,0.07919999957084656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.2,0.7795584201812744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,2,power_law_1.2,19.05878448486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,2,power_law_1.2,0.10374399423599243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,2,power_law_1.2,0.8384896278381347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.2,0.9831168174743652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,2,power_law_1.2,0.18497920036315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.2,1.3889535903930663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,2,power_law_1.2,1.101433563232422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,2,power_law_1.2,0.18325120210647583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.2,1.8185600280761718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,2,power_law_1.2,1.4579903602600097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.2,3.2437950134277345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,2,power_law_1.2,33.221432495117185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,2,power_law_1.2,0.2232896089553833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.2,3.9195838928222657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,2,power_law_1.2,0.1437376022338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,2,power_law_1.2,1.9877376556396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,2,power_law_1.2,5.453081512451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,2,power_law_1.2,0.12550400495529174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,2,power_law_1.2,2.8755071640014647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,2,power_law_1.2,0.16223360300064088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,2,power_law_1.2,4.106060791015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,2,power_law_1.2,7.947430419921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,1,balanced,0.11847999691963196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,2,power_law_1.2,0.21171200275421143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,2,power_law_1.2,5.403462219238281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,1,balanced,0.21091733376185098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.2,0.2751039981842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,2,power_law_1.2,15.403173828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,1,balanced,0.39192001024882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.2,0.39114880561828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,2,power_law_1.2,6.773554992675781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.2,0.5194111824035644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,1,balanced,0.3847839832305908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.2,0.7418111801147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,2,power_law_1.2,10.007154846191407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,1,balanced,0.384117325146993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.2,0.988697624206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,1,balanced,0.3837706645329793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.2,1.4537983894348145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,1,balanced,0.38549331823984784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,1,balanced,0.38894399007161456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.2,1.9157503128051758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,2,power_law_1.2,21.24751434326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,1,balanced,0.3843253453572591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.2,2.873619270324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,1,balanced,0.38815999031066895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,1,balanced,0.3935893376668294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.2,3.7973438262939454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,1,balanced,0.39470934867858887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,1,balanced,0.39875201384226483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,2,power_law_1.2,4.7335552215576175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,1,balanced,0.42019200325012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,1,balanced,0.42005332310994464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,1,balanced,0.42606933911641437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,2,power_law_1.2,7.556403350830078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,1,balanced,0.4798986514409383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,1,balanced,0.5177280108133951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,1,balanced,0.529584010442098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,2,power_law_1.2,15.444032287597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,1,balanced,0.7167306741078695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,1,balanced,0.7790613174438477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,1,balanced,1.0616479714711506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,1,balanced,1.42083740234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,1,balanced,2.0578826268514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,1,balanced,2.699359893798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,1,balanced,4.035749435424805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,1,balanced,5.4159895579020185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,1,balanced,6.712127685546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,1,balanced,10.686138153076172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,1,balanced,20.886810302734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,1,balanced,0.0783733328183492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,1,balanced,0.06445866823196411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,1,balanced,0.1225920021533966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,1,balanced,0.09404266873995464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,1,balanced,0.2157599925994873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,1,balanced,0.1387999951839447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,1,balanced,0.21409066518147787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,1,balanced,0.14432000120480856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,1,balanced,0.2092906634012858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,1,balanced,0.1434879998366038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,1,balanced,0.21215999126434326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,1,balanced,0.14497066537539163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,1,balanced,0.21612266699473062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,1,balanced,0.14807466665903726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,1,balanced,0.1481119990348816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,1,balanced,0.21732266743977866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,1,balanced,0.1476800044377645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,1,balanced,0.2173759937286377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,1,balanced,0.2199839949607849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,1,balanced,0.1499573290348053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,1,balanced,0.2254506746927897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,1,balanced,0.15149866541226706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,1,balanced,0.22893865903218588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,1,balanced,0.027893332143624622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,1,balanced,0.15412799517313638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,1,balanced,0.23028266429901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,1,balanced,0.1548533340295156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,1,power_law_1.01,0.1173632025718689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,1,balanced,0.028666667640209198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,1,balanced,0.24100800355275473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,1,balanced,0.16168000300725302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,1,balanced,0.2567253311475118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,1,balanced,0.02752533306678136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,1,balanced,0.179802676041921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,1,power_law_1.01,0.17487360239028932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,1,balanced,0.2612266739209493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,1,balanced,0.18101332585016885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,1,balanced,0.03348266581694285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,1,power_law_1.01,0.21142399311065674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,1,balanced,0.26505066951115924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,1,balanced,0.18548266092936197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,1,balanced,0.045834665497144066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,1,balanced,0.3247893253962199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,1,power_law_1.01,0.3757760047912598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,1,balanced,0.22617600361506143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,1,balanced,0.06238399942715963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,1,balanced,0.34379732608795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,1,power_law_1.01,0.37888638973236083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,1,balanced,0.24593599637349448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,1,power_law_1.01,0.07589120268821717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,1,balanced,0.05365333457787832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,1,power_law_1.01,0.3842751979827881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,1,balanced,0.4341973463694255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,1,balanced,0.30867733558019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,1,power_law_1.01,0.06280320286750793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,1,balanced,0.06243733565012614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,1,power_law_1.01,0.37585279941558836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,1,power_law_1.01,0.10300159454345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,1,balanced,0.38171199957529706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,1,balanced,0.501477320988973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,1,balanced,0.052154665191968284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,1,power_law_1.01,0.07976319789886474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,1,power_law_1.01,0.3888767957687378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,1,power_law_1.01,0.1125440001487732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,1,power_law_1.01,0.2029952049255371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,1,balanced,0.057333335280418396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,1,power_law_1.01,0.3981760025024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,1,power_law_1.01,0.09858559966087341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,1,power_law_1.01,0.20122880935668946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,1,balanced,0.5168053309122721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,1,balanced,0.06683200101057689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,1,balanced,0.6962347030639648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,1,power_law_1.01,0.21311359405517577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,1,power_law_1.01,0.4072768211364746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,1,power_law_1.01,0.13934719562530518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,1,balanced,0.060602664947509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,1,power_law_1.01,0.2122112035751343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,1,power_law_1.01,0.02632319927215576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,1,power_law_1.01,0.41328001022338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,1,power_law_1.01,0.14049919843673705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,1,power_law_1.01,0.21794559955596923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,1,balanced,0.06517333288987477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,1,balanced,0.6619093418121338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,1,balanced,0.9452213446299235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,1,power_law_1.01,0.22176001071929932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,1,power_law_1.01,0.4340223789215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,1,balanced,0.07776533563931783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,1,power_law_1.01,0.14142719507217408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,1,power_law_1.01,0.22359039783477783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,1,power_law_1.01,0.026655998826026917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,1,power_law_1.01,0.43532161712646483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,1,power_law_1.01,0.23023359775543212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,1,balanced,0.1611840029557546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,1,power_law_1.01,0.14237439632415771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,1,power_law_1.01,0.23733758926391602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,1,power_law_1.01,0.44852480888366697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,1,power_law_1.01,0.02661119997501373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,1,balanced,0.9344000021616617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,1,balanced,0.18491200606028238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,1,balanced,1.3479520479838054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,1,power_law_1.01,0.2473344087600708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,1,power_law_1.01,0.510316801071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,1,balanced,0.2201706568400065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,1,power_law_1.01,0.14600319862365724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,1,power_law_1.01,0.24673280715942383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,1,power_law_1.01,0.03139199912548065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,1,power_law_1.01,0.5174911975860595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,1,power_law_1.01,0.2736255884170532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,1,balanced,0.11957866946856181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,1,power_law_1.01,0.04366079866886139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,1,power_law_1.01,0.1443711996078491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,1,power_law_1.01,0.5235136032104493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,1,power_law_1.01,0.2828351974487305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,1,balanced,1.197914679845174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,1,balanced,1.6620426177978516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,1,balanced,0.14640532930692038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,1,power_law_1.01,0.060812801122665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,1,power_law_1.01,0.2982912063598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,1,power_law_1.01,0.6061503887176514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,1,power_law_1.01,0.14723199605941772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,1,balanced,0.20388267437616983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,1,power_law_1.01,0.35509119033813474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,1,power_law_1.01,0.0521664023399353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,1,power_law_1.01,0.6769919872283936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,1,power_law_1.01,0.15370240211486816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,1,power_law_1.01,0.38623359203338625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,1,balanced,0.2595146695772807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,1,power_law_1.01,0.06028159856796265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,1,power_law_1.01,0.8307135581970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,1,power_law_1.01,0.4868480205535889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,1,balanced,2.440000057220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,1,power_law_1.01,0.050406402349472045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,1,power_law_1.01,0.987820816040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,1,power_law_1.01,0.15640319585800172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,1,balanced,1.7070399920145671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,1,power_law_1.01,0.5880447864532471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,1,balanced,0.362335999806722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,1,power_law_1.01,0.054124802350997925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,1,power_law_1.01,1.2981375694274901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,1,power_law_1.01,0.7844672203063965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,1,power_law_1.01,0.1607103943824768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,1,power_law_1.01,0.06500480175018311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,1,power_law_1.01,1.601215934753418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,1,power_law_1.01,0.1679744005203247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,1,power_law_1.01,0.998969554901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,1,power_law_1.01,0.05850239992141724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,1,balanced,0.4694399833679199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,1,power_law_1.01,1.4202303886413574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,1,power_law_1.01,0.06269440054893494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,1,power_law_1.01,2.319584083557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,1,power_law_1.01,0.1829632043838501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,1,balanced,3.1990346908569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,1,power_law_1.01,0.0753920018672943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,1,power_law_1.01,1.8157567977905273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,1,power_law_1.01,0.1851904034614563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,1,power_law_1.01,3.0326847076416015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,1,balanced,2.2784694035847983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,1,power_law_1.01,0.15928319692611695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,1,balanced,0.6825493176778158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,1,power_law_1.01,0.2052992105484009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,1,power_law_1.01,2.5821439743041994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,1,power_law_1.01,0.18315520286560058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,1,power_law_1.01,4.497420883178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,1,power_law_1.01,0.23476479053497315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,1,power_law_1.01,3.3630912780761717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,1,power_law_1.01,0.21662719249725343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.01,0.27500159740448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,1,power_law_1.01,5.7309825897216795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,1,power_law_1.01,0.11832319498062134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,1,power_law_1.01,4.075929641723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,1,balanced,0.8918240070343018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.01,0.34247679710388185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.01,0.1422271966934204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,1,balanced,3.9565601348876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.01,0.4110079765319824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,1,power_law_1.01,7.1171714782714846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.01,0.20018560886383058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,1,power_law_1.01,6.4115455627441404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,1,balanced,2.8273280461629233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.01,0.5510655879974365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.01,0.25571839809417723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.01,0.6984320163726807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.01,0.35841920375823977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,1,balanced,1.3117706775665283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,1,power_law_1.01,10.961459350585937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.01,0.4691648006439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.01,0.9983615875244141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,1,power_law_1.01,12.532550048828124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.01,0.676032018661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.01,1.261171245574951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.01,0.8886848449707031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.01,1.8204992294311524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.01,1.3043647766113282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,1,power_law_1.01,21.100633239746095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.01,2.416748809814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.01,1.723174476623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,1,balanced,1.7287786801656086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,1,balanced,6.257738749186198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.01,3.015091133117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,1,balanced,4.723349253336589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.01,2.145702362060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,1,power_law_1.01,4.865606307983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,1,power_law_1.01,3.4014209747314452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,1,power_law_1.01,6.739379119873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,1,power_law_1.01,9.4299072265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,1,balanced,2.1476906140645347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,1,balanced,3.4042027791341147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,1,balanced,9.374325434366861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,1,balanced,12.446629842122396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,1,power_law_1.2,0.11777280569076538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,1,power_law_1.2,0.16957440376281738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,1,power_law_1.2,0.2106623888015747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,1,balanced,6.751189549763997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,1,power_law_1.2,0.38543999195098877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,1,power_law_1.2,0.36833279132843016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,1,power_law_1.2,0.38220160007476806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,1,power_law_1.2,0.38392319679260256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,1,power_law_1.2,0.3928447961807251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,1,power_law_1.2,0.4011199951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,1,power_law_1.2,0.406771183013916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,1,power_law_1.2,0.42181758880615233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,1,power_law_1.2,0.43274240493774413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,1,power_law_1.2,0.4336512088775635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,1,power_law_1.2,0.4663871765136719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,1,power_law_1.2,0.509388780593872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,1,power_law_1.2,0.5356351852416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,1,power_law_1.2,0.5391168117523193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,1,power_law_1.2,0.5956480026245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,1,power_law_1.2,0.6714816093444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,1,power_law_1.2,0.8440511703491211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,1,power_law_1.2,0.9635071754455566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,1,power_law_1.2,1.2779583930969238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,1,power_law_1.2,1.6071552276611327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,1,power_law_1.2,2.270694351196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,1,power_law_1.2,3.0038143157958985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,1,power_law_1.2,4.422444915771484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,1,power_law_1.2,5.785030364990234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,1,power_law_1.2,7.054342651367188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,1,power_law_1.2,10.93151397705078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,1,power_law_1.2,21.113626098632814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,1,power_law_1.2,0.0767359972000122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,1,power_law_1.2,0.06257920265197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,1,power_law_1.2,0.02619520127773285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,1,power_law_1.2,0.09793919920921326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,1,power_law_1.2,0.07845119833946228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,1,power_law_1.2,0.026624000072479247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,1,power_law_1.2,0.11337599754333497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,1,power_law_1.2,0.08312320113182067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,1,power_law_1.2,0.026476800441741943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,1,power_law_1.2,0.21096320152282716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,1,power_law_1.2,0.13891839981079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,1,power_law_1.2,0.03133440017700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,1,power_law_1.2,0.13636480569839476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,1,power_law_1.2,0.2086656093597412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,1,power_law_1.2,0.04350079894065857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,1,power_law_1.2,0.14105600118637085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,1,power_law_1.2,0.21240320205688476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,1,power_law_1.2,0.14235520362854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,1,power_law_1.2,0.06079999804496765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,1,power_law_1.2,0.2161344051361084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,1,power_law_1.2,0.14562560319900514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,1,power_law_1.2,0.050944000482559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,1,power_law_1.2,0.2151616096496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,1,power_law_1.2,0.14723199605941772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,1,power_law_1.2,0.06085759997367859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,1,power_law_1.2,0.2229248046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,1,power_law_1.2,0.14816639423370362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,1,power_law_1.2,0.05064319968223572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,1,power_law_1.2,0.15045759677886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,1,power_law_1.2,0.2267967939376831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,1,power_law_1.2,0.05438079833984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,1,power_law_1.2,0.15633920431137086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,1,power_law_1.2,0.23911681175231933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,1,power_law_1.2,0.16223360300064088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,1,power_law_1.2,0.06472960114479065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,1,power_law_1.2,0.2481920003890991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,1,power_law_1.2,0.16711679697036744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,1,power_law_1.2,0.058387202024459836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,1,power_law_1.2,0.2465343952178955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,1,power_law_1.2,0.18206080198287963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,1,power_law_1.2,0.24455680847167968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,1,power_law_1.2,0.06279039978981019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,1,power_law_1.2,0.19036799669265747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,1,power_law_1.2,0.2737279891967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,1,power_law_1.2,0.20249600410461427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,1,power_law_1.2,0.07550719976425171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,1,power_law_1.2,0.23451519012451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,1,power_law_1.2,0.2803328037261963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,1,power_law_1.2,0.15955840349197387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.2,0.27190399169921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,1,power_law_1.2,0.3039360046386719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,1,power_law_1.2,0.18292479515075682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.2,0.3432703971862793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,1,power_law_1.2,0.3417407989501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.2,0.41244797706604003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,1,power_law_1.2,0.21619200706481934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,1,power_law_1.2,0.39407360553741455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.2,0.5528768062591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,1,power_law_1.2,0.11789439916610718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,1,power_law_1.2,0.48991999626159666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.2,0.6946559906005859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.2,0.14268159866333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,1,power_law_1.2,0.5869696140289307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.2,0.9898431777954102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.2,0.20037760734558105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,1,power_law_1.2,0.7888959884643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.2,1.2643839836120605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.2,0.25596160888671876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,1,power_law_1.2,0.9922687530517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.2,1.8538047790527343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.2,0.3577791929244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,1,power_law_1.2,1.4422911643981933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.2,2.4104896545410157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.2,0.4691455841064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,1,power_law_1.2,1.807302474975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.2,3.041734313964844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.2,0.677568006515503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,1,power_law_1.2,2.566982460021973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,1,power_law_1.2,4.911859130859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.2,0.8851903915405274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,1,power_law_1.2,3.3602622985839843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.2,1.3063679695129395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,1,power_law_1.2,9.413407897949218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,1,power_law_1.2,4.098182296752929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.2,1.7249536514282227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.2,2.1444608688354494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,1,power_law_1.2,6.420262145996094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,1,power_law_1.2,3.402201461791992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,1,power_law_1.2,12.519725036621093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,2,balanced,0.09431466460227966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,1,power_law_1.2,6.746905517578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,2,balanced,0.06989333530267079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,2,balanced,0.14653333028157553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,2,balanced,0.06931733091672261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,2,balanced,0.24231467644373575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,2,balanced,0.09371200203895569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,2,balanced,0.4408106803894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,2,balanced,0.1437173287073771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,2,balanced,0.8347413539886475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,2,balanced,0.2472533384958903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,2,balanced,0.08853866656621297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,2,balanced,1.632922649383545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,2,balanced,0.44650665918986004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,2,balanced,0.12046933174133301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,2,balanced,1.6345866521199544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,2,balanced,0.8423360188802084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,2,balanced,0.18440000216166177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,2,balanced,1.6277920405069988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,2,balanced,0.30663466453552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,2,balanced,1.6287253697713215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,2,balanced,0.8435040314992269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,2,balanced,0.5493173201878866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,2,balanced,1.6396479606628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,2,balanced,0.8428053061167399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,2,balanced,0.5528800090154012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,2,balanced,1.6318027178446453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,2,balanced,0.06591466565926869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,2,balanced,0.8481067021687826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,2,balanced,0.0642986645301183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,2,balanced,0.555295983950297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,2,balanced,1.6410239537556965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,2,balanced,0.8496479988098145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,2,balanced,0.09357333183288574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,2,balanced,0.854634682337443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,2,balanced,1.6415680249532063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,2,balanced,0.5569706757863363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,2,balanced,0.14412267009417215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,2,balanced,0.8633920351664225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,2,balanced,0.255023996035258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,2,balanced,1.658890724182129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,2,balanced,0.5605173508326212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,2,balanced,0.4678933223088582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,2,balanced,0.8685333728790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,2,balanced,0.48054933547973633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,2,balanced,1.6671679814656575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,2,balanced,0.5626506805419922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,2,balanced,0.8748479684193929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,2,balanced,0.49216000239054364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,2,balanced,0.5664373238881429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,2,balanced,0.8826986948649088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,2,balanced,1.676693280537923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,2,balanced,0.504965345064799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,2,balanced,0.5690666834513346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,2,balanced,0.8892052968343099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,2,balanced,0.5166613260904948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,2,balanced,1.6951573689778645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,2,balanced,0.5787093242009481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,2,balanced,0.5428586800893148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,2,balanced,0.9117333094278971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,2,balanced,0.5132266680399576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,2,balanced,0.5839466651280721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,2,balanced,1.7344533602396648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,2,balanced,0.9394986629486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,2,balanced,0.5244799852371216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,2,balanced,0.5910026629765829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,2,balanced,0.5493866602579752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,2,balanced,0.6039199829101562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,2,balanced,1.7783519426981609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,2,balanced,0.9612747033437093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,2,balanced,0.5538026491800944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,2,balanced,0.6255733172098795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,2,balanced,0.5567520062128702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,2,balanced,1.0182666778564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,2,balanced,1.8269386291503906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,2,balanced,0.5740266640981039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,2,balanced,0.6452373266220093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,2,balanced,1.0845599969228108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,2,balanced,0.7425333658854166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,2,balanced,1.8799734115600586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,2,balanced,0.6984586715698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,2,balanced,0.7113813559214274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,2,balanced,1.1806399822235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,2,balanced,0.7617706457773844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,2,balanced,1.9679946899414062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,2,balanced,1.3737227121988933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,2,balanced,0.8598293463389078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,2,balanced,1.5005866686503093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,2,balanced,1.28165864944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,2,balanced,2.1848373413085938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,2,balanced,1.1499253114064534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,2,balanced,0.9588267008463541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,2,balanced,1.6744906107584636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,2,balanced,1.1927839914957683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,2,balanced,2.3976319630940757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,2,balanced,1.3173333009084065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,2,balanced,1.760682741800944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,2,balanced,1.2778560320536296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,2,balanced,2.5247999827067056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,2,balanced,1.5217067400614421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,2,balanced,1.3813014030456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,2,power_law_1.01,0.14368640184402465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,2,balanced,2.7251628239949546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,2,power_law_1.01,0.23892478942871093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,2,power_law_1.01,0.21403520107269286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,2,balanced,3.7206080754597983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,2,power_law_1.01,0.33948800563812254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,2,power_law_1.01,0.5691840171813964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,2,balanced,2.1405332883199057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,2,power_law_1.01,0.7999167919158936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,2,power_law_1.01,1.1335871696472168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,2,power_law_1.01,1.2031935691833495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,2,balanced,2.4285546938578286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,2,power_law_1.01,1.272531223297119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,2,power_law_1.01,1.2089920043945312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,2,power_law_1.01,1.280742359161377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,2,power_law_1.01,1.3562623977661132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,2,balanced,2.946410814921061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,2,power_law_1.01,1.3778176307678223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,2,power_law_1.01,1.441215991973877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,2,balanced,4.552650769551595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,2,power_law_1.01,1.4465920448303222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,2,power_law_1.01,1.4700799942016602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,2,balanced,2.7097225189208984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,2,power_law_1.01,1.542905616760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,2,power_law_1.01,1.6501888275146483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,2,power_law_1.01,1.7627328872680663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,2,balanced,2.532543977101644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,2,power_law_1.01,1.9667903900146484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,2,power_law_1.01,1.890118408203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,2,power_law_1.01,2.1528255462646486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,2,power_law_1.01,2.421036720275879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,2,power_law_1.01,2.9191360473632812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,2,power_law_1.01,3.2798015594482424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,2,balanced,4.128570556640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,2,power_law_1.01,4.23966064453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,2,balanced,5.7011464436848955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,2,power_law_1.01,5.424531173706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,2,power_law_1.01,6.5474494934082035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,2,balanced,3.4380693435668945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,2,power_law_1.01,9.71310043334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,2,balanced,3.618197441101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,2,power_law_1.01,18.195333862304686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,2,power_law_1.01,0.09347839951515198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,2,balanced,5.513386408487956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,2,power_law_1.01,0.14176000356674195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,2,power_law_1.01,0.11096960306167603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,2,balanced,7.469194412231445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,2,power_law_1.01,0.18670079708099366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,2,power_law_1.01,0.29738240242004393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,2,power_law_1.01,0.44375038146972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,2,power_law_1.01,0.604307222366333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,2,power_law_1.01,0.631935977935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,2,balanced,5.125775973002116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,2,balanced,4.9026187260945635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,2,power_law_1.01,0.6502143859863281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,2,power_law_1.01,0.08562560081481933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,2,power_law_1.01,0.6741184234619141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,2,power_law_1.01,0.11774719953536987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,2,power_law_1.01,0.6858880043029785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,2,power_law_1.01,0.10913920402526855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,2,power_law_1.01,0.7050879955291748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,2,power_law_1.01,0.15164159536361693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,2,power_law_1.01,0.7438464164733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,2,power_law_1.01,0.2237567901611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,2,power_law_1.01,0.7616384029388428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,2,power_law_1.01,0.2943295955657959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,2,power_law_1.01,0.04895359873771667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,2,power_law_1.01,0.7820479869842529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,2,power_law_1.01,0.3880320072174072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,2,power_law_1.01,0.8132351875305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,2,power_law_1.01,0.41443839073181155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,2,power_law_1.01,0.07519360184669495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,2,power_law_1.01,0.42312960624694823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,2,power_law_1.01,0.8342911720275878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,2,power_law_1.01,0.07331200242042542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,2,power_law_1.01,0.4387968063354492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,2,power_law_1.01,0.1100160002708435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,2,power_law_1.01,0.9193856239318847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,2,power_law_1.01,0.4671807765960693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,2,power_law_1.01,0.17594239711761475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,2,power_law_1.01,0.9277631759643554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,2,power_law_1.01,0.46990718841552737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,2,power_law_1.01,0.23144960403442383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,2,power_law_1.01,1.089459228515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,2,power_law_1.01,0.48049278259277345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,2,power_law_1.01,0.3211647987365723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,2,power_law_1.01,0.492844820022583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,2,power_law_1.01,1.085644817352295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,2,power_law_1.01,0.34180479049682616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,2,power_law_1.01,0.5190400123596192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,2,power_law_1.01,0.35920000076293945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,2,power_law_1.01,1.2737407684326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,2,power_law_1.01,0.5234943866729737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,2,power_law_1.01,0.37648639678955076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,2,power_law_1.01,0.5482624053955079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,2,power_law_1.01,1.4581695556640626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,2,power_law_1.01,0.5860095977783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,2,power_law_1.01,0.393503999710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,2,power_law_1.01,1.828883171081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.01,0.6278528213500977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,2,power_law_1.01,0.3971776008605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,2,power_law_1.01,2.1818239212036135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.01,0.7291391849517822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,2,power_law_1.01,0.40906238555908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,2,power_law_1.01,2.8974143981933596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.01,0.7776447772979737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,2,power_law_1.01,0.4293759822845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.01,0.9603520393371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,2,power_law_1.01,3.6564159393310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,2,power_law_1.01,0.42821760177612306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.01,1.0903167724609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,2,power_law_1.01,4.24224624633789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,2,power_law_1.01,0.43520641326904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.01,1.4007488250732423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,2,power_law_1.01,0.46129279136657714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,2,power_law_1.01,6.6602943420410154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.01,1.779199981689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,2,power_law_1.01,0.4958847999572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.01,2.4106239318847655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.01,0.5222720146179199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,2,power_law_1.01,12.290201568603516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.01,3.090239906311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.01,1.1675135612487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.01,1.3522175788879394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,2,power_law_1.01,3.8885311126708983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,2,balanced,10.868133544921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.01,0.8584256172180176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,2,power_law_1.01,6.095507049560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.01,0.9515263557434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,2,balanced,14.535760243733725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.01,1.139033603668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,2,power_law_1.01,11.344230651855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.01,1.3062463760375977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.01,1.7244735717773438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.01,2.3405439376831056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,2,balanced,9.905765533447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,2,balanced,9.680170694986979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,2,power_law_1.01,2.7176511764526365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,2,power_law_1.01,3.8580032348632813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,2,power_law_1.01,7.3833473205566404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,2,power_law_1.2,0.14447360038757323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,2,power_law_1.2,0.218451189994812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,2,power_law_1.2,0.18257280588150024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,2,power_law_1.2,0.32451839447021485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,2,power_law_1.2,0.5392767906188964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,2,power_law_1.2,0.7153024196624755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,2,power_law_1.2,1.0918463706970214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,2,power_law_1.2,1.1883647918701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,2,power_law_1.2,1.1996159553527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,2,power_law_1.2,1.2602496147155762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,2,power_law_1.2,1.302342414855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,2,power_law_1.2,1.3610943794250487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,2,power_law_1.2,1.3586560249328614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,2,power_law_1.2,1.4379839897155762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,2,power_law_1.2,1.4921343803405762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,2,power_law_1.2,1.491596794128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,2,power_law_1.2,1.5773887634277344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,2,power_law_1.2,1.7503231048583985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,2,power_law_1.2,1.8590143203735352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,2,power_law_1.2,2.1038911819458006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,2,power_law_1.2,2.035788726806641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,2,power_law_1.2,2.206950378417969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,2,power_law_1.2,2.5961408615112305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,2,power_law_1.2,3.045702362060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,2,power_law_1.2,3.5334720611572266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,2,power_law_1.2,4.592569732666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,2,power_law_1.2,5.777439880371094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,2,power_law_1.2,6.986329650878906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,2,power_law_1.2,10.715558624267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,2,power_law_1.2,18.373158264160157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,1,balanced,0.09697066744168599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,2,power_law_1.2,0.09452160000801087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,2,power_law_1.2,0.08565760254859925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,1,balanced,0.14703466494878134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,2,power_law_1.2,0.13967360258102418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,2,power_law_1.2,0.11475839614868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,1,balanced,0.24449066321055093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,1,balanced,0.4453973372777303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,2,power_law_1.2,0.04922240078449249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,2,power_law_1.2,0.11421439647674561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,2,power_law_1.2,0.09882240295410157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,1,balanced,0.8375413417816162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,2,power_law_1.2,0.15396480560302733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,2,power_law_1.2,0.19415680170059205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,1,balanced,1.6184693972269695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,2,power_law_1.2,0.07539839744567871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,1,balanced,1.636522610982259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,2,power_law_1.2,0.20710399150848388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,2,power_law_1.2,0.27340800762176515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,2,power_law_1.2,0.06970239877700805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,1,balanced,1.6264479955037434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,2,power_law_1.2,0.2590719938278198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,2,power_law_1.2,0.37601919174194337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,2,power_law_1.2,0.10583679676055908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,1,balanced,1.6334187189737956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,2,power_law_1.2,0.3861504077911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,2,power_law_1.2,0.5761536121368408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,1,balanced,1.635957400004069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,2,power_law_1.2,0.1590399980545044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,1,balanced,1.63701327641805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,2,power_law_1.2,0.41896958351135255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,2,power_law_1.2,0.6001920223236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,2,power_law_1.2,0.21518080234527587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,1,balanced,1.650986671447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,2,power_law_1.2,0.6369023799896241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,2,power_law_1.2,0.4178431987762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,1,balanced,1.6568053563435872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,2,power_law_1.2,0.6589056015014648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,2,power_law_1.2,0.32079999446868895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,2,power_law_1.2,0.4462719917297363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,1,balanced,1.6675893465677898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,2,power_law_1.2,0.3488447904586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,2,power_law_1.2,0.674886417388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,2,power_law_1.2,0.4492095947265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,1,balanced,0.07171200215816498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,2,power_law_1.2,0.36462719440460206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,1,balanced,1.681536038716634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,2,power_law_1.2,0.7142720222473145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,2,power_law_1.2,0.47162880897521975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,2,power_law_1.2,0.7285312175750732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,1,balanced,1.7011094093322754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,1,balanced,0.09742933511734009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,2,power_law_1.2,0.49028477668762205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,2,power_law_1.2,0.3706687927246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,2,power_law_1.2,0.7751552104949951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,1,balanced,1.7367520332336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,1,balanced,0.1481706698735555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,2,power_law_1.2,0.5082687854766845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,2,power_law_1.2,0.7977791786193847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,2,power_law_1.2,0.5295423984527587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,2,power_law_1.2,0.4005375862121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,1,balanced,0.2515626748402913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,1,balanced,0.08467732866605122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,1,balanced,0.03577066709597906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,1,balanced,1.7606933911641438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,2,power_law_1.2,0.5534527778625489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,2,power_law_1.2,0.8283967971801758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,1,balanced,0.44897600015004474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,2,power_law_1.2,0.4019775867462158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,1,balanced,0.043562665581703186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,2,power_law_1.2,0.5782527923583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,1,balanced,0.10543466607729594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,2,power_law_1.2,0.8707072257995605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,1,balanced,0.8425386746724447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,2,power_law_1.2,0.39609599113464355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,1,balanced,0.057914664347966514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,1,balanced,1.830949306488037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,1,balanced,0.8483359813690186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,1,balanced,0.08594133456548055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,2,power_law_1.2,0.630182409286499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,1,balanced,0.13843199610710144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,2,power_law_1.2,0.9939968109130859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,2,power_law_1.2,0.4282495975494385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,1,balanced,0.1426293353239695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,1,balanced,0.8518079916636149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.2,0.6657408237457275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,1,balanced,0.20671466986338297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,1,balanced,0.2586933374404907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,1,balanced,0.8598612944285074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,1,balanced,1.914639949798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,2,power_law_1.2,0.4447360038757324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,2,power_law_1.2,0.9964991569519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,1,balanced,0.2658720016479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.2,0.769593620300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,1,balanced,0.33634666601816815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,1,balanced,0.8618613084157308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,2,power_law_1.2,0.44393601417541506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,1,balanced,0.27196266253789264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.2,0.8114239692687988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,2,power_law_1.2,1.1462847709655761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,1,balanced,0.5852640072504679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,1,balanced,1.9893280665079753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,1,balanced,0.870959997177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,1,balanced,0.2797013322512309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,2,power_law_1.2,0.48508801460266116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.2,1.0131711959838867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,1,balanced,0.595088005065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,2,power_law_1.2,1.1241408348083497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,1,balanced,0.8798560301462809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,1,balanced,0.28706133365631104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,2,power_law_1.2,0.538969612121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,1,balanced,0.6028213500976562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.2,1.1531519889831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,1,balanced,0.8807786305745443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,1,balanced,0.3019040028254191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,1,balanced,2.1548213958740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,2,power_law_1.2,1.372556781768799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.2,0.5513919830322266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.2,1.5133567810058595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,1,balanced,0.6056906779607137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,1,balanced,0.31806399424870807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,1,balanced,0.894261360168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.2,1.2645183563232423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,2,power_law_1.2,1.5540096282958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.2,1.85230712890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,1,balanced,0.6059573491414388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,1,balanced,0.33136000235875446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,1,balanced,2.330026626586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,1,balanced,0.907152016957601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,1,balanced,0.6245013475418091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.2,1.3850751876831056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,2,power_law_1.2,2.0126272201538087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.2,2.5071231842041017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,1,balanced,0.3615786631902059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,1,balanced,0.9225227038065592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.2,0.8891648292541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,1,balanced,0.6145066817601522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.2,3.320780944824219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,2,power_law_1.2,2.381888008117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,1,balanced,0.3259306748708089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,1,balanced,2.760944048563639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.2,1.032089614868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,1,balanced,0.3386506636937459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,1,balanced,0.9461066722869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,1,balanced,0.6243093411127726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.2,1.1475839614868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,2,power_law_1.2,4.379423904418945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,2,power_law_1.2,2.9722047805786134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,1,balanced,0.36694931983947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,1,balanced,0.9979893366495768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,1,balanced,0.6437546809514364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.2,1.427295970916748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,1,balanced,3.017519950866699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,2,power_law_1.2,6.016960144042969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,2,power_law_1.2,4.042572784423828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,1,balanced,0.36052266756693524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,1,balanced,1.0311413606007893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,1,balanced,0.6556959946950277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.2,1.8934080123901367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,2,power_law_1.2,4.649651336669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,1,balanced,0.3945173422495524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,2,power_law_1.2,11.166099548339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,1,balanced,0.6652426719665527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,1,balanced,1.1201972961425781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.2,2.243935966491699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,2,power_law_1.2,6.891744232177734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,1,balanced,0.6874826749165853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,1,balanced,0.4026026725769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,2,power_law_1.2,2.8321151733398438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,1,balanced,1.2121013005574544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,1,balanced,4.718554814656575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,1,balanced,0.7371093432108561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,1,balanced,0.44113067785898846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,2,power_law_1.2,4.148998260498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,2,power_law_1.2,13.363084411621093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,1,balanced,0.7736586729685465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,1,balanced,1.376570701599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,2,power_law_1.2,7.658547210693359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,1,balanced,0.9525973002115885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,1,balanced,0.8638613224029541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,1,balanced,1.0996747016906738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,1,balanced,1.560090700785319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,1,balanced,0.9432160059611002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,1,balanced,5.024197260538737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,1,balanced,0.7597333590189616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,1,balanced,1.1167466640472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,1,balanced,2.1175626118977866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,1,balanced,1.2890453338623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,1,balanced,0.8427519798278809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,1,balanced,2.3449920018514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,1,balanced,1.8334879875183105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,1,balanced,6.547152201334636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,1,balanced,1.4046719868977864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,1,balanced,3.438490549723307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,1,balanced,2.181829293568929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,1,balanced,1.5452106793721516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,1,balanced,3.1461334228515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,1,balanced,4.218288103739421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,1,balanced,10.265541076660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,1,balanced,3.9996213912963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,1,balanced,2.127552032470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,1,balanced,5.382106781005859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,1,balanced,5.104832013448079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,1,balanced,2.9759626388549805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,1,balanced,7.964485168457031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,1,power_law_1.01,0.09397760033607483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,1,power_law_1.01,0.1434880018234253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,1,power_law_1.01,0.16587519645690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,1,power_law_1.01,0.35185918807983396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,1,power_law_1.01,0.5626688003540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,1,balanced,7.527210871378581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,1,balanced,18.93780263264974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,1,power_law_1.01,0.766707181930542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,1,power_law_1.01,1.0971199989318847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,1,power_law_1.01,1.1759936332702636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,1,power_law_1.01,1.2286016464233398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,1,power_law_1.01,1.2362815856933593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,1,power_law_1.01,1.3067008018493653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,1,power_law_1.01,1.316652774810791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,1,power_law_1.01,1.3523327827453613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,1,power_law_1.01,1.4182656288146973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,1,power_law_1.01,1.4499263763427734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,1,power_law_1.01,1.4905023574829102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,1,power_law_1.01,1.5697728157043458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,1,power_law_1.01,1.710246467590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,1,power_law_1.01,1.706719970703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,1,power_law_1.01,1.893471908569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,1,balanced,5.8639678955078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,1,power_law_1.01,1.9672191619873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,1,power_law_1.01,2.2492095947265627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,1,balanced,15.531168619791666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,1,power_law_1.01,2.553548812866211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,1,power_law_1.01,3.1069055557250977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,1,power_law_1.01,3.5745792388916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,1,power_law_1.01,4.680198287963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,1,balanced,14.556203206380209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,1,power_law_1.01,5.865337753295899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,1,power_law_1.01,7.111398315429687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,1,power_law_1.01,10.942752075195312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,1,power_law_1.01,19.639546203613282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,1,power_law_1.01,0.0701312005519867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,1,power_law_1.01,0.09542400240898133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,1,power_law_1.01,0.07986559867858886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,1,power_law_1.01,0.10549759864807129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,1,power_law_1.01,0.029151999950408937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,1,power_law_1.01,0.10094079971313477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,1,power_law_1.01,0.19795199632644653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,1,power_law_1.01,0.04928640127182007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,1,power_law_1.01,0.10965759754180908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,1,power_law_1.01,0.2971456050872803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,1,power_law_1.01,0.04837760031223297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,1,power_law_1.01,0.1773311972618103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,1,power_law_1.01,0.41415038108825686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,1,power_law_1.01,0.07534080147743225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,1,power_law_1.01,0.2373568058013916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,1,power_law_1.01,0.5873472213745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,1,power_law_1.01,0.310809588432312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,1,power_law_1.01,0.14362239837646484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,1,power_law_1.01,0.6170688152313233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,1,power_law_1.01,0.42743678092956544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,1,power_law_1.01,0.20352001190185548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,1,power_law_1.01,0.6523200035095215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,1,power_law_1.01,0.4481215953826904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,1,power_law_1.01,0.2561471939086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,1,power_law_1.01,0.6720064163208008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,1,power_law_1.01,0.47742719650268556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,1,power_law_1.01,0.3159679889678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,1,power_law_1.01,0.6990143775939941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,1,power_law_1.01,0.4810688018798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,1,power_law_1.01,0.3283967971801758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,1,power_law_1.01,0.7199808120727539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,1,power_law_1.01,0.5025728225708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,1,power_law_1.01,0.3347071886062622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,1,power_law_1.01,0.7371776103973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,1,power_law_1.01,0.5165887832641601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,1,power_law_1.01,0.35376639366149903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,1,power_law_1.01,0.751961612701416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,1,power_law_1.01,0.5294784069061279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,1,power_law_1.01,0.35800321102142335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,1,power_law_1.01,0.7956799983978271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,1,power_law_1.01,0.5536831855773926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,1,power_law_1.01,0.3805824041366577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,1,power_law_1.01,0.8197823524475097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,1,power_law_1.01,0.5731647968292236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,1,power_law_1.01,0.42928638458251955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,1,power_law_1.01,0.8659263610839844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,1,power_law_1.01,0.5905151844024659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,1,power_law_1.01,0.4007296085357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,1,power_law_1.01,0.952780818939209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,1,power_law_1.01,0.6320320129394531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,1,power_law_1.01,0.4170559883117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,1,power_law_1.01,0.7086016178131104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,1,power_law_1.01,0.981811237335205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,1,power_law_1.01,0.45209598541259766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.01,0.7500607967376709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,1,power_law_1.01,1.1543744087219239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,1,power_law_1.01,0.45934081077575684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.01,0.8746560096740723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,1,power_law_1.01,1.21527042388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.01,0.48696317672729494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.01,0.96627836227417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,1,power_law_1.01,1.4280511856079101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.01,1.1840703964233399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.01,0.5207295894622803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,1,power_law_1.01,1.7022527694702148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.01,1.4038528442382812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.01,0.5715648174285889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,1,power_law_1.01,2.2209087371826173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.01,1.8685184478759767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.01,1.3599743843078613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,1,power_law_1.01,2.713804817199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.01,2.3327615737915037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.01,1.6191871643066407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,1,power_law_1.01,3.6329151153564454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.01,3.2488449096679686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.01,0.9996864318847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,1,power_law_1.01,4.622041702270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.01,4.170342254638672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.01,1.1388287544250488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.01,1.5055744171142578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,1,power_law_1.01,5.5696063995361325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,1,power_law_1.01,5.036979293823242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.01,1.9210304260253905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,1,power_law_1.2,0.09303039908409119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,1,power_law_1.01,8.22375030517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,1,power_law_1.01,7.6331840515136715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,1,power_law_1.01,2.2443199157714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,1,power_law_1.2,0.14055039882659912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,1,power_law_1.01,3.332896041870117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,1,power_law_1.01,15.57019500732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,1,power_law_1.2,0.16228480339050294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,1,power_law_1.01,14.807539367675782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,1,power_law_1.2,0.3090431928634644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,1,power_law_1.01,6.2871551513671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,1,power_law_1.2,0.48869757652282714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,1,power_law_1.2,0.6543168067932129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,1,power_law_1.2,1.0612671852111817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,1,power_law_1.2,1.1343680381774903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,1,power_law_1.2,1.1778752326965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,1,power_law_1.2,1.230016040802002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,1,power_law_1.2,1.3086400032043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,1,power_law_1.2,1.3537343978881835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,1,power_law_1.2,1.425881576538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,1,power_law_1.2,0.07009919881820678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,1,power_law_1.2,1.4372735977172852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,1,power_law_1.2,0.0806335985660553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,1,power_law_1.2,0.09500160217285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,1,power_law_1.2,1.476038360595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,1,power_law_1.2,0.10016000270843506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,1,power_law_1.2,0.10728319883346557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,1,power_law_1.2,1.5571200370788574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,1,power_law_1.2,0.10758399963378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,1,power_law_1.2,0.18468480110168456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,1,power_law_1.2,1.7005119323730469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,1,power_law_1.2,0.1644863963127136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,1,power_law_1.2,0.029311999678611755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,1,power_law_1.2,0.27854080200195314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,1,power_law_1.2,1.8133184432983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,1,power_law_1.2,0.20718719959259033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,1,power_law_1.2,0.048467200994491574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,1,power_law_1.2,0.38047358989715574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,1,power_law_1.2,0.2845312118530273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,1,power_law_1.2,1.809164810180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,1,power_law_1.2,0.0474368005990982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,1,power_law_1.2,0.5665791988372803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,1,power_law_1.2,0.41434879302978517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,1,power_law_1.2,0.08042240142822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,1,power_law_1.2,2.002521514892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,1,power_law_1.2,0.5969727993011474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,1,power_law_1.2,0.13292800188064574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,1,power_law_1.2,0.43520641326904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,1,power_law_1.2,2.0646976470947265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,1,power_law_1.2,0.18211840391159057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,1,power_law_1.2,0.652569580078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,1,power_law_1.2,0.4626304149627686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,1,power_law_1.2,0.2824512004852295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,1,power_law_1.2,2.3587520599365233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,1,power_law_1.2,0.4688576221466064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,1,power_law_1.2,0.6616896152496338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,1,power_law_1.2,0.30234880447387696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,1,power_law_1.2,0.666918420791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,1,power_law_1.2,0.5000383853912354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,1,power_law_1.2,2.656902313232422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,1,power_law_1.2,0.3250816106796265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,1,power_law_1.2,0.7130112171173095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,1,power_law_1.2,0.5137087821960449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,1,power_law_1.2,0.32877440452575685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,1,power_law_1.2,0.758022403717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,1,power_law_1.2,3.21978874206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,1,power_law_1.2,0.3387840032577515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,1,power_law_1.2,0.5363647937774658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,1,power_law_1.2,0.7967679977416993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,2,balanced,0.0276853342851003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,1,power_law_1.2,0.3674688100814819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,1,power_law_1.2,3.7668991088867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,1,power_law_1.2,0.5627583980560302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,1,power_law_1.2,0.8199808120727539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,1,power_law_1.2,0.38670079708099364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,2,balanced,0.033887999753157295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,1,power_law_1.2,0.43168001174926757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,1,power_law_1.2,0.5833663940429688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,1,power_law_1.2,0.8589311599731445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,1,power_law_1.2,4.883539199829102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,2,balanced,0.04966933528582255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,1,power_law_1.2,0.4170688152313232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,2,balanced,0.07850666840871175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,1,power_law_1.2,0.9095104217529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,1,power_law_1.2,0.6040448188781739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,1,power_law_1.2,0.4288447856903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,2,balanced,0.11921599507331848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,1,power_law_1.2,6.135776138305664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,1,power_law_1.2,0.4720448017120361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,1,power_law_1.2,1.0111104011535645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,1,power_law_1.2,0.657587194442749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,2,balanced,0.12082667152086894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,1,power_law_1.2,0.4882368087768555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,1,power_law_1.2,7.293580627441406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,1,power_law_1.2,1.0436863899230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,1,power_law_1.2,0.7394559860229493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,2,balanced,0.12099732955296834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.2,0.5201280117034912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,1,power_law_1.2,1.2107647895812987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,1,power_law_1.2,11.076038360595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,2,balanced,0.12316800157229106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.2,0.7807040214538574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.2,0.5395391941070556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,1,power_law_1.2,1.276524829864502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,2,balanced,0.1223413348197937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.2,0.9011903762817383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.2,0.5890175819396972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,1,power_law_1.2,19.862886047363283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,2,balanced,0.12190933028856914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,1,power_law_1.2,1.502617645263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.2,0.9934399604797364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.2,1.4302528381347657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,2,balanced,0.12215466300646464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.2,1.6742975234985351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.2,1.2259200096130372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,1,power_law_1.2,1.7660032272338868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,2,balanced,0.12469866871833801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.2,1.0546303749084474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.2,1.4425087928771974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,2,balanced,0.12531733512878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,1,power_law_1.2,2.286092758178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.2,1.250380802154541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,2,balanced,0.12850133577982584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.2,1.9013824462890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,1,power_law_1.2,2.763667106628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.2,1.5645888328552247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,2,balanced,0.13313600420951843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.2,2.3579391479492187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,1,power_law_1.2,3.7184574127197267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.2,1.9500415802001954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,2,balanced,0.13479466239611307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.2,3.2643009185791017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,1,power_law_1.2,4.693427276611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,1,power_law_1.2,2.279532814025879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,2,balanced,0.1387999951839447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.2,4.198624038696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,1,power_law_1.2,5.643526458740235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,2,balanced,0.14570666352907816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,1,power_law_1.2,3.4078720092773436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,1,power_law_1.2,5.101107025146485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,2,balanced,0.1555466651916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,1,power_law_1.2,8.264185333251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,1,power_law_1.2,6.334553527832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,1,power_law_1.2,7.73694076538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,2,balanced,0.17892267306645712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,1,power_law_1.2,15.61071319580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,1,power_law_1.2,14.9902587890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,2,balanced,0.18987733125686646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,2,balanced,0.2336426575978597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,2,balanced,0.2684160073598226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,2,balanced,0.36882134278615314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,2,balanced,0.43254931767781574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,2,balanced,0.6029066642125448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,2,balanced,0.781381368637085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,2,balanced,0.9518453280131022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,2,balanced,1.4893439610799153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,2,balanced,0.04167466859022776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,2,balanced,0.03976533313592275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,2,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,2,balanced,0.04490133126576742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,2,balanced,0.019424000134070713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,2,balanced,0.055573334296544395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,2,balanced,0.033759998778502144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,2,balanced,0.02107733239730199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,2,balanced,0.07877333462238312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,2,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,2,balanced,0.02362666775782903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,2,balanced,0.08064533273379008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,2,balanced,0.058245331048965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,2,balanced,0.029333333174387615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,2,balanced,0.0830026666323344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,2,balanced,0.08335999647776286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,2,balanced,0.04725333551565806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,2,balanced,0.08290666838486989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,2,balanced,0.08319999774297078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,2,balanced,0.0491946687301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,2,balanced,2.8641227086385093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,2,balanced,0.08311466872692108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,2,balanced,0.08505599697430928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,2,balanced,0.04986133178075155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,2,balanced,0.08602133393287659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,2,balanced,0.08541867136955261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,2,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,2,balanced,0.0860746701558431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,2,balanced,0.08673600355784099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,2,balanced,0.053632001082102455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,2,balanced,0.088837335507075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,2,balanced,0.08835200468699138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,2,balanced,0.050069332122802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,2,balanced,0.08891200025876363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,2,balanced,0.08867733677228291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,2,balanced,0.053717335065205894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,2,balanced,0.09276800354321797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,2,balanced,0.09307199716567993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,2,balanced,0.059903999169667564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,2,balanced,0.09867200255393982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,2,balanced,0.0925439993540446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,2,balanced,0.060453335444132485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,2,balanced,0.10088533163070679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,2,balanced,0.09733333190282185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,2,balanced,0.060533334811528526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,2,balanced,0.10331199566523235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,2,balanced,0.10599467158317566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,2,balanced,0.07828266421953838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,2,balanced,0.11658133069674174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,2,balanced,0.1070240040620168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,2,balanced,0.07673066854476929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,2,balanced,0.12999467055002847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,2,balanced,0.10987200339635213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,2,balanced,0.07606933514277141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,2,balanced,0.1497066617012024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,2,balanced,0.12291199962298076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,2,balanced,0.11212266484896342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,2,balanced,0.17246933778127035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,2,balanced,0.136543999115626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,2,balanced,0.1190880040327708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,2,balanced,0.159360001484553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,2,balanced,0.2327679991722107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,2,balanced,0.13985066612561545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,2,balanced,0.18095999956130981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,2,balanced,0.14597333470980325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,2,balanced,0.2745866576830546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,2,balanced,0.23918400208155313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,2,balanced,0.14190933108329773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,2,balanced,0.27957866589228314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,2,balanced,0.39868799845377606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,2,balanced,0.14481600125630698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,2,balanced,0.4036213159561157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,2,balanced,0.2571679949760437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,2,balanced,0.4978239933649699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,2,balanced,0.5066080093383789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,2,balanced,0.25327465931574505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,2,balanced,0.7218453089396158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,2,balanced,0.7323199907938639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,2,balanced,0.36315735181172687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,2,balanced,0.9447466532389323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,2,balanced,0.9608480135599772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,2,balanced,0.473311980565389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,2,balanced,1.171610673268636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,2,balanced,1.188970645268758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,2,balanced,0.5846986770629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,2,power_law_1.01,0.030112001299858093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,2,balanced,1.842682679494222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,2,power_law_1.01,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,2,power_law_1.01,0.04351359903812409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,2,power_law_1.01,0.061408001184463504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,2,power_law_1.01,0.07397119998931885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,2,balanced,1.853530724843343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,2,balanced,0.9181599617004395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,2,power_law_1.01,0.10115840435028076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,2,power_law_1.01,0.037574398517608645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,2,power_law_1.01,0.10735360383987427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,2,power_law_1.01,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,2,power_law_1.01,0.11214720010757447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,2,power_law_1.01,0.037299200892448425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,2,power_law_1.01,0.11237119436264038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,2,power_law_1.01,0.04857600033283234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,2,power_law_1.01,0.11399680376052856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,2,power_law_1.01,0.05727360248565674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,2,power_law_1.01,0.12088320255279542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,2,power_law_1.01,0.07571840286254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,2,power_law_1.01,0.1251263976097107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,2,power_law_1.01,0.07736960053443909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,2,power_law_1.01,0.04172160029411316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,2,power_law_1.01,0.1273344039916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,2,power_law_1.01,0.018080000579357148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,2,power_law_1.01,0.07898880243301391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,2,power_law_1.01,0.1257151961326599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,2,power_law_1.01,0.03828479945659637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,2,power_law_1.01,0.018719999492168425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,2,power_law_1.01,0.08328319787979126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,2,power_law_1.01,0.1344704031944275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,2,power_law_1.01,0.042822399735450746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,2,power_law_1.01,0.02332800030708313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,2,balanced,3.585125287373861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,2,power_law_1.01,0.08378880023956299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,2,power_law_1.01,0.136735999584198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,2,power_law_1.01,0.05012480020523071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,2,power_law_1.01,0.026060798764228822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,2,power_law_1.01,0.08775039911270141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,2,power_law_1.01,0.13994239568710326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,2,power_law_1.01,0.02948479950428009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,2,power_law_1.01,0.05720959901809693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,2,power_law_1.01,0.09171199798583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,2,power_law_1.01,0.15646719932556152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,2,power_law_1.01,0.0395583987236023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,2,balanced,3.641962687174479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,2,power_law_1.01,0.07265920042991639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,2,balanced,1.8151146570841472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,2,power_law_1.01,0.09370239973068237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,2,power_law_1.01,0.07256320118904114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,2,power_law_1.01,0.1700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,2,power_law_1.01,0.040934398770332336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,2,power_law_1.01,0.09562240242958069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,2,power_law_1.01,0.0766975998878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,2,power_law_1.01,0.20755200386047362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,2,power_law_1.01,0.042259201407432556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,2,power_law_1.01,0.10349440574645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,2,power_law_1.01,0.07724159955978394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,2,power_law_1.01,0.2099071979522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,2,power_law_1.01,0.044819200038909913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,2,power_law_1.01,0.1086016058921814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,2,power_law_1.01,0.2554816007614136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,2,power_law_1.01,0.08106240034103393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,2,power_law_1.01,0.04488320052623749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,2,power_law_1.01,0.11255680322647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,2,power_law_1.01,0.3076224088668823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,2,power_law_1.01,0.08309119939804077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,2,power_law_1.01,0.04643200039863586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,2,power_law_1.01,0.12787840366363526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,2,power_law_1.01,0.4081535816192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,2,power_law_1.01,0.04776960015296936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,2,power_law_1.01,0.08545920252799988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,2,power_law_1.01,0.1385983943939209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,2,power_law_1.01,0.5071936130523682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,2,power_law_1.01,0.04976640045642853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,2,power_law_1.01,0.08792960047721862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,2,power_law_1.01,0.17320319414138793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,2,power_law_1.01,0.7009471893310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,2,power_law_1.01,0.05085440278053284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,2,power_law_1.01,0.09381759762763978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,2,power_law_1.01,0.19996800422668456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,2,power_law_1.01,0.8972672462463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,2,power_law_1.01,0.05604479908943176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,2,power_law_1.01,0.09845119714736938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,2,power_law_1.01,0.26481919288635253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,2,power_law_1.01,0.05446400046348572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,2,power_law_1.01,0.1051584005355835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,2,power_law_1.01,1.1160639762878417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,2,power_law_1.01,0.32444798946380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,2,power_law_1.01,0.05627520084381103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,2,power_law_1.01,0.10496000051498414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,2,power_law_1.01,1.5981056213378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,2,power_law_1.01,0.44780797958374025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,2,power_law_1.01,0.09876480102539062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,2,power_law_1.01,0.11717120409011841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,2,power_law_1.01,0.55000319480896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,2,power_law_1.01,3.325785446166992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,2,power_law_1.01,0.13786879777908326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,2,power_law_1.01,0.10890239477157593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,2,power_law_1.01,0.8208895683288574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,2,power_law_1.01,0.16161279678344725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,2,power_law_1.01,0.10058879852294922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,2,power_law_1.01,1.1100288391113282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,2,power_law_1.01,0.19568639993667603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,2,power_law_1.01,0.10673279762268066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,2,power_law_1.01,1.3622719764709472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,2,power_law_1.01,0.25559680461883544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,2,power_law_1.01,0.12359039783477783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,2,power_law_1.01,2.0590911865234376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,2,power_law_1.01,0.3074559926986694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,2,power_law_1.01,0.14670079946517944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,2,power_law_1.01,0.44089598655700685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,2,power_law_1.01,0.18481279611587526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,2,power_law_1.01,4.08155517578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,2,power_law_1.01,0.5582079887390137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,2,power_law_1.01,0.2247999906539917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,2,power_law_1.01,0.864192008972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,2,power_law_1.01,0.30599040985107423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,2,power_law_1.01,1.0613632202148438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,2,power_law_1.01,0.38601601123809814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,2,power_law_1.01,1.3110207557678222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,2,power_law_1.01,0.4897791862487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,2,power_law_1.01,1.8829824447631835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,2,power_law_1.01,0.7550528049468994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,2,power_law_1.01,3.9385791778564454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,2,power_law_1.01,1.4456000328063965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,2,power_law_1.2,0.030854400992393494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,2,power_law_1.2,0.03516800105571747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,2,power_law_1.2,0.040838399529457094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,2,power_law_1.2,0.052691197395324706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,2,power_law_1.2,0.0685696005821228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,2,power_law_1.2,0.10274560451507568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,2,power_law_1.2,0.10876799821853637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,2,power_law_1.2,0.11270400285720825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,2,power_law_1.2,0.11676160097122193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,2,power_law_1.2,0.11648639440536498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,2,power_law_1.2,0.12396160364151002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,2,power_law_1.2,0.12625919580459594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,2,power_law_1.2,0.12608640193939208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,2,power_law_1.2,0.12726399898529053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,2,power_law_1.2,0.13731199502944946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,2,power_law_1.2,0.141702401638031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,2,power_law_1.2,0.14268800020217895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,2,power_law_1.2,0.15590399503707886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,2,power_law_1.2,0.17367039918899535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,2,power_law_1.2,0.21452798843383789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,2,power_law_1.2,0.21721599102020264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,2,power_law_1.2,0.2666304111480713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,2,power_law_1.2,0.3221951961517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,2,power_law_1.2,0.428934383392334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,2,power_law_1.2,0.5450431823730468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,2,power_law_1.2,0.7755072116851807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,2,power_law_1.2,0.9600576400756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,2,power_law_1.2,1.254047966003418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,2,power_law_1.2,1.7075199127197265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,2,power_law_1.2,3.574233627319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,2,power_law_1.2,0.03692159950733185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,2,power_law_1.2,0.03317759931087494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,2,power_law_1.2,0.03585920035839081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,2,power_law_1.2,0.046112000942230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,2,power_law_1.2,0.05278720259666443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,2,power_law_1.2,0.07619199752807618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,2,power_law_1.2,0.07854080200195312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,2,power_law_1.2,0.08163840174674988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,2,power_law_1.2,0.08212479948997498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,2,power_law_1.2,0.08439679741859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,2,power_law_1.2,0.018137599527835845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,2,power_law_1.2,0.04188160002231598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,2,power_law_1.2,0.08672000169754028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,2,power_law_1.2,0.019385600090026857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,2,power_law_1.2,0.037785598635673524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,2,power_law_1.2,0.09099519848823548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,2,power_law_1.2,0.022073599696159362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,2,power_law_1.2,0.04408960044384003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,2,power_law_1.2,0.09553279876708984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,2,power_law_1.2,0.02550399899482727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,2,power_law_1.2,0.04979200065135956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,2,power_law_1.2,0.09667199850082397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,2,power_law_1.2,0.028019198775291444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,2,power_law_1.2,0.055276799201965335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,2,power_law_1.2,0.10487680435180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,2,power_law_1.2,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,2,power_law_1.2,0.07348480224609374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,2,power_law_1.2,0.11401599645614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,2,power_law_1.2,0.04229120016098022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,2,power_law_1.2,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,2,power_law_1.2,0.11668479442596436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,2,power_law_1.2,0.04322560131549835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,2,power_law_1.2,0.07783039808273315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,2,power_law_1.2,0.1345919966697693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,2,power_law_1.2,0.04556800127029419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,2,power_law_1.2,0.0787775993347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,2,power_law_1.2,0.14520319700241088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,2,power_law_1.2,0.04417920112609863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,2,power_law_1.2,0.08179839849472045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,2,power_law_1.2,0.1803455948829651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,2,power_law_1.2,0.04764800071716309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,2,power_law_1.2,0.0858560025691986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,2,power_law_1.2,0.20413439273834227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,2,power_law_1.2,0.04828799962997436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,2,power_law_1.2,0.08642560243606567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,2,power_law_1.2,0.2728895902633667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,2,power_law_1.2,0.049747198820114136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,2,power_law_1.2,0.08992000222206116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,2,power_law_1.2,0.334937596321106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,2,power_law_1.2,0.04951040148735046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,2,power_law_1.2,0.09372159838676453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,2,power_law_1.2,0.4721856117248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,2,power_law_1.2,0.053907197713851926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,2,power_law_1.2,0.09909759759902954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,2,power_law_1.2,0.5837503910064697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,2,power_law_1.2,0.05490559935569763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,2,power_law_1.2,0.10698239803314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,2,power_law_1.2,0.8937472343444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,2,power_law_1.2,0.058361601829528806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,2,power_law_1.2,0.10861439704895019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,2,power_law_1.2,1.1187071800231934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,2,power_law_1.2,0.10123519897460938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,2,power_law_1.2,0.1269055962562561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,2,power_law_1.2,1.4486207962036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,2,power_law_1.2,0.10826879739761353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,2,power_law_1.2,0.14161280393600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,2,power_law_1.2,2.1910272598266602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,2,power_law_1.2,0.10314879417419434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,2,power_law_1.2,0.1679360032081604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,1,balanced,0.03367999941110611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,2,power_law_1.2,0.10896639823913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,2,power_law_1.2,0.20277121067047119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,2,power_law_1.2,4.732870483398438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,1,balanced,0.03552533437808355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,2,power_law_1.2,0.13000320196151732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,2,power_law_1.2,0.2603327989578247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,1,balanced,0.050053333242734276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,2,power_law_1.2,0.14264960289001466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,1,balanced,0.0804746647675832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,2,power_law_1.2,0.31499519348144533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,2,power_law_1.2,0.18997759819030763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,1,balanced,0.123690664768219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,2,power_law_1.2,0.4548351764678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,1,balanced,0.12542933225631714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,2,power_law_1.2,0.24106240272521973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,1,balanced,0.12709866960843405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,2,power_law_1.2,0.6240960121154785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,2,power_law_1.2,0.331660795211792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,1,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,1,balanced,0.12640000383059183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,2,power_law_1.2,0.9233983993530274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,2,power_law_1.2,0.4131775856018066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,1,balanced,0.03752533346414566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,1,balanced,0.1281760036945343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,2,power_law_1.2,1.1755328178405762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,2,power_law_1.2,0.5021696090698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,1,balanced,0.04182933270931244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,1,balanced,0.12773866454760233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,2,power_law_1.2,1.405081558227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,2,power_law_1.2,0.7569471836090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,1,balanced,0.13130666812260947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,1,balanced,0.01727466657757759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,1,balanced,0.06237866481145223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,1,balanced,0.05607999861240387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,2,power_law_1.2,2.186854362487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,1,balanced,0.13337066769599915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,2,power_law_1.2,1.591443157196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,1,balanced,0.019317333896954853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,1,balanced,0.08896000186602275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,1,balanced,0.06367999811967213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,1,balanced,0.1344106694062551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,2,power_law_1.2,4.051731109619141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,1,balanced,0.021216000119845074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,1,balanced,0.08835200468699138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,1,balanced,0.1402186652024587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,1,balanced,0.07449600100517273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,1,balanced,0.02443733314673106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,1,balanced,0.14522133270899454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,1,balanced,0.09145599603652954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,1,balanced,0.09851732850074768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,1,balanced,0.032042667269706726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,1,balanced,0.14455466469128928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,1,balanced,0.1502346694469452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,1,balanced,0.09211200475692749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,1,balanced,0.03278400003910065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,1,balanced,0.15944000085194907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,1,balanced,0.13776533802350363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,1,balanced,0.09379733602205913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,1,balanced,0.033530667424201965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,1,balanced,0.171450674533844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,1,balanced,0.13778133193651834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,1,balanced,0.0951039989789327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,1,balanced,0.035349334279696144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,1,balanced,0.18595200777053833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,1,balanced,0.13547199964523315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,1,balanced,0.09715732932090759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,1,balanced,0.03760000069936117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,1,balanced,0.22659200429916382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,1,balanced,0.13781332969665527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,1,balanced,0.1013706624507904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,1,balanced,0.03775999943415324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,1,balanced,0.1388213336467743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,1,balanced,0.10266133149464925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,1,balanced,0.24791999657948813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,1,balanced,0.03774933268626531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,1,balanced,0.14044800400733948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,1,balanced,0.11091199517250061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,1,balanced,0.03968533376852671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,1,balanced,0.3256266713142395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,1,balanced,0.14697600404421488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,1,balanced,0.04142933338880539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,1,balanced,0.12236266334851582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,1,balanced,0.04318933188915253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,1,balanced,0.1421280006567637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,1,balanced,0.3800533215204875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,1,balanced,0.12590932846069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,1,balanced,0.045925334095954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,1,balanced,0.14969600240389505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,1,balanced,0.13359466195106506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,1,balanced,0.04573333263397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,1,balanced,0.5271733204523722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,1,balanced,0.15828266739845276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,1,balanced,0.15230400363604227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,1,balanced,0.04874666531880697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,1,balanced,0.16223999857902527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,1,balanced,0.17493865887324014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,1,balanced,0.06644266843795776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,1,balanced,0.1705333391825358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,1,balanced,0.660207986831665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,1,balanced,0.0671253353357315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,1,balanced,0.21112000942230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,1,balanced,0.19783467054367065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,1,balanced,0.08790399630864461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,1,balanced,0.24761066834131876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,1,balanced,0.22217599550882974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,1,balanced,0.08649599552154541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,1,balanced,0.942624012629191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,1,balanced,0.26716800530751544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,1,balanced,0.35183465480804443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,1,balanced,0.11271466811498006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,1,power_law_1.01,0.033369600772857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,1,balanced,0.3123253385225932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,1,balanced,0.4304266770680745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,1,balanced,0.11495467027028401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,1,power_law_1.01,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,1,balanced,0.47565333048502606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,1,power_law_1.01,0.044435200095176694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,1,balanced,1.2364319960276287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,1,balanced,0.6234613259633383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,1,power_law_1.01,0.06259199976921082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,1,balanced,0.18439465761184692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,1,balanced,0.5752960046132406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,1,power_law_1.01,0.07739520072937012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,1,power_law_1.01,0.11036159992218017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,1,power_law_1.01,0.11323519945144653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,1,balanced,0.8067946434020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,1,balanced,0.19548267126083374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,1,power_law_1.01,0.11793919801712036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,1,balanced,0.8330720265706381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,1,power_law_1.01,0.12383999824523925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,1,power_law_1.01,0.12441600561141967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,1,balanced,1.531786600748698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,1,power_law_1.01,0.12999039888381958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,1,balanced,1.0869653224945068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,1,balanced,1.1720213095347087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,1,balanced,0.2759360074996948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,1,power_law_1.01,0.1307584047317505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,1,power_law_1.01,0.13581440448760987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,1,power_law_1.01,0.1358847975730896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,1,power_law_1.01,0.14300800561904908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,1,balanced,1.5711466471354167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,1,power_law_1.01,0.14394240379333495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,1,power_law_1.01,0.15433599948883056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,1,balanced,1.5379039446512859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,1,balanced,0.35716267426808673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,1,power_law_1.01,0.1737920045852661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,1,power_law_1.01,0.19148160219192506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,1,balanced,2.313546657562256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,1,power_law_1.01,0.2354111909866333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,1,balanced,2.055562655131022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,1,power_law_1.01,0.25347840785980225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,1,power_law_1.01,0.3313024044036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,1,power_law_1.01,0.40417919158935545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,1,balanced,1.90938138961792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,1,balanced,0.43729599316914874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,1,power_law_1.01,0.5429183959960937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,1,power_law_1.01,0.6861248016357422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,1,balanced,2.5276479721069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,1,power_law_1.01,0.9665535926818848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,1,power_law_1.01,1.2435711860656737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,1,power_law_1.01,1.5354880332946776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,1,power_law_1.01,2.3144575119018556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,1,power_law_1.01,4.469728088378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,1,balanced,3.018618583679199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,1,balanced,0.6786293188730875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,1,balanced,3.9474452336629233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,1,balanced,4.521018664042155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,1,balanced,5.898799896240234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,1,balanced,1.3316586812337239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,1,balanced,7.643434524536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,1,power_law_1.01,0.03449600040912628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,1,power_law_1.01,0.033907198905944826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,1,power_law_1.01,0.03940480053424835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,1,power_law_1.01,0.048441600799560544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,1,power_law_1.01,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,1,power_law_1.01,0.07794560194015503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,1,power_law_1.01,0.08176640272140503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,1,power_law_1.01,0.08486400246620178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,1,power_law_1.01,0.08777599930763244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,1,power_law_1.01,0.08989440202713013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,1,power_law_1.01,0.0927232027053833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,1,power_law_1.01,0.0976256012916565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,1,power_law_1.01,0.10291839838027954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,1,power_law_1.01,0.11030399799346924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,1,power_law_1.01,0.11792639493942261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,1,power_law_1.01,0.12583680152893068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,1,power_law_1.01,0.13458559513092042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,1,power_law_1.01,0.15783679485321045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,1,power_law_1.01,0.052665597200393675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,1,power_law_1.01,0.17770880460739136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,1,power_law_1.01,0.05614079833030701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,1,power_law_1.01,0.2214143991470337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,1,power_law_1.01,0.06812160015106201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,1,power_law_1.01,0.26031999588012694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,1,power_law_1.01,0.08284800052642823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,1,power_law_1.01,0.3541759967803955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,1,power_law_1.01,0.09249920248985291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,1,power_law_1.01,0.44599041938781736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,1,power_law_1.01,0.017900800704956053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,1,power_law_1.01,0.11890560388565063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,1,power_law_1.01,0.634227180480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,1,power_law_1.01,0.018028800189495087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,1,power_law_1.01,0.12863999605178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,1,power_law_1.01,0.8156096458435058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,1,power_law_1.01,0.01976960003376007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,1,power_law_1.01,0.12959359884262084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,1,power_law_1.01,1.1738752365112304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,1,power_law_1.01,0.023839999735355378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,1,power_law_1.01,0.13248000144958497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,1,power_law_1.01,1.5364416122436524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,1,power_law_1.01,0.02913280129432678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,1,power_law_1.01,0.13319040536880494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,1,power_law_1.01,1.8990272521972655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,1,power_law_1.01,0.03705599904060364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,1,power_law_1.01,0.1370687961578369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,1,power_law_1.01,2.987558364868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,1,power_law_1.01,0.03960959911346436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,1,power_law_1.01,0.13936640024185182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,1,power_law_1.01,5.837196731567383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,1,power_law_1.01,0.04366079866886139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,1,power_law_1.01,0.141484797000885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,1,power_law_1.01,0.045075199007987975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,1,power_law_1.01,0.1465407967567444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,1,power_law_1.01,0.044588801264762876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,1,power_law_1.01,0.15541759729385377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,1,power_law_1.01,0.04705919921398163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,1,power_law_1.01,0.1598847985267639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,1,power_law_1.01,0.046854400634765626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,1,power_law_1.01,0.1732480049133301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,1,power_law_1.01,0.04856959879398346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,1,power_law_1.01,0.20821120738983154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,1,power_law_1.01,0.0494592010974884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,1,power_law_1.01,0.23608319759368895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,1,power_law_1.01,0.05106559991836548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,1,power_law_1.01,0.2951488018035889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,1,power_law_1.01,0.05287039875984192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,1,power_law_1.01,0.3541120052337646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,1,power_law_1.01,0.05727999806404114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,1,power_law_1.01,0.4765632152557373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,1,power_law_1.01,0.06817920207977295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,1,power_law_1.01,0.5970176219940185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,1,power_law_1.01,0.07268480062484742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,1,power_law_1.01,0.8356863975524902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,1,power_law_1.01,0.09237120151519776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,1,power_law_1.01,1.065561580657959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,1,power_law_1.01,0.09627519845962525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,1,power_law_1.01,1.5349184036254884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,1,power_law_1.01,0.12101119756698608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,1,power_law_1.01,1.9948352813720702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,1,power_law_1.01,0.14659839868545532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,1,power_law_1.01,2.4580223083496096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,1,power_law_1.01,0.18201600313186644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,1,power_law_1.01,3.8424896240234374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,1,power_law_1.01,0.22821760177612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,1,power_law_1.01,7.5428611755371096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,1,power_law_1.01,0.3064127922058105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,1,power_law_1.01,0.3856895923614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,1,power_law_1.01,0.468671989440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,1,power_law_1.01,0.7109119892120361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,1,power_law_1.01,1.3560064315795899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,1,power_law_1.2,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,1,power_law_1.2,0.035206401348114015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,1,power_law_1.2,0.04191359877586365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,1,power_law_1.2,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,1,power_law_1.2,0.056992000341415404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,1,power_law_1.2,0.03576320111751556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,1,power_law_1.2,0.06763520240783691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,1,power_law_1.2,0.03932160139083862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,1,power_law_1.2,0.11004159450531006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,1,power_law_1.2,0.04867840111255646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,1,power_law_1.2,0.1150272011756897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,1,power_law_1.2,0.05431680083274841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,1,power_law_1.2,0.12133760452270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,4,balanced,0.0888159970442454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,1,power_law_1.2,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,1,power_law_1.2,0.12303999662399293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,4,balanced,0.08870933453241985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,1,power_law_1.2,0.08324480056762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,1,power_law_1.2,0.1265920042991638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,4,balanced,0.14825600385665894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,1,power_law_1.2,0.08497920036315917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,1,power_law_1.2,0.1292672038078308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,4,balanced,0.13914133111635843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,1,power_law_1.2,0.08796160221099854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,1,power_law_1.2,0.13220479488372802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,1,power_law_1.2,0.053913599252700804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,4,balanced,0.14198399583498636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,1,power_law_1.2,0.09000319838523865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,1,power_law_1.2,0.13978240489959717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,1,power_law_1.2,0.05566080212593079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,4,balanced,0.1405173341433207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,1,power_law_1.2,0.09568639993667602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,1,power_law_1.2,0.13778560161590575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,4,balanced,0.14199999968210855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,1,power_law_1.2,0.09966719746589661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,1,power_law_1.2,0.06573439836502075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,1,power_law_1.2,0.0166143998503685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,1,power_law_1.2,0.14394240379333495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,4,balanced,0.14272000392278036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,1,power_law_1.2,0.10312960147857667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,1,power_law_1.2,0.07838720083236694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,1,power_law_1.2,0.017113600671291352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,1,power_law_1.2,0.14931199550628663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,4,balanced,0.1514346698919932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,1,power_law_1.2,0.11182719469070435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,4,balanced,0.15176533659299216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,1,power_law_1.2,0.15686399936676027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,1,power_law_1.2,0.12048640251159667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,1,power_law_1.2,0.08736640214920044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,1,power_law_1.2,0.01964160054922104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,4,balanced,0.14797332882881165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,1,power_law_1.2,0.1784832000732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,1,power_law_1.2,0.12621439695358277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,4,balanced,0.14657066265741983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,1,power_law_1.2,0.19733760356903077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,1,power_law_1.2,0.11902719736099243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,1,power_law_1.2,0.023219199478626253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,1,power_law_1.2,0.13537919521331787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,4,balanced,0.14621866742769876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,1,power_law_1.2,0.2387455940246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,1,power_law_1.2,0.1291327953338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,1,power_law_1.2,0.028505599498748778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,1,power_law_1.2,0.15777280330657958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,4,balanced,0.15637333194414774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,1,power_law_1.2,0.26199679374694823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,1,power_law_1.2,0.13175040483474731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,4,balanced,0.16718933979670206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,1,power_law_1.2,0.0380160003900528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,1,power_law_1.2,0.1778048038482666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,1,power_law_1.2,0.3341824054718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,1,power_law_1.2,0.1322368025779724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,4,balanced,0.16681599617004395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,1,power_law_1.2,0.04036479890346527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,1,power_law_1.2,0.22334721088409423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,1,power_law_1.2,0.40536961555480955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,1,power_law_1.2,0.1361407995223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,4,balanced,0.17933867375055948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,1,power_law_1.2,0.04322560131549835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,1,power_law_1.2,0.14088959693908693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,1,power_law_1.2,0.26139519214630125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,1,power_law_1.2,0.54683518409729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,4,balanced,0.19850667317708334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,1,power_law_1.2,0.043289598822593686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,1,power_law_1.2,0.13941119909286498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,1,power_law_1.2,0.35653119087219237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,1,power_law_1.2,0.6927743911743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,1,power_law_1.2,0.14391679763793946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,1,power_law_1.2,0.44929919242858884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,4,balanced,0.2222506602605184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,1,power_law_1.2,0.04434559941291809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,4,balanced,0.2747146685918172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,1,power_law_1.2,0.971168041229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,1,power_law_1.2,0.14766720533370972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,1,power_law_1.2,0.6345151901245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,1,power_law_1.2,0.0458624005317688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,4,balanced,0.29522132873535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,1,power_law_1.2,1.2521408081054688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,1,power_law_1.2,0.15827840566635132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,1,power_law_1.2,0.813804817199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,1,power_law_1.2,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,4,balanced,0.39472532272338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,1,power_law_1.2,1.5419327735900878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,1,power_law_1.2,0.16421120166778563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,1,power_law_1.2,1.1706879615783692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,1,power_law_1.2,0.047884801030159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,1,power_law_1.2,1.5281472206115723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,4,balanced,0.4740159908930461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,1,power_law_1.2,2.3168703079223634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,1,power_law_1.2,0.17709439992904663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,1,power_law_1.2,0.04947839975357056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,1,power_law_1.2,4.439443206787109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,1,power_law_1.2,0.20872321128845214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,1,power_law_1.2,1.8841279983520507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,1,power_law_1.2,0.23843839168548583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,4,balanced,0.6779733498891195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,1,power_law_1.2,0.05062400102615357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,1,power_law_1.2,2.976358413696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,1,power_law_1.2,0.2969536066055298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,4,balanced,0.880944013595581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,1,power_law_1.2,0.05392640233039856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,1,power_law_1.2,5.81091194152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,1,power_law_1.2,0.35387520790100097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,4,balanced,1.2870933214823406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,1,power_law_1.2,0.056524801254272464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,1,power_law_1.2,0.47450881004333495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,1,power_law_1.2,0.06858239769935608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,4,balanced,1.7105706532796223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,1,power_law_1.2,0.5917695999145508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,1,power_law_1.2,0.07274240255355835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,1,power_law_1.2,0.8272192001342773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,4,balanced,2.0850346883138022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,1,power_law_1.2,0.09338880181312562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,1,power_law_1.2,1.0594112396240234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,1,power_law_1.2,0.09466239809989929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,1,power_law_1.2,1.5227264404296874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,4,balanced,3.3845227559407554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,1,power_law_1.2,0.12238080501556396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,1,power_law_1.2,1.9784896850585938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,1,power_law_1.2,0.14982399940490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,1,power_law_1.2,2.4447999954223634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,1,power_law_1.2,0.1858944058418274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,1,power_law_1.2,3.817087936401367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,4,balanced,6.590437571207683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,1,power_law_1.2,0.2247999906539917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,1,power_law_1.2,7.5211647033691404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,1,power_law_1.2,0.3087296009063721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,1,power_law_1.2,0.3896320104598999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,1,power_law_1.2,0.46851201057434083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,1,power_law_1.2,0.7143936157226562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,1,power_law_1.2,1.365516757965088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,4,balanced,0.06322666505972545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,4,balanced,0.06396799782911937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,4,balanced,0.09055466453234355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,4,balanced,0.09145599603652954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,4,balanced,0.09099733829498291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,4,balanced,0.0920853316783905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,4,balanced,0.09430932998657227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,4,balanced,0.0921493371327718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,4,balanced,0.09203733007113139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,4,balanced,0.09335466225941975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,4,balanced,0.0955573320388794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,4,balanced,0.09993599851926167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,4,balanced,0.09947199622790019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,4,balanced,0.10260799527168274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,4,balanced,0.11422933141390483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,4,balanced,0.11507733662923177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,4,balanced,0.12140799562136333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,4,balanced,0.13800533612569174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,4,balanced,0.1458453337351481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,4,balanced,0.18688533703486124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,4,balanced,0.21364800135294595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,4,balanced,0.28725866476694745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,4,balanced,0.3426719903945923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,4,balanced,0.058037335673967995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,4,balanced,0.4850240151087443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,4,balanced,0.05762133498986562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,4,balanced,0.07183466851711273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,4,balanced,0.6231093406677246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,4,balanced,0.035189333061377205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,4,balanced,0.07709866762161255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,4,balanced,0.9052373568216959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,4,balanced,0.033413333197434746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,4,balanced,0.07551466425259908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,4,balanced,0.03331200033426285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,4,balanced,1.1531626383463542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,4,balanced,0.0786186655362447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,4,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,4,balanced,0.07894933223724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,4,balanced,0.038058665891488395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,4,balanced,1.4347200393676758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,4,balanced,0.08016000191370647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,4,balanced,0.03956266740957896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,4,balanced,0.07860800127188365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,4,balanced,0.05871999760468801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,4,balanced,2.2760799725850425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,4,balanced,0.07978133360544841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,4,balanced,0.046021332343419395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,4,balanced,0.0800906668106715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,4,balanced,0.09713600079218547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,4,balanced,0.08043733239173889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,4,balanced,0.09716799855232239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,4,balanced,0.0787360022465388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,4,balanced,0.09902399778366089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,4,balanced,4.504048029581706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,4,balanced,0.08382933338483174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,4,balanced,0.09573866923650105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,4,balanced,0.08931199709574382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,4,balanced,0.08828799923261006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,4,balanced,0.08942932883898418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,4,balanced,0.07445866862932841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,4,balanced,0.0949173370997111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,4,balanced,0.1088106632232666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,4,balanced,0.10904000202814738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,4,balanced,0.08488532900810242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,4,balanced,0.11775466799736023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,4,balanced,0.0990133285522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,4,balanced,0.14271466930707297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,4,balanced,0.13924800356229147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,4,balanced,0.16141333182652792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,4,balanced,0.16451733311017355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,4,balanced,0.2082186738650004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,4,balanced,0.2312693397204081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,4,balanced,0.25712533791859943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,4,balanced,0.2941280007362366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,4,balanced,0.3442399899164836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,4,balanced,0.42045867443084717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,4,balanced,0.5307039817174276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,4,balanced,0.4416266679763794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,4,balanced,0.7854186693827311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,4,balanced,0.6194239854812622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,4,balanced,1.0219733715057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,4,balanced,0.8124106725056967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,4,balanced,1.514560063680013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,4,balanced,1.0162879625956218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,4,balanced,2.0063145955403647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,4,balanced,1.5732533137003581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,4,balanced,2.4953813552856445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,4,balanced,3.944618542989095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,4,balanced,3.2008374532063804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,4,balanced,7.8439788818359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,4,power_law_1.01,0.12272640466690063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,4,power_law_1.01,0.0698303997516632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,4,power_law_1.2,0.07226240038871765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,4,power_law_1.2,0.12236160039901733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,4,power_law_1.2,0.03266560137271881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,4,power_law_1.01,0.03277440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,4,power_law_1.01,0.08282880187034607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,4,power_law_1.2,0.07793920040130616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,4,power_law_1.01,0.0977728009223938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,4,power_law_1.01,0.06725760102272034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,4,power_law_1.2,0.07012479901313781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,4,power_law_1.2,0.14640640020370482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,4,power_law_1.2,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,4,power_law_1.01,0.03175680041313171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,4,power_law_1.01,0.07310079932212829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,4,power_law_1.2,0.0837119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,4,power_law_1.01,0.1352255940437317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,4,power_law_1.01,0.06339200139045716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,4,power_law_1.2,0.06999679803848266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,4,power_law_1.2,0.13447680473327636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,4,power_law_1.2,0.031481599807739256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,4,power_law_1.01,0.031200000643730165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,4,power_law_1.01,0.08300160169601441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,4,power_law_1.2,0.03617280125617981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,4,power_law_1.2,0.07262719869613647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,4,power_law_1.01,0.13775999546051027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,4,power_law_1.01,0.07560319900512695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,4,power_law_1.2,0.07604479789733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,4,power_law_1.2,0.13781119585037233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,4,power_law_1.01,0.03616639971733093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,4,power_law_1.01,0.09146239757537841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,4,power_law_1.2,0.03733119964599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,4,power_law_1.2,0.08908799886703492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,4,power_law_1.01,0.1392832040786743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,4,power_law_1.01,0.0759872019290924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,4,power_law_1.2,0.07522559762001038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,4,power_law_1.2,0.13635200262069702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,4,power_law_1.01,0.037171199917793274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,4,power_law_1.01,0.0910975992679596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,4,power_law_1.2,0.038387200236320494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,4,power_law_1.2,0.08993279933929443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,4,power_law_1.01,0.14382079839706421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,4,power_law_1.2,0.05677440166473389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,4,power_law_1.01,0.07667199969291687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,4,power_law_1.2,0.07769600152969361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,4,power_law_1.2,0.04496639966964722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,4,power_law_1.2,0.1422592043876648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,4,power_law_1.01,0.0383296012878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,4,power_law_1.01,0.0908352017402649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,4,power_law_1.2,0.09036160111427308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,4,power_law_1.01,0.15024640560150146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,4,power_law_1.01,0.07760000228881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,4,power_law_1.2,0.07761279940605163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,4,power_law_1.2,0.09566720128059387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,4,power_law_1.2,0.15873279571533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,4,power_law_1.01,0.05728639960289002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,4,power_law_1.01,0.09372159838676453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,4,power_law_1.2,0.09239680171012879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,4,power_law_1.01,0.19043200016021727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,4,power_law_1.01,0.07871999740600585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,4,power_law_1.2,0.07893120050430298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,4,power_law_1.01,0.07983360290527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,4,power_law_1.2,0.09654399752616882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,4,power_law_1.2,0.17742079496383667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,4,power_law_1.01,0.04526079893112182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,4,power_law_1.01,0.09832320213317872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,4,power_law_1.2,0.09512320160865784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,4,power_law_1.01,0.2068095922470093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,4,power_law_1.2,0.07985919713973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,4,power_law_1.01,0.08013439774513245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,4,power_law_1.2,0.09761919975280761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,4,power_law_1.01,0.08083840012550354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,4,power_law_1.2,0.20914559364318847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,4,power_law_1.2,0.0943231999874115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,4,power_law_1.01,0.0954688012599945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,4,power_law_1.01,0.10381439924240113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,4,power_law_1.2,0.10553599596023559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,4,power_law_1.01,0.2009023904800415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,4,power_law_1.2,0.08206080198287964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,4,power_law_1.01,0.08158079981803894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,4,power_law_1.2,0.21388161182403564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,4,power_law_1.2,0.0899392008781433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,4,power_law_1.01,0.09646720290184022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,4,power_law_1.01,0.10625920295715333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,4,power_law_1.2,0.10429439544677735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,4,power_law_1.01,0.15961600542068483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,4,power_law_1.2,0.08069120049476623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,4,power_law_1.01,0.08450559973716736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,4,power_law_1.2,0.15735039710998536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,4,power_law_1.2,0.07255039811134338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,4,power_law_1.01,0.09791359901428223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,4,power_law_1.01,0.11023999452590942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,4,power_law_1.2,0.10595840215682983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,4,power_law_1.01,0.18295040130615234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,4,power_law_1.2,0.08691840171813965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,4,power_law_1.01,0.09270399808883667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,4,power_law_1.2,0.11099519729614257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,4,power_law_1.2,0.17916159629821776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,4,power_law_1.2,0.10480639934539795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,4,power_law_1.2,0.12773120403289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,4,power_law_1.01,0.09738879799842834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,4,power_law_1.01,0.11338880062103271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,4,power_law_1.01,0.1786239981651306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,4,power_law_1.2,0.0859008014202118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,4,power_law_1.01,0.10586880445480347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,4,power_law_1.2,0.17722879648208617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,4,power_law_1.01,0.10611200332641602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,4,power_law_1.2,0.08284159898757934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,4,power_law_1.2,0.11939200162887573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,4,power_law_1.01,0.08913279771804809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,4,power_law_1.01,0.12476160526275634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,4,power_law_1.01,0.1943743944168091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,4,power_law_1.01,0.12343039512634277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,4,power_law_1.2,0.0941760003566742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,4,power_law_1.2,0.1933568000793457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,4,power_law_1.01,0.13646080493927001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,4,power_law_1.2,0.09786239862442017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,4,power_law_1.2,0.22941439151763915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,4,power_law_1.2,0.14086400270462035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,4,power_law_1.01,0.07278720140457154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,4,power_law_1.01,0.2048959970474243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,4,power_law_1.01,0.14220160245895386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,4,power_law_1.2,0.10662399530410767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,4,power_law_1.01,0.14302719831466676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,4,power_law_1.2,0.1384703993797302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.01,0.18332159519195557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,4,power_law_1.2,0.24753921031951903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.01,0.24147839546203614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,4,power_law_1.2,0.1509760022163391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,4,power_law_1.01,0.1080896019935608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,4,power_law_1.01,0.23169920444488526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,4,power_law_1.2,0.16651519536972045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,4,power_law_1.01,0.1513152003288269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,4,power_law_1.2,0.10802559852600098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.2,0.16261119842529298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,4,power_law_1.2,0.23594241142272948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.01,0.3136831998825073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,4,power_law_1.2,0.33911681175231934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,4,power_law_1.01,0.08309760093688964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,4,power_law_1.01,0.21498239040374756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,4,power_law_1.01,0.09809280037879944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,4,power_law_1.2,0.20674560070037842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,4,power_law_1.01,0.17203840017318725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,4,power_law_1.2,0.2791935920715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,4,power_law_1.2,0.12961920499801635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.2,0.23045120239257813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.01,0.37304959297180174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,4,power_law_1.2,0.3918911933898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,4,power_law_1.01,0.30542080402374266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,4,power_law_1.01,0.13991039991378784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,4,power_law_1.01,0.20485761165618896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,4,power_law_1.2,0.33240959644317625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,4,power_law_1.2,0.15301120281219482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.2,0.29341440200805663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.01,0.4794303894042969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,4,power_law_1.2,0.5893184185028076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,4,power_law_1.01,0.3282687902450562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.01,0.1631872057914734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,4,power_law_1.01,0.25012478828430174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,4,power_law_1.2,0.44879999160766604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.2,0.1940287947654724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.2,0.42138237953186036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.01,0.7635519981384278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,4,power_law_1.2,0.6387775897979736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,4,power_law_1.01,0.581990385055542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.01,0.23121919631958007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,4,power_law_1.01,0.2866175889968872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,4,power_law_1.2,0.5812992095947266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.2,0.2873215913772583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.2,0.5285247802734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.01,1.1785408020019532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,4,power_law_1.2,0.8743488311767578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,4,power_law_1.01,0.5427072048187256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.01,0.2936000108718872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,4,power_law_1.01,0.36690559387207033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.01,0.423199987411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,4,power_law_1.2,0.7845503807067871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.2,0.2803328037261963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.2,0.7855807781219483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.01,1.8562240600585938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,4,power_law_1.2,1.159724807739258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,4,power_law_1.01,0.7903808116912842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,4,power_law_1.01,0.5081984043121338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,4,power_law_1.01,1.1154560089111327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.01,0.5294015884399415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,4,power_law_1.2,0.9508223533630371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.2,0.425113582611084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.2,1.0207807540893554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.01,2.092025566101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,4,power_law_1.2,1.647871971130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,4,power_law_1.01,0.6910016059875488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,4,power_law_1.01,1.2032832145690917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.01,0.784665584564209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,4,power_law_1.01,1.997599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,4,power_law_1.2,1.6501760482788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.2,0.47467517852783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.2,1.5133824348449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,4,power_law_1.2,2.263961601257324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,4,power_law_1.01,2.474527931213379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,4,power_law_1.2,2.388819122314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,4,power_law_1.01,0.9290623664855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.01,1.0196736335754395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,4,power_law_1.01,2.6293888092041016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,4,power_law_1.01,1.2813695907592773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.2,0.8979455947875976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.2,2.003571128845215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,4,power_law_1.2,3.515494537353516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,4,power_law_1.01,4.218150329589844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,4,power_law_1.2,3.563564682006836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.01,1.517024040222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,4,power_law_1.01,4.328851318359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,4,power_law_1.01,1.6213119506835938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.2,1.217036819458008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,4,power_law_1.2,2.4868288040161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,4,power_law_1.2,3.768563079833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,4,power_law_1.01,7.076364898681641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,4,power_law_1.2,6.119692611694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.01,2.0088319778442383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,4,power_law_1.01,5.45074577331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,4,power_law_1.01,2.5425472259521484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.2,1.995699119567871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,4,power_law_1.2,3.9450496673583983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,4,power_law_1.2,4.979033660888672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,4,power_law_1.2,7.335052490234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,4,power_law_1.01,2.4938304901123045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,4,power_law_1.01,9.779603576660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,4,power_law_1.01,3.4714561462402345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.2,2.076710319519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,4,power_law_1.2,7.832511901855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,4,power_law_1.01,3.9507774353027343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,4,power_law_1.2,11.187270355224609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,4,power_law_1.2,9.09850845336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,4,power_law_1.01,5.34777603149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,4,power_law_1.01,20.092640686035157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,4,power_law_1.2,3.218431854248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,4,power_law_1.01,7.839730834960937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,4,power_law_1.01,12.060870361328124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,4,power_law_1.2,21.126463317871092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,4,power_law_1.2,5.152403259277344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,4,power_law_1.2,9.899769592285157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,2,balanced,0.03843733419974645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,2,balanced,0.06393066545327504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,2,balanced,0.060165335734685264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,2,balanced,0.023242667317390442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,2,power_law_1.01,0.082777601480484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,2,power_law_1.01,0.02240000069141388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,2,power_law_1.01,0.05122560262680054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,2,power_law_1.01,0.07726719975471497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,2,balanced,0.059418668349583946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,2,balanced,0.07878399888674419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,2,balanced,0.07997333506743114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,2,balanced,0.02348800003528595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,2,power_law_1.01,0.022380800545215608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,2,power_law_1.01,0.08821759819984436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,2,balanced,0.08427733182907104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,2,power_law_1.01,0.0572160005569458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,2,power_law_1.01,0.09089279770851136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,2,balanced,0.13452800114949545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,2,balanced,0.11728533109029134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,2,balanced,0.02316266546646754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,2,power_law_1.01,0.021798400580883025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,2,power_law_1.01,0.09724799990653991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,2,balanced,0.0230880007147789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,2,balanced,0.08866133292516072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,2,power_law_1.01,0.09676799774169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,2,power_law_1.01,0.06782720088958741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,2,balanced,0.1390666663646698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,2,balanced,0.11569066842397054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,2,power_law_1.01,0.02082560062408447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,2,balanced,0.023189333577950794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,2,power_law_1.01,0.11964800357818603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,2,balanced,0.13828266660372415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,2,balanced,0.08665600419044495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,2,power_law_1.01,0.13329919576644897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,2,power_law_1.01,0.08711040019989014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,2,power_law_1.01,0.02210559993982315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,2,balanced,0.123471995194753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,2,balanced,0.027290667096773785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,2,power_law_1.01,0.11781760454177856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,2,balanced,0.13929599523544312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,2,power_law_1.01,0.12296960353851319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,2,balanced,0.08763733506202698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,2,power_law_1.01,0.13490560054779052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,2,power_law_1.01,0.02529920041561127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,2,power_law_1.01,0.08731520175933838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,2,balanced,0.11452800035476685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,2,balanced,0.03818666686614355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,2,power_law_1.01,0.11477760076522828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,2,balanced,0.14032000303268433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,2,balanced,0.09011733531951904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,2,power_law_1.01,0.13856639862060546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,2,power_law_1.01,0.036185601353645326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,2,power_law_1.01,0.08731520175933838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,2,balanced,0.029445332785447437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,2,balanced,0.12339199582735698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,2,power_law_1.01,0.11623680591583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,2,balanced,0.14041067163149515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,2,power_law_1.01,0.09054080247879029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,2,power_law_1.01,0.02975359857082367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,2,balanced,0.0899679958820343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,2,power_law_1.01,0.1443071961402893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,2,balanced,0.03882666677236557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,2,power_law_1.01,0.03758080005645752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,2,power_law_1.01,0.09343360066413879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,2,balanced,0.11586667100588481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,2,power_law_1.01,0.11807359457015991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,2,balanced,0.1402613321940104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,2,balanced,0.0897173285484314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,2,power_law_1.01,0.14493440389633178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,2,balanced,0.03324799984693527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,2,power_law_1.01,0.03299840092658997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,2,power_law_1.01,0.09453439712524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,2,balanced,0.123471995194753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,2,power_law_1.01,0.03540480136871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,2,power_law_1.01,0.15083520412445067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,2,power_law_1.01,0.11441279649734497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,2,balanced,0.09167466560999553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,2,balanced,0.13961600263913473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,2,balanced,0.037530665596326195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,2,power_law_1.01,0.09570559859275818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,2,power_law_1.01,0.04056319892406464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,2,balanced,0.1244533360004425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,2,power_law_1.01,0.15221760272979737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,2,power_law_1.01,0.04480000138282776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,2,power_law_1.01,0.11114879846572875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,2,power_law_1.01,0.15058560371398927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,2,balanced,0.09308266639709473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,2,balanced,0.1446506679058075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,2,power_law_1.01,0.09926400184631348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,2,balanced,0.0428959975639979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,2,power_law_1.01,0.0463808000087738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,2,balanced,0.10997333129247029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,2,power_law_1.01,0.1617087960243225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,2,power_law_1.01,0.1179967999458313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,2,balanced,0.09715732932090759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,2,balanced,0.14324800173441568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,2,power_law_1.01,0.1004480004310608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,2,balanced,0.04612799982229868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,2,power_law_1.01,0.06458240151405334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,2,balanced,0.1169599990049998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,2,power_law_1.01,0.16155519485473632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,2,balanced,0.09530666470527649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,2,balanced,0.14410133163134256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,2,power_law_1.01,0.10774400234222412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,2,power_law_1.01,0.12655359506607056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,2,balanced,0.04823466638724009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,2,power_law_1.01,0.05534719824790955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,2,balanced,0.11002133289972942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,2,balanced,0.1025279959042867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,2,power_law_1.01,0.17256319522857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,2,balanced,0.154448002576828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,2,power_law_1.01,0.11190400123596192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,2,power_law_1.01,0.062009602785110474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,2,power_law_1.01,0.1309056043624878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,2,balanced,0.06983466446399689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,2,balanced,0.10596266388893127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,2,balanced,0.11095999677975972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,2,power_law_1.01,0.12257280349731445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,2,balanced,0.16354667147000632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,2,power_law_1.01,0.19143680334091187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,2,power_law_1.01,0.0827455997467041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,2,power_law_1.01,0.1398848056793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,2,balanced,0.05726400017738342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,2,balanced,0.09916266798973083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,2,balanced,0.11111467083295186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,2,power_law_1.01,0.13870079517364503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,2,balanced,0.16614933808644614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,2,power_law_1.01,0.19971840381622313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.01,0.09884799718856811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,2,power_law_1.01,0.14145920276641846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,2,balanced,0.06424533327420552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.01,0.1374400019645691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,2,balanced,0.17657599846522012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,2,balanced,0.11517332990964253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,2,balanced,0.10264533758163452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,2,power_law_1.01,0.12767360210418702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,2,power_law_1.01,0.2209023952484131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.01,0.17108479738235474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,2,power_law_1.01,0.15215359926223754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,2,balanced,0.0849120020866394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,2,balanced,0.1954560081164042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,2,balanced,0.13795199990272522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,2,balanced,0.0995199978351593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,2,power_law_1.01,0.17103359699249268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,2,power_law_1.01,0.2578752040863037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.01,0.23922560214996338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,2,power_law_1.01,0.1754240036010742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,2,balanced,0.14473600188891092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,2,balanced,0.100490669409434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,2,balanced,0.2128480076789856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,2,balanced,0.16639999548594156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,2,power_law_1.01,0.19032319784164428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.01,0.3009216070175171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,2,balanced,0.14088533322016397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,2,power_law_1.01,0.2626240015029907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.01,0.2220479965209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,2,balanced,0.1794933279355367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,2,balanced,0.14061333735783896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,2,balanced,0.2567039926846822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.01,0.43553919792175294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,2,power_law_1.01,0.2396480083465576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,2,balanced,0.30426132678985596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,2,balanced,0.1766293247540792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,2,power_law_1.01,0.36912639141082765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.01,0.5635647773742676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.01,0.25229439735412595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,2,balanced,0.21215466658274332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,2,balanced,0.17364799976348877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,2,power_law_1.01,0.2759552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,2,balanced,0.4028640190760295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,2,balanced,0.21667200326919556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.01,0.8295807838439941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,2,power_law_1.01,0.4367231845855713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.01,0.34561920166015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.01,1.0920063972473144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,2,balanced,0.2818506757418315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,2,balanced,0.24226667483647665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,2,power_law_1.01,0.39295361042022703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.01,0.4965951919555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,2,power_law_1.01,0.5244160175323487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,2,power_law_1.01,1.3541760444641113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,2,balanced,0.2940000096956889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,2,balanced,0.49094398816426593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,2,balanced,0.3498400052388509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.01,0.4978176116943359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,2,power_law_1.01,0.4629312038421631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,2,balanced,0.3036959966023763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,2,power_law_1.01,0.6708928108215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,2,power_law_1.01,2.1451391220092773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,2,balanced,0.37417598565419513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,2,balanced,0.6960426966349283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,2,power_law_1.01,0.7251071929931641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.01,0.8424192428588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,2,balanced,0.49038398265838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,2,balanced,0.43875734011332196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,2,power_law_1.01,1.141484832763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,2,power_law_1.01,4.254118347167969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,2,balanced,0.525434652964274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,2,balanced,0.8825706640879313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,2,power_law_1.01,0.9244223594665527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.01,1.0212096214294433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,2,balanced,0.61626664797465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,2,power_law_1.01,1.4702207565307617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,2,balanced,0.5677226781845093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,2,power_law_1.01,1.0818047523498535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,2,balanced,0.6850773493448893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.01,1.6076160430908204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,2,balanced,1.3056906859079997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,2,power_law_1.01,1.7796991348266602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,2,balanced,0.8996853033701578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,2,balanced,0.8325013319651285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,2,power_law_1.01,1.8248064041137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,2,balanced,0.999013344446818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.01,1.8693952560424805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,2,balanced,1.1632533073425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,2,power_law_1.01,2.5255807876586913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,2,balanced,1.673749287923177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,2,power_law_1.01,2.1273088455200195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,2,power_law_1.01,2.3222591400146486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,2,balanced,1.0962560176849365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,2,power_law_1.01,3.2027713775634767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,2,balanced,1.305792013804118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,2,balanced,1.4484586715698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,2,balanced,2.0777014096577964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,2,power_law_1.01,4.0422721862792965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,2,power_law_1.01,3.6774784088134767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,2,power_law_1.01,4.909555053710937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,2,balanced,1.3590614000956218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,2,power_law_1.01,9.291372680664063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,2,power_law_1.01,7.172787475585937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,2,balanced,1.6172800064086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,2,power_law_1.01,10.289695739746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,2,balanced,3.2420374552408853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,2,balanced,2.2533226013183594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,2,balanced,2.14793062210083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,2,balanced,2.572373390197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,2,balanced,6.5088850657145185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,2,balanced,4.415514628092448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,2,balanced,4.262314796447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,2,balanced,5.178192138671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,1,balanced,0.05819199979305267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,1,balanced,0.021173333128293354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,1,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,1,balanced,0.057946667075157166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,2,power_law_1.2,0.022265599668025972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,2,power_law_1.2,0.07255039811134338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,2,power_law_1.2,0.08731520175933838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,1,balanced,0.08041599889596303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,2,power_law_1.2,0.05603839755058289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,1,balanced,0.07834133505821228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,1,balanced,0.021189334491888683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,1,balanced,0.05690133571624756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,2,power_law_1.2,0.022233599424362184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,2,power_law_1.2,0.07828480005264282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,2,power_law_1.2,0.08364160060882568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,1,balanced,0.021194666624069214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,1,balanced,0.1346773306528727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,1,balanced,0.1172320048014323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,1,balanced,0.08247466882069905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,2,power_law_1.2,0.060102397203445436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,2,power_law_1.2,0.09428480267524719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,1,balanced,0.019674666225910187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,2,power_law_1.2,0.021753600239753722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,2,power_law_1.2,0.09660159945487976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,1,balanced,0.020869334538777668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,1,balanced,0.11213333408037822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,1,balanced,0.13498133420944214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,1,balanced,0.08566400408744812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,2,power_law_1.2,0.06605439782142639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,2,power_law_1.2,0.13735040426254272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,2,power_law_1.2,0.020921599864959717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,1,balanced,0.022042666872342426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,2,power_law_1.2,0.08459519743919372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,2,power_law_1.2,0.12145919799804687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,1,balanced,0.11539733409881592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,1,balanced,0.1372106671333313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,2,power_law_1.2,0.022099199891090392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,1,balanced,0.08481599887212117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,2,power_law_1.2,0.13485440015792846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,1,balanced,0.02657066782315572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,2,power_law_1.2,0.08480640053749085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,2,power_law_1.2,0.12403199672698975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,1,balanced,0.10612266262372334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,2,power_law_1.2,0.025273600220680238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,1,balanced,0.1283573309580485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,1,balanced,0.1381600002447764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,1,balanced,0.025248001019159954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,1,balanced,0.08687466382980347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,2,power_law_1.2,0.13925119638442993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,2,power_law_1.2,0.11929600238800049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,2,power_law_1.2,0.0876416027545929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,1,balanced,0.10307733217875163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,2,power_law_1.2,0.037836799025535585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,1,balanced,0.029253333806991577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,1,balanced,0.13846932848294577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,1,balanced,0.08892800410588582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,2,power_law_1.2,0.1439296007156372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,2,power_law_1.2,0.11946239471435546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,2,power_law_1.2,0.09048960208892823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,1,balanced,0.10735467076301575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,2,power_law_1.2,0.029542401432991028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,1,balanced,0.027306665976842243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,1,balanced,0.13970667123794556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,1,balanced,0.08750933408737183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,2,power_law_1.2,0.14312959909439088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,2,power_law_1.2,0.11345280408859253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,2,power_law_1.2,0.09287679791450501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,1,balanced,0.11692266662915547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,1,balanced,0.029658667743206024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,2,power_law_1.2,0.037920001149177554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,1,balanced,0.13940266768137613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,1,balanced,0.08899199962615967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,2,power_law_1.2,0.15278719663619994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,2,power_law_1.2,0.11409280300140381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,2,power_law_1.2,0.09371520280838012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,2,power_law_1.2,0.14859520196914672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,2,power_law_1.2,0.09581440091133117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,2,power_law_1.2,0.032646399736404416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,1,balanced,0.03146133323510488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,2,power_law_1.2,0.11126400232315063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,1,balanced,0.08918933073679607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,1,balanced,0.10711466272672017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,1,balanced,0.13901866475741068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,2,power_law_1.2,0.15745279788970948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,2,power_law_1.2,0.09813119769096375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,2,power_law_1.2,0.03561600148677826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,1,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,2,power_law_1.2,0.15998079776763915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,2,power_law_1.2,0.11390080451965331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,1,balanced,0.09131200114885966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,2,power_law_1.2,0.040633600950241086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,2,power_law_1.2,0.10293760299682617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,1,balanced,0.10523733496665955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,1,balanced,0.14179733395576477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,1,balanced,0.03602666656176249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,2,power_law_1.2,0.1641088008880615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,2,power_law_1.2,0.11624959707260132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,1,balanced,0.09771200021107991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,2,power_law_1.2,0.04478079974651337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,2,power_law_1.2,0.10867840051651001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,1,balanced,0.10506133238474528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,1,balanced,0.14177067081133524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,2,power_law_1.2,0.17598079442977904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,1,balanced,0.045941332976023354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,1,balanced,0.09622933467229207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,2,power_law_1.2,0.11975040435791015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,2,power_law_1.2,0.046265599131584165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,2,power_law_1.2,0.11177599430084229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,1,balanced,0.10746133327484131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,2,power_law_1.2,0.1913472056388855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,1,balanced,0.1446613371372223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,2,power_law_1.2,0.0627839982509613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,1,balanced,0.04394133388996124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,2,power_law_1.2,0.125164794921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,1,balanced,0.10039466619491577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,2,power_law_1.2,0.12486399412155151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,1,balanced,0.1095413366953532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,2,power_law_1.2,0.19765119552612304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,2,power_law_1.2,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,1,balanced,0.15043733517328897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,1,balanced,0.04561600089073181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,2,power_law_1.2,0.1418239951133728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,1,balanced,0.11245866616566975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,2,power_law_1.2,0.13703680038452148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,1,balanced,0.10333866874376933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,2,power_law_1.2,0.22956159114837646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,2,power_law_1.2,0.061791998147964475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,1,balanced,0.17273600896199545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,1,balanced,0.058330665032068886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,2,power_law_1.2,0.14073599576950074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,1,balanced,0.11348266402880351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,2,power_law_1.2,0.13733760118484498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,2,power_law_1.2,0.08299520015716552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,1,balanced,0.10499200224876404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,2,power_law_1.2,0.2618304014205933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,1,balanced,0.17294400930404663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,1,balanced,0.06778133412202199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.2,0.0989184021949768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,2,power_law_1.2,0.15750399827957154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,1,balanced,0.08760533730189006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,1,balanced,0.11776000261306763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,2,power_law_1.2,0.16814719438552855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,1,balanced,0.16060800353686014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,2,power_law_1.2,0.2847935914993286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,1,balanced,0.17286932468414307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.2,0.1374783992767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,2,power_law_1.2,0.18125439882278443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,1,balanced,0.10779733459154765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,1,balanced,0.14035733540852866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,2,power_law_1.2,0.19309439659118652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,2,power_law_1.2,0.36078720092773436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,1,balanced,0.1511360009511312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.2,0.17081600427627563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,1,balanced,0.20453333854675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.2,0.20722560882568358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,1,balanced,0.14702399571736655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,1,balanced,0.14818666378657022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,2,power_law_1.2,0.2649087905883789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,2,power_law_1.2,0.4606336116790771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,1,balanced,0.19370667139689127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.2,0.24011518955230712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.2,0.29614720344543455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,1,balanced,0.2148639957110087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,1,balanced,0.18413333098093668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,1,balanced,0.18261865774790445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.2,0.3007040023803711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,2,power_law_1.2,0.2872960090637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,2,power_law_1.2,0.6631040096282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,1,balanced,0.24016000827153525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.2,0.33953280448913575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,1,balanced,0.2799253265062968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,1,balanced,0.21335466702779135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,1,balanced,0.2559573252995809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,2,power_law_1.2,0.7586112022399902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.2,0.4362175941467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,2,power_law_1.2,0.3887295961380005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.2,0.45661439895629885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,1,balanced,0.3224746584892273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,1,balanced,0.3021120031674703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,1,balanced,0.28917866945266724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.2,0.5759103775024415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,2,power_law_1.2,1.0059519767761231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.2,0.5647808074951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,2,power_law_1.2,0.47733120918273925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,1,balanced,0.32864532868067425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,1,balanced,0.40430935223897296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,1,balanced,0.35452266534169513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,1,balanced,0.4118080139160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.2,0.9377599716186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,2,power_law_1.2,1.3024319648742675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.2,0.830508804321289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,2,power_law_1.2,0.8016832351684571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,1,balanced,0.4920213222503662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,1,balanced,0.47552533944447833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,1,balanced,0.4987573226292928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,1,balanced,0.5829066832860311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.2,1.0804672241210938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,2,power_law_1.2,1.0412416458129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.2,1.096243190765381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,2,power_law_1.2,2.408793640136719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,1,balanced,0.7009920279184977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,2,power_law_1.2,1.2232640266418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.2,1.7089792251586915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,1,balanced,0.624127984046936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,1,balanced,0.646506667137146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,1,balanced,0.7655680179595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,2,power_law_1.2,1.3592767715454102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,2,power_law_1.2,2.9436031341552735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,2,power_law_1.2,1.9482879638671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.2,2.104863929748535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,1,balanced,0.894261360168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,1,balanced,0.9297706286112467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,2,power_law_1.2,2.150111961364746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,1,balanced,1.1129226684570312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,1,balanced,0.7699786822001139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,2,power_law_1.2,3.132953643798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,2,power_law_1.2,2.4194368362426757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,2,power_law_1.2,2.602873611450195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,2,power_law_1.2,4.259065628051758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,1,balanced,1.309930642445882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,2,power_law_1.2,5.863206481933593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,2,power_law_1.2,3.754169464111328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,1,balanced,1.2125173409779866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,2,power_law_1.2,4.5607551574707035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,1,balanced,1.4749174118041992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,1,balanced,1.2118773460388184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,2,power_law_1.2,11.364595031738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,1,balanced,1.708554744720459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,2,power_law_1.2,8.843539428710937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,2,power_law_1.2,6.714272308349609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,1,balanced,1.488965352376302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,1,balanced,1.828378677368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,1,balanced,2.1157280604044595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,1,balanced,2.3340373039245605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,1,balanced,2.394399960835775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,1,balanced,2.8914079666137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,1,balanced,3.3050880432128906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,1,balanced,4.5794986089070635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,1,balanced,5.690783818562825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,1,balanced,6.693386713663737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,1,power_law_1.01,0.056492799520492555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,1,power_law_1.01,0.06883199810981751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,1,power_law_1.01,0.071424001455307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,1,power_law_1.01,0.1309056043624878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,1,power_law_1.01,0.13162879943847655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,1,power_law_1.01,0.13814400434494017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,1,power_law_1.01,0.14149119853973388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,1,power_law_1.01,0.14156800508499146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,1,power_law_1.01,0.14179199934005737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,1,power_law_1.01,0.14235520362854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,1,power_law_1.01,0.13972480297088624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,1,power_law_1.01,0.14962559938430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,1,power_law_1.01,0.15094399452209473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,1,power_law_1.01,0.156876802444458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,1,power_law_1.01,0.17278079986572265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,1,power_law_1.01,0.1844032049179077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,1,power_law_1.01,0.17946879863739013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,1,power_law_1.01,0.03852800130844116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,1,power_law_1.01,0.019577600061893463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,1,power_law_1.01,0.06294400095939637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,1,power_law_1.01,0.2106112003326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,1,power_law_1.01,0.24021759033203124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,1,power_law_1.01,0.0487744003534317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,1,power_law_1.01,0.062054401636123656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,1,power_law_1.01,0.2922368049621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,1,power_law_1.01,0.052665597200393675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,1,power_law_1.2,0.056467199325561525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,1,power_law_1.01,0.08119680285453797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,1,power_law_1.01,0.019763199985027312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,1,power_law_1.01,0.08071680068969726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,1,power_law_1.01,0.337990403175354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,1,power_law_1.01,0.11622400283813476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,1,power_law_1.2,0.06655359864234925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,1,power_law_1.01,0.018918399512767792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,1,power_law_1.01,0.08261119723320007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,1,power_law_1.01,0.10742399692535401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,1,power_law_1.01,0.4465151786804199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,1,power_law_1.2,0.08705919981002808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,1,power_law_1.01,0.020006400346755982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,1,power_law_1.01,0.08325120210647582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,1,power_law_1.01,0.5425983905792237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,1,power_law_1.2,0.1336575984954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,1,power_law_1.2,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,1,power_law_1.01,0.11311359405517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,1,power_law_1.01,0.08693119883537292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,1,power_law_1.2,0.130348801612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,1,power_law_1.01,0.7593344211578369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,1,power_law_1.01,0.11151360273361206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,1,power_law_1.01,0.020134399831295013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,1,power_law_1.01,0.08788480162620545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,1,power_law_1.2,0.1368064045906067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,1,power_law_1.2,0.05003520250320435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,1,power_law_1.01,0.9678591728210449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,1,power_law_1.01,0.11113599538803101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,1,power_law_1.01,0.08967040181159973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,1,power_law_1.2,0.1371392011642456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,1,power_law_1.01,0.021561600267887115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,1,power_law_1.01,1.350387191772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,1,power_law_1.2,0.050937598943710326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,1,power_law_1.01,0.11030399799346924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,1,power_law_1.01,0.08978559970855712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,1,power_law_1.2,0.13972480297088624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,1,power_law_1.01,1.7672063827514648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,1,power_law_1.01,0.02550399899482727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,1,power_law_1.2,0.14306559562683105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,1,power_law_1.01,0.08967040181159973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,1,power_law_1.01,0.024460799992084503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,1,power_law_1.01,2.166361618041992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,1,power_law_1.2,0.0836736023426056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,1,power_law_1.01,0.11244800090789794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,1,power_law_1.2,0.142848002910614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,1,power_law_1.01,0.0959231972694397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,1,power_law_1.01,0.028288000822067262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,1,power_law_1.2,0.14317439794540404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,1,power_law_1.2,0.08302080035209655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,1,power_law_1.01,0.11022080183029175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,1,power_law_1.01,3.3829952239990235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,1,power_law_1.01,0.09876480102539062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,1,power_law_1.2,0.15160319805145264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,1,power_law_1.01,0.02675839960575104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,1,power_law_1.2,0.08508800268173218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,1,power_law_1.01,0.11649279594421387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,1,power_law_1.01,0.10131200551986694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,1,power_law_1.2,0.1501312017440796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,1,power_law_1.01,0.027769601345062254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,1,power_law_1.2,0.08673279881477355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,1,power_law_1.01,6.781868743896484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,1,power_law_1.01,0.11514879465103149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,1,power_law_1.2,0.019097599387168884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,1,power_law_1.2,0.06275839805603027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,1,power_law_1.2,0.15595519542694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,1,power_law_1.01,0.11091840267181396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,1,power_law_1.01,0.030444800853729248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,1,power_law_1.2,0.088128000497818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,1,power_law_1.01,0.12058240175247192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,1,power_law_1.2,0.17457280158996583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,1,power_law_1.01,0.11813759803771973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,1,power_law_1.2,0.019513599574565887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,1,power_law_1.2,0.06480000019073487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,1,power_law_1.01,0.12885119915008544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,1,power_law_1.01,0.03299840092658997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,1,power_law_1.2,0.09012479782104492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,1,power_law_1.2,0.07652480006217957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,1,power_law_1.2,0.18602880239486694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,1,power_law_1.01,0.1229248046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,1,power_law_1.2,0.019270400702953338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,1,power_law_1.2,0.09019520282745361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,1,power_law_1.01,0.13320959806442262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,1,power_law_1.01,0.034764799475669864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,1,power_law_1.2,0.12197120189666748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,1,power_law_1.2,0.18339840173721314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,1,power_law_1.01,0.14138879776000976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,1,power_law_1.2,0.09210240244865417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,1,power_law_1.2,0.11491199731826782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,1,power_law_1.2,0.21027839183807373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,1,power_law_1.2,0.01971839964389801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,1,power_law_1.01,0.13903360366821288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,1,power_law_1.01,0.04568960070610047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,1,power_law_1.01,0.15923839807510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,1,power_law_1.2,0.24344959259033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,1,power_law_1.2,0.10974080562591552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,1,power_law_1.2,0.09607679843902588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,1,power_law_1.2,0.01987839937210083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,1,power_law_1.01,0.16727039813995362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,1,power_law_1.01,0.04268800020217896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,1,power_law_1.2,0.29704320430755615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,1,power_law_1.2,0.11566720008850098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,1,power_law_1.01,0.19903359413146973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,1,power_law_1.2,0.09801599979400635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,1,power_law_1.2,0.02131199985742569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.01,0.18655359745025635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,1,power_law_1.01,0.044870400428771974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,1,power_law_1.2,0.34318718910217283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,1,power_law_1.01,0.23819520473480224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,1,power_law_1.2,0.11349120140075683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,1,power_law_1.2,0.026924800872802735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,1,power_law_1.2,0.10114560127258301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.01,0.2267904043197632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,1,power_law_1.01,0.05768960118293762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,1,power_law_1.2,0.44525442123413084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,1,power_law_1.01,0.31492478847503663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,1,power_law_1.01,0.3807935953140259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.01,0.06636160016059875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.01,0.27508480548858644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,1,power_law_1.2,0.5406079769134522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,1,power_law_1.2,0.10841599702835084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,1,power_law_1.2,0.02465279996395111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,1,power_law_1.2,0.11184639930725097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,1,power_law_1.01,0.5265471935272217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.01,0.08524799942970276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,1,power_law_1.2,0.7532288074493408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.01,0.3694400072097778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,1,power_law_1.2,0.1119488000869751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,1,power_law_1.01,0.6654208183288575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,1,power_law_1.2,0.11912959814071655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,1,power_law_1.2,0.028512001037597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.01,0.10739200115203858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,1,power_law_1.2,0.9500224113464355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.01,0.4517888069152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,1,power_law_1.2,0.10959999561309815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,1,power_law_1.2,0.12387200593948364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,1,power_law_1.01,0.9455360412597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,1,power_law_1.2,0.02709760069847107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,1,power_law_1.2,1.3560064315795899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.01,0.6250751972198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.01,0.14680960178375244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,1,power_law_1.2,1.7557760238647462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,1,power_law_1.2,0.14169600009918212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.01,0.7990655899047852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,1,power_law_1.01,1.226137638092041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.01,0.17969919443130494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,1,power_law_1.2,0.10954240560531617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,1,power_law_1.2,0.028019198775291444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,1,power_law_1.2,0.11745280027389526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,1,power_law_1.2,2.160268783569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,1,power_law_1.2,0.15804799795150756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.01,1.1434240341186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,1,power_law_1.01,1.5092351913452149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.01,0.25550079345703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,1,power_law_1.2,0.12182400226593018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,1,power_law_1.2,0.03030399978160858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,1,power_law_1.2,0.19892480373382568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,1,power_law_1.2,3.3735294342041016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.01,1.4922687530517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,1,power_law_1.2,0.13479039669036866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.01,0.3267584085464478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,1,power_law_1.01,2.34466552734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,1,power_law_1.2,0.0331712007522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,1,power_law_1.2,0.23284480571746827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,1,power_law_1.2,0.139628803730011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.01,0.47526397705078127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.01,1.8639104843139649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,1,power_law_1.2,6.773760223388672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,1,power_law_1.01,4.583270263671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,1,power_law_1.2,0.0344895988702774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,1,power_law_1.2,0.306931209564209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,1,power_law_1.2,0.14334720373153687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.01,0.6216639995574951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,1,power_law_1.01,2.913510322570801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,1,power_law_1.2,0.04426240026950836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,1,power_law_1.2,0.38172159194946287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,1,power_law_1.2,0.16331520080566406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.01,0.7695104122161865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,1,power_law_1.2,0.0429504007101059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,1,power_law_1.01,5.787801742553711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,1,power_law_1.2,0.526585578918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,1,power_law_1.2,0.045407998561859134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.2,0.1863168001174927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,1,power_law_1.01,1.210374355316162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,1,power_law_1.2,0.6641151905059814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,1,power_law_1.2,0.058284801244735715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.2,0.22892160415649415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.2,0.2759936094284058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.2,0.06629120111465454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,1,power_law_1.2,0.9476544380187988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.2,0.3595776081085205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,1,power_law_1.01,2.3945344924926757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.2,0.08589439988136291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.2,0.44452481269836425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.2,0.10845439434051514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.2,0.6299263954162597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,1,power_law_1.2,1.2316479682922363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.2,0.14661760330200196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.2,0.7952191829681396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.2,0.18021759986877442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.2,1.1544384002685546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,1,power_law_1.2,1.5145792007446288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.2,0.25502719879150393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.2,1.518496036529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.2,0.32542080879211427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,1,power_law_1.2,2.34322566986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.2,1.8482879638671874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.2,0.47537918090820314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,1,power_law_1.2,2.929209518432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.2,0.6219071865081787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,1,power_law_1.2,4.581504058837891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,1,power_law_1.2,5.842937469482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.2,0.7689663887023925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,1,power_law_1.2,1.2123071670532226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,1,power_law_1.2,2.3937471389770506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,4,balanced,0.12693867087364197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,4,balanced,0.12146133184432983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,4,balanced,0.21370667219161987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,4,balanced,0.212719996770223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,4,balanced,0.2136639952659607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,4,balanced,0.21419199307759604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,4,balanced,0.21498666206995645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,4,balanced,0.2148373325665792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,4,balanced,0.21681066354115805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,4,balanced,0.21594667434692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,4,balanced,0.2274293303489685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,4,balanced,0.22324266036351523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,4,balanced,0.22459199031194052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,4,balanced,0.23751999934514365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,4,balanced,0.2550026575724284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,4,balanced,0.25702399015426636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,4,balanced,0.24740266799926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,4,balanced,0.32123200098673504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,4,balanced,0.29844266176223755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,4,balanced,0.42630934715270996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,4,balanced,0.4294293324152629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,4,balanced,0.5863146781921387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,4,balanced,0.7304693063100179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,4,balanced,1.0255786577860515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,4,balanced,1.4570026397705078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,4,balanced,2.0855305989583335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,4,balanced,2.750613212585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,4,balanced,3.42849063873291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,4,balanced,0.07613866527875264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,4,balanced,0.08158933122952779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,4,balanced,5.477338790893555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,4,balanced,0.07594666878382365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,4,balanced,0.09607999523480733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,4,balanced,0.08246933420499165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,4,balanced,0.10223467151323955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,4,balanced,0.1284266710281372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,4,balanced,0.09986666838328044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,4,balanced,0.12896000345547995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,4,balanced,0.10178666313489278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,4,balanced,0.1029919981956482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,4,balanced,0.1295840044816335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,4,balanced,0.1030453344186147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,4,balanced,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,4,balanced,0.13038399815559387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,4,balanced,0.10290132959683736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,4,balanced,0.04832000037034353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,4,balanced,0.10359467069307964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,4,balanced,0.13127999504407248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,4,balanced,0.10545066992441814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,4,balanced,0.04784533381462097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,4,balanced,0.10497066378593445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,4,balanced,0.13200533390045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,4,balanced,0.10938133796056111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,4,balanced,10.805039723714193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,4,balanced,0.04810666541258494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,4,balanced,0.13915200034777322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,4,balanced,0.12148800492286682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,4,power_law_1.01,0.16042879819869996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,4,balanced,0.14017599821090698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,4,balanced,0.059205333391825356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,4,balanced,0.12181867162386577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,4,balanced,0.06001600126425425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,4,balanced,0.13843199610710144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,4,power_law_1.01,0.15665919780731202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,4,balanced,0.122079998254776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,4,balanced,0.1381173332532247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,4,balanced,0.06198933223883311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,4,balanced,0.14546133081118265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,4,power_law_1.01,0.19633280038833617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,4,balanced,0.07268799841403961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,4,balanced,0.13808533549308777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,4,balanced,0.1520799994468689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,4,power_law_1.01,0.21245439052581788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,4,balanced,0.14218133687973022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,4,balanced,0.06811733543872833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,4,balanced,0.181658665339152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,4,balanced,0.16457600394884744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,4,power_law_1.01,0.2122112035751343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,4,balanced,0.13412266969680786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,4,balanced,0.16571199893951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,4,balanced,0.21821866432825723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,4,balanced,0.13571733236312866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,4,power_law_1.01,0.22085120677947997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,4,balanced,0.15887999534606934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,4,balanced,0.13979732990264893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,4,power_law_1.01,0.22672638893127442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,4,balanced,0.281551996866862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,4,balanced,0.2066239913304647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,4,balanced,0.13327999909718832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,4,power_law_1.01,0.2195904016494751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,4,balanced,0.2012959917386373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,4,balanced,0.35884801546732586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,4,balanced,0.12788800398508707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,4,power_law_1.01,0.24472320079803467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,4,balanced,0.10198932886123657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,4,power_law_1.01,0.24356479644775392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,4,balanced,0.2794293363889058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,4,balanced,0.14603199561436972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,4,balanced,0.480405330657959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,4,power_law_1.01,0.23749759197235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,4,balanced,0.2937013308207194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,4,balanced,0.1264639993508657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,4,power_law_1.01,0.10168319940567017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,4,power_law_1.01,0.2549760103225708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,4,balanced,0.1623840034008026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,4,balanced,0.39574400583902997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,4,power_law_1.01,0.2688767910003662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,4,balanced,0.6202880144119263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,4,power_law_1.01,0.09859200119972229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,4,balanced,0.2123039960861206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,4,power_law_1.01,0.28575360774993896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,4,balanced,0.4962773323059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,4,power_law_1.01,0.09184640049934387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,4,power_law_1.01,0.32188799381256106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,4,balanced,0.2746826608975728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,4,power_law_1.01,0.04759039878845215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,4,power_law_1.01,0.10883840322494506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,4,power_law_1.01,0.30814080238342284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,4,balanced,0.39131732781728107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,4,balanced,0.8815840085347494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,4,power_law_1.01,0.1269695997238159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,4,power_law_1.01,0.09718400239944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,4,balanced,0.6997600396474203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,4,power_law_1.01,0.04767360091209412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,4,power_law_1.01,0.35383679866790774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,4,power_law_1.01,0.12814719676971437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,4,power_law_1.01,0.04650239944458008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,4,balanced,0.5217653512954712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,4,power_law_1.01,0.475878381729126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,4,power_law_1.01,0.07997440099716187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,4,power_law_1.01,0.13264000415802002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,4,power_law_1.01,0.058182400465011594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,4,power_law_1.01,0.5355199813842774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,4,balanced,0.9481546878814697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,4,power_law_1.01,0.05868160128593445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,4,power_law_1.01,0.14104959964752198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,4,power_law_1.01,0.1001855969429016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,4,balanced,0.7448266347249349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,4,power_law_1.01,0.05991680026054382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,4,power_law_1.01,0.7090752124786377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,4,balanced,1.1692746480305989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,4,power_law_1.01,0.14239360094070436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,4,power_law_1.01,0.09982720017433167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,4,power_law_1.01,0.074099200963974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,4,power_law_1.01,1.1052032470703126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,4,power_law_1.01,0.13983360528945923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,4,power_law_1.01,0.06616320013999939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,4,power_law_1.01,0.10097919702529908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,4,balanced,0.9918239911397299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,4,power_law_1.01,1.281107234954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,4,balanced,1.353925387064616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,4,power_law_1.01,0.1328384041786194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,4,power_law_1.01,0.15807360410690308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,4,power_law_1.01,0.1338688015937805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,4,power_law_1.01,0.10160640478134156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,4,power_law_1.01,0.1410367965698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,4,power_law_1.01,1.7777536392211915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,4,power_law_1.01,0.13699840307235717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,4,power_law_1.01,0.14924800395965576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,4,power_law_1.01,0.10250240564346313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,4,power_law_1.01,0.1397760033607483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,4,power_law_1.01,2.866841506958008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,4,balanced,1.4576586087544758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,4,balanced,1.4502612749735515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,4,power_law_1.01,0.15614080429077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,4,power_law_1.01,0.13334399461746216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,4,power_law_1.01,0.1028223991394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,4,power_law_1.01,4.282700729370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,4,power_law_1.01,0.17193599939346313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,4,power_law_1.01,0.10057599544525146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,4,power_law_1.01,0.10395519733428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,4,balanced,1.7921013832092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,4,power_law_1.01,0.20818560123443602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,4,power_law_1.01,0.1463104009628296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,4,power_law_1.01,0.10896639823913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,4,balanced,1.9184373219807942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,4,power_law_1.01,0.2353663921356201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,4,power_law_1.01,0.12524800300598143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,4,power_law_1.01,7.438534545898437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,4,power_law_1.01,0.11569279432296753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,4,power_law_1.01,0.16216319799423218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,4,power_law_1.01,0.23459839820861816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,4,power_law_1.01,0.11636480093002319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,4,power_law_1.01,0.21175680160522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,4,power_law_1.01,7.46890869140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,4,power_law_1.01,0.2684416055679321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,4,balanced,2.3541919390360513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,4,balanced,2.8738346099853516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,4,balanced,2.1996000607808432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,4,power_law_1.01,0.12145919799804687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.01,0.2731136083602905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,4,power_law_1.01,0.34387199878692626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,4,power_law_1.01,8.896729278564454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,4,power_law_1.01,0.1407807946205139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.01,0.3889983892440796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,4,power_law_1.01,0.5108160018920899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.01,0.5181568145751954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,4,balanced,3.8022133509318032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,4,power_law_1.01,0.542080020904541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,4,power_law_1.01,0.13916800022125245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.01,0.7430912017822265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,4,power_law_1.01,0.6547135829925537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.01,0.9876416206359864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,4,power_law_1.01,0.15367679595947265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,4,power_law_1.01,13.8163330078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,4,balanced,3.473423957824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,4,power_law_1.01,1.2940863609313964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.01,1.455078411102295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,4,power_law_1.01,0.1900287985801697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.01,1.919968032836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.01,0.2211519956588745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,4,power_law_1.01,1.7307584762573243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.01,0.3083264112472534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.01,2.8725631713867186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,4,power_law_1.01,2.2801792144775392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,4,balanced,4.731184005737305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.01,0.3821824073791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.01,3.8008190155029298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,4,power_law_1.01,3.86798095703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,4,balanced,4.753445307413737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.01,0.560422420501709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,4,power_law_1.01,30.670208740234376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,4,power_law_1.01,4.729087829589844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.01,0.7500224113464355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,4,power_law_1.01,4.337644958496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.01,1.0812352180480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,4,power_law_1.01,7.561990356445312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.01,1.570905590057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,4,power_law_1.01,5.326950454711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.01,2.4308223724365234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,4,power_law_1.01,15.205191040039063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,4,power_law_1.01,7.844691467285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.01,2.96362247467041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,4,balanced,7.56387201944987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,4,power_law_1.01,2.9426559448242187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,4,power_law_1.01,16.98869171142578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,4,balanced,6.931802749633789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,4,power_law_1.01,6.27050895690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,4,power_law_1.01,12.927897644042968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,4,balanced,15.09655507405599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,4,power_law_1.2,0.17807359695434571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,4,power_law_1.2,0.1968384027481079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,4,power_law_1.2,0.21288959980010985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,4,power_law_1.2,0.21288321018218995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,4,power_law_1.2,0.2120959997177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,4,power_law_1.2,0.21749119758605956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,4,power_law_1.2,0.2268928050994873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,4,power_law_1.2,0.21856000423431396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,4,power_law_1.2,0.24101760387420654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,4,power_law_1.2,0.23819520473480224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,4,power_law_1.2,0.2312256097793579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,4,power_law_1.2,0.24116480350494385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,4,power_law_1.2,0.2634880065917969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,4,power_law_1.2,0.2861056089401245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,4,power_law_1.2,0.3211967945098877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,4,power_law_1.2,0.33486080169677734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,4,power_law_1.2,0.3972991943359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,4,power_law_1.2,0.43323521614074706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,4,power_law_1.2,0.5685632228851318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,4,power_law_1.2,0.7432447910308838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,4,power_law_1.2,0.8733887672424316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,4,power_law_1.2,1.4771583557128907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,4,power_law_1.2,1.8875135421752929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,4,power_law_1.2,2.772755241394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,4,power_law_1.2,4.008307266235351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,4,power_law_1.2,5.598483276367188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,4,power_law_1.2,7.320665740966797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,4,power_law_1.2,0.09133440256118774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,4,power_law_1.2,9.234130859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,4,power_law_1.2,0.10303360223770142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,4,power_law_1.2,0.10889600515365601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,4,power_law_1.2,0.12830079793930055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,4,power_law_1.2,13.69488067626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,4,power_law_1.2,0.12705279588699342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,4,power_law_1.2,0.12788480520248413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,4,power_law_1.2,0.0912447988986969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,4,power_law_1.2,0.13423999547958373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,4,power_law_1.2,0.14572160243988036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,4,power_law_1.2,0.08511999845504761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,4,power_law_1.2,0.15345280170440673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,4,power_law_1.2,0.09994879961013795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,4,power_law_1.2,0.15048960447311402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,4,power_law_1.2,0.16547839641571044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,4,power_law_1.2,0.0998207986354828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,4,power_law_1.2,40.242105102539064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,4,power_law_1.2,0.14396799802780152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,4,power_law_1.2,0.10088959932327271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,4,power_law_1.2,0.16417920589447021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,4,power_law_1.2,0.10101759433746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,4,power_law_1.2,0.15864959955215455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,4,power_law_1.2,0.10210560560226441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,4,power_law_1.2,0.17527040243148803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,4,power_law_1.2,0.1016319990158081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,4,power_law_1.2,0.20515201091766358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,4,power_law_1.2,0.10677759647369385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,4,power_law_1.2,0.04919039905071258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,4,power_law_1.2,0.20828158855438234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,4,power_law_1.2,0.10871039628982544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,4,power_law_1.2,0.22933759689331054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,2,balanced,0.08195733527342479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,4,power_law_1.2,0.11636480093002319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,4,power_law_1.2,0.04790399968624115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,4,power_law_1.2,0.2842751979827881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,4,power_law_1.2,0.1177664041519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,4,power_law_1.2,0.046700799465179445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,4,power_law_1.2,0.12228480577468873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,4,power_law_1.2,0.34075520038604734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,2,balanced,0.11786666512489319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,4,power_law_1.2,0.058233600854873654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,4,power_law_1.2,0.14370559453964232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,4,power_law_1.2,0.5481791973114014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,2,balanced,0.20956265926361084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,4,power_law_1.2,0.058713597059249875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,2,balanced,0.056314667065938316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,2,balanced,0.04801600178082784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,2,balanced,0.027509334186712902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,4,power_law_1.2,0.5548223972320556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,4,power_law_1.2,0.059961599111557004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,2,balanced,0.21655466159184775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,4,power_law_1.2,0.9913087844848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,4,power_law_1.2,0.14807679653167724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,2,balanced,0.06637333333492279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,4,power_law_1.2,0.07519360184669495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,2,balanced,0.07943999767303467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,2,balanced,0.0276053324341774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,4,power_law_1.2,1.3138367652893066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,2,balanced,0.2111519972483317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,4,power_law_1.2,0.06626560091972351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,2,balanced,0.09306666254997253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,2,balanced,0.02716800073782603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,4,power_law_1.2,1.6981184005737304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,2,balanced,0.1241973340511322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,4,power_law_1.2,0.15593600273132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,4,power_law_1.2,0.132915198802948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,2,balanced,0.20998400449752808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,4,power_law_1.2,2.361017608642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,2,balanced,0.033402666449546814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,2,balanced,0.09880000352859497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,2,balanced,0.21239999930063883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,2,balanced,0.12772267063458762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,4,power_law_1.2,0.13279999494552613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,4,power_law_1.2,0.20332159996032714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,4,power_law_1.2,3.197427177429199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,2,balanced,0.21241066853205362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,2,balanced,0.045408000548680626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,2,balanced,0.09573866923650105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,2,balanced,0.12460800011952718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,4,power_law_1.2,0.13691519498825072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.2,0.210265588760376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,4,power_law_1.2,5.2073406219482425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,2,balanced,0.04568533102671305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,2,balanced,0.21180800596872965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,2,balanced,0.09711999694506328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,4,power_law_1.2,0.13383040428161622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,2,balanced,0.1276639997959137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.2,0.3440576076507568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,4,power_law_1.2,6.243360137939453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,2,balanced,0.04055466751257578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,2,balanced,0.21177599827448526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,4,power_law_1.2,0.12781440019607543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,2,balanced,0.1301653285821279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,2,balanced,0.09882666667302449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.2,0.41760640144348143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,4,power_law_1.2,8.883052825927734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,2,balanced,0.04781866570313772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,2,balanced,0.2227999965349833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,4,power_law_1.2,0.10012160539627075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,2,balanced,0.09859200318654378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,2,balanced,0.12784000237782797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.2,0.6367680072784424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,2,balanced,0.0536106675863266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.2,0.8204799652099609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,2,balanced,0.09920000036557515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,2,balanced,0.21319466829299927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,4,power_law_1.2,0.14958720207214354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,4,power_law_1.2,18.22344970703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,2,balanced,0.1273973286151886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,2,balanced,0.04576533536116282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,2,balanced,0.09943999846776326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,2,balanced,0.12945066889127096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,4,power_law_1.2,0.12507519721984864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,2,balanced,0.2165279984474182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.2,1.184921646118164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,2,balanced,0.054085334142049156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,2,balanced,0.13285866379737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,4,power_law_1.2,0.16194560527801513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,2,balanced,0.23771733045578003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,2,balanced,0.10058666268984477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,2,balanced,0.10773866375287373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.2,1.4793408393859864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,2,balanced,0.13531733552614847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,2,balanced,0.10212799906730652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,2,balanced,0.2412160038948059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,4,power_law_1.2,0.21151359081268312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,2,balanced,0.10166399677594502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,2,balanced,0.13502400120099387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,2,balanced,0.10549333691596985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.2,2.141119956970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,2,balanced,0.246288001537323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.2,0.2734272003173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,2,balanced,0.10709333419799805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,2,balanced,0.14103999733924866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,2,balanced,0.13618133465449014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,2,balanced,0.25144533316294354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.2,4.125356674194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,2,balanced,0.09278933207194011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.2,0.3891072034835815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,2,balanced,0.11377066373825073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,2,balanced,0.15727466344833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,2,balanced,0.30962133407592773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,2,balanced,0.07507200042406718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.2,0.5204031944274903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,2,balanced,0.11558933059374492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,2,balanced,0.16088533401489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,4,power_law_1.2,3.8913089752197267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,2,balanced,0.3022720019022624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,2,balanced,0.09585066636403401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.2,0.7462143898010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,2,balanced,0.12061867117881775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,2,balanced,0.16225600242614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,4,power_law_1.2,8.220588684082031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,2,balanced,0.3854666550954183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,2,balanced,0.11929066975911458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.2,0.9896832466125488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,2,balanced,0.14255467057228088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,2,balanced,0.19677333037058511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.2,1.4565759658813477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,4,power_law_1.2,12.93966064453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,2,balanced,0.14592533310254416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,2,balanced,0.416869322458903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.2,1.9175167083740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,2,balanced,0.1534986694653829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,2,balanced,0.20266133546829224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,2,balanced,0.5724853277206421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,2,balanced,0.20367467403411865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.2,2.8713151931762697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,2,balanced,0.18731733163197836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,2,balanced,0.25872000058492023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,2,balanced,0.25303467114766437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,2,balanced,0.7463733355204264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.2,3.799737548828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,2,balanced,0.22613867123921713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,2,balanced,0.3613599936167399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,2,balanced,0.29201066493988037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,4,power_law_1.2,4.730559921264648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,2,balanced,0.29517332712809247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,2,balanced,1.047887961069743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,2,balanced,0.3996373414993286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,2,balanced,0.4719626506169637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,4,power_law_1.2,7.564678192138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,2,balanced,0.37350932757059735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,2,balanced,1.4695146878560383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,2,balanced,0.5050026575724283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,2,balanced,0.515770673751831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,2,balanced,0.6811520258585612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,4,power_law_1.2,15.363801574707031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,2,balanced,0.7197439670562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,2,balanced,2.1734026273091636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,2,balanced,0.656602660814921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,2,balanced,0.8898560206095377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,2,balanced,0.9409973621368408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,2,balanced,0.9253866672515869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,2,balanced,2.7649758656819663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,2,balanced,1.308832009633382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,2,balanced,1.3710932731628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,2,balanced,1.212656021118164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,2,balanced,1.7295519510904949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,2,balanced,3.431999842325846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,2,balanced,1.759130636850993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,2,balanced,1.5190399487813313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,2,balanced,2.1498026847839355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,2,power_law_1.01,0.09649279713630676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,2,balanced,2.1732799212137857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,2,power_law_1.01,0.11198079586029053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,2,balanced,5.412682851155599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,2,power_law_1.01,0.15365760326385497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,2,power_law_1.01,0.21552639007568358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,2,power_law_1.01,0.21016960144042968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,2,balanced,2.416031996409098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,2,power_law_1.01,0.20964479446411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,2,power_law_1.01,0.22016639709472657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,2,power_law_1.01,0.21855359077453612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,2,balanced,3.4039894739786782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,2,balanced,3.4046506881713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,2,power_law_1.01,0.22172160148620607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,2,power_law_1.01,0.22576639652252198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,2,power_law_1.01,0.23620479106903075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,2,power_law_1.01,0.24581120014190674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,2,power_law_1.01,0.2627583980560303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,2,power_law_1.01,0.27387518882751466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,2,power_law_1.01,0.308953595161438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,2,power_law_1.01,0.3112384080886841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,2,balanced,10.877012888590494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,2,power_law_1.01,0.3130111932754517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,2,power_law_1.01,0.3619647979736328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,2,power_law_1.01,0.41634559631347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,2,power_law_1.01,0.5403200149536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,2,power_law_1.01,0.6353663921356201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,2,balanced,4.879679997762044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,2,power_law_1.01,0.9639424324035645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,2,balanced,6.704837163289388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,2,balanced,6.746682484944661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,2,power_law_1.01,0.9941887855529785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,2,power_law_1.01,1.5991807937622071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,2,power_law_1.01,2.374367904663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,2,power_law_1.01,3.2174846649169924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,2,power_law_1.01,3.998463821411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,2,power_law_1.01,5.2899009704589846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,2,power_law_1.01,7.663897705078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,2,power_law_1.01,18.731430053710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,2,power_law_1.01,0.06946560144424438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,2,power_law_1.01,0.07873280048370361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,2,power_law_1.01,0.08753920197486878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,2,power_law_1.01,0.12410880327224731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,2,power_law_1.01,0.12483199834823608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,2,power_law_1.01,0.12691839933395385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,2,power_law_1.01,0.13283200263977052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,2,power_law_1.01,0.1349503993988037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,2,power_law_1.01,0.13092479705810547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,2,power_law_1.01,0.13838080167770386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,2,power_law_1.01,0.13720959424972534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,2,power_law_1.01,0.1410752058029175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,2,power_law_1.01,0.145414400100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,2,power_law_1.01,0.156876802444458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,2,power_law_1.01,0.17504639625549318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,2,power_law_1.01,0.17802879810333253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,2,power_law_1.01,0.19312000274658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,2,power_law_1.01,0.22558720111846925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,2,power_law_1.01,0.25360000133514404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,2,power_law_1.01,0.35528318881988524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,2,power_law_1.01,0.41143040657043456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,2,power_law_1.01,0.5952767848968505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,2,power_law_1.01,0.6923264026641845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,2,power_law_1.01,0.9089407920837402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,2,power_law_1.01,1.4304256439208984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,2,power_law_1.01,1.9956287384033202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,2,power_law_1.01,2.6840383529663088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,2,power_law_1.01,2.938912010192871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,2,power_law_1.01,0.06801279783248901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,2,power_law_1.01,5.260031890869141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,2,power_law_1.01,0.07038080096244811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,2,power_law_1.01,0.07080960273742676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,2,power_law_1.01,10.16872329711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,2,power_law_1.01,0.09244800209999085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,2,power_law_1.01,0.09220479726791382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,2,power_law_1.01,0.09596800208091735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,2,power_law_1.01,0.09657599925994872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,2,power_law_1.01,0.09712640047073365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,2,power_law_1.01,0.09793919920921326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,2,power_law_1.01,0.09847679734230042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,2,power_law_1.01,0.10093439817428589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,2,power_law_1.01,0.10042239427566528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,2,power_law_1.01,0.10720640420913696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,2,power_law_1.01,0.11050239801406861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,2,power_law_1.01,0.027084800601005554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,2,power_law_1.01,0.12152960300445556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,2,power_law_1.01,0.027289599180221558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,2,power_law_1.01,0.12657920122146607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,2,power_law_1.01,0.025785601139068602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,2,power_law_1.01,0.14018559455871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,2,power_law_1.01,0.031539198756217954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,2,power_law_1.01,0.04407680034637451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,2,power_law_1.01,0.16069120168685913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,2,power_law_1.01,0.0433023989200592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.01,0.19861760139465331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,2,power_law_1.01,0.03847039937973022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.01,0.2357503890991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,2,power_law_1.01,0.046316799521446225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.01,0.31254398822784424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,2,power_law_1.01,0.05028480291366577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.01,0.37201919555664065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,2,power_law_1.01,0.04368639886379242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.01,0.5930560111999512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,2,power_law_1.01,0.053235197067260744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.01,0.8049407958984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,2,power_law_1.01,0.10400639772415161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.01,0.9320256233215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,2,power_law_1.01,0.1045375943183899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,2,power_law_1.01,0.13495680093765258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.01,1.4215167999267577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,2,power_law_1.01,0.0923200011253357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.01,1.949894332885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,2,power_law_1.01,0.07417600154876709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,2,power_law_1.01,0.09474560022354125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,2,power_law_1.01,2.307481575012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,2,power_law_1.01,0.11733759641647339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,2,power_law_1.01,3.836057662963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.01,0.14304640293121337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.01,0.20101759433746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,2,power_law_1.01,7.555654144287109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.01,0.25635199546813964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.01,0.3600383996963501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.01,0.46874241828918456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,2,power_law_1.2,0.0688256025314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,2,power_law_1.2,0.1122048020362854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.01,0.6789887905120849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,2,power_law_1.2,0.06886399984359741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.01,0.8907456398010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,2,power_law_1.2,0.1298815965652466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,2,power_law_1.2,0.08456959724426269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.01,1.3088512420654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,2,power_law_1.2,0.15537919998168945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,2,power_law_1.2,0.12314239740371705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.01,1.7267072677612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,2,power_law_1.2,0.2087552070617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,2,power_law_1.2,0.12357759475708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,2,power_law_1.01,2.150598335266113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,2,power_law_1.2,0.21437439918518067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,2,power_law_1.2,0.12632960081100464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,2,power_law_1.2,0.2133631944656372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,2,power_law_1.01,3.4025089263916017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,2,power_law_1.2,0.1293503999710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,2,power_law_1.2,0.22726399898529054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,2,power_law_1.2,0.13276159763336182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,2,power_law_1.01,6.7459266662597654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,2,power_law_1.2,0.22471680641174316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,2,power_law_1.2,0.13341439962387086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,2,power_law_1.2,0.2266239881515503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,2,power_law_1.2,0.13820159435272217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,2,power_law_1.2,0.06255360245704651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,2,power_law_1.2,0.23209600448608397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,2,power_law_1.2,0.13838080167770386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,2,power_law_1.2,0.027059200406074523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,2,power_law_1.2,0.23336319923400878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,2,power_law_1.2,0.06526079773902893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,1,balanced,0.07527466615041097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,2,power_law_1.2,0.14655359983444213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,2,power_law_1.2,0.0685375988483429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,2,power_law_1.2,0.2737600088119507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,2,power_law_1.2,0.027302399277687073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,1,balanced,0.1188800036907196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,2,power_law_1.2,0.09280639886856079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,2,power_law_1.2,0.14482560157775878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,2,power_law_1.2,0.09485440254211426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,2,power_law_1.2,0.026214399933815004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,2,power_law_1.2,0.24391679763793944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,1,balanced,0.2060533364613851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,2,power_law_1.2,0.15843839645385743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,2,power_law_1.2,0.09545599818229675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,2,power_law_1.2,0.03180800080299377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,1,balanced,0.21013333400090536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,2,power_law_1.2,0.27217280864715576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,2,power_law_1.2,0.09608960151672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,2,power_law_1.2,0.18120319843292237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,1,balanced,0.20958934227625528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,2,power_law_1.2,0.0438400000333786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,2,power_law_1.2,0.09711999893188476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,2,power_law_1.2,0.29828479290008547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,2,power_law_1.2,0.17642879486083984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,2,power_law_1.2,0.04337919950485229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,1,balanced,0.20920000473658243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,2,power_law_1.2,0.0982591986656189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,2,power_law_1.2,0.18794239759445192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,2,power_law_1.2,0.3466559886932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,2,power_law_1.2,0.03839359879493713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,1,balanced,0.21203200022379556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,2,power_law_1.2,0.09879040122032165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,1,balanced,0.21344532569249472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,2,power_law_1.2,0.04593279957771301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,2,power_law_1.2,0.3182015895843506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,2,power_law_1.2,0.09904639720916748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,2,power_law_1.2,0.22569599151611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,1,balanced,0.21133333444595337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,2,power_law_1.2,0.04856959879398346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,2,power_law_1.2,0.04658559858798981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,2,power_law_1.2,0.1044927954673767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,1,balanced,0.21070400873819986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,2,power_law_1.2,0.37087359428405764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,2,power_law_1.2,0.27127039432525635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,2,power_law_1.2,0.05329279899597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,2,power_law_1.2,0.10686719417572021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,1,balanced,0.2206773360570272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,2,power_law_1.2,0.4314432144165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,2,power_law_1.2,0.3575680017471313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,2,power_law_1.2,0.1043328046798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,2,power_law_1.2,0.11185280084609986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,2,power_law_1.2,0.10505599975585937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,1,balanced,0.21771732966105142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,2,power_law_1.2,0.4441472053527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,2,power_law_1.2,0.5760511875152587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,2,power_law_1.2,0.12450560331344604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,2,power_law_1.2,0.13506560325622557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,2,power_law_1.2,0.5411136150360107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,1,balanced,0.21991999944051108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,2,power_law_1.2,0.7092351913452148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,2,power_law_1.2,0.08795520067214965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,2,power_law_1.2,0.12823679447174072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,2,power_law_1.2,0.07304959893226623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,2,power_law_1.2,0.8449215888977051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,1,balanced,0.2307413419087728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,2,power_law_1.2,0.14017280340194702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,2,power_law_1.2,0.839571189880371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,2,power_law_1.2,0.09428480267524719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,2,power_law_1.2,0.16520320177078246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,2,power_law_1.2,0.9938495635986329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,1,balanced,0.24764800071716309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,2,power_law_1.2,0.11710079908370971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,2,power_law_1.2,1.2163200378417969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.2,0.18531839847564696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,2,power_law_1.2,1.505305576324463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,1,balanced,0.2514293392499288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.2,0.14248960018157958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.2,0.24077439308166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,2,power_law_1.2,1.7021823883056642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.2,0.2883968114852905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.2,0.19968639612197875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,1,balanced,0.24914666016896567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,2,power_law_1.2,2.2540351867675783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,2,power_law_1.2,2.2267583847045898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.2,0.44208641052246095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.2,0.25502719879150393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,2,power_law_1.2,2.5808576583862304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,1,balanced,0.30220266183217365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.2,0.5877247810363769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.2,0.35820159912109373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,2,power_law_1.2,3.953657531738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,1,balanced,0.30666667222976685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,2,power_law_1.2,3.1655168533325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,1,balanced,0.053530668218930565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.2,0.7818496227264404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.2,0.4678976058959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,1,balanced,0.39632534980773926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,2,power_law_1.2,5.3833984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,2,power_law_1.2,5.164300918579102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.2,0.6772672176361084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,1,balanced,0.0769760012626648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.2,0.9962112426757812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,1,balanced,0.47062933444976807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.2,0.890835189819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.2,1.420851230621338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,1,balanced,0.11764267086982727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,2,power_law_1.2,5.7003520965576175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,1,balanced,0.6205706596374512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,2,power_law_1.2,11.440364837646484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.2,1.31080322265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,1,balanced,0.12194133798281352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.2,1.7953407287597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.2,1.7276735305786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,1,balanced,0.12097600102424622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,1,balanced,0.7508320013682047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,2,power_law_1.2,9.71435546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,2,power_law_1.2,2.6661184310913084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,2,power_law_1.2,2.1477376937866213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,1,balanced,0.12363732854525249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,1,balanced,0.12586133678754172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,2,power_law_1.2,3.895769500732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,2,power_law_1.2,3.40302734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,1,balanced,1.0746666590372722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,1,balanced,0.125109334786733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,2,power_law_1.2,17.563763427734376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,2,power_law_1.2,6.8211006164550785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,2,power_law_1.2,8.78371810913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,1,balanced,0.12503467003504434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,1,balanced,1.451072057088216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,1,balanced,0.12774933377901712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,1,balanced,0.1300373375415802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,1,balanced,0.1348373293876648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,1,balanced,0.13408533732096353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,1,balanced,2.122858683268229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,1,balanced,0.14044800400733948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,1,balanced,0.1593706707159678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,1,balanced,2.7705065409342446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,1,balanced,0.16065067052841187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,1,balanced,0.16074132919311523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,1,balanced,0.19248533248901367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,1,balanced,3.416341463724772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,1,balanced,0.20431999365488687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,1,balanced,0.26127467552820843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,1,balanced,0.2991679906845093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,1,balanced,0.043765331308046974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,1,balanced,0.4071413278579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,1,balanced,0.06402133405208588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,1,balanced,0.08930666248003642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,1,balanced,0.5165599981943766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,1,balanced,0.09365866581598918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,1,balanced,5.515125274658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,1,balanced,0.09239466985066731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,1,balanced,0.7272960344950358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,1,balanced,0.09267733494440715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,1,balanced,0.09426132837931316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,1,balanced,0.09486933549245198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,1,balanced,0.9423946539560953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,1,balanced,0.09454933802286784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,1,balanced,0.09620267152786255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,1,balanced,0.0972160001595815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,1,balanced,0.09956266482671101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,1,balanced,1.3697120348612468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,1,balanced,0.10008000334103902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,1,balanced,0.10950400431950887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,1,balanced,0.11896533767382304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,1,balanced,0.12127466996510823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,1,balanced,0.02332266668478648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,1,balanced,1.8029120763142903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,1,balanced,0.13405332962671915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,1,balanced,0.15930666526158652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,1,balanced,0.02325333406527837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,1,balanced,10.778848012288412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,1,balanced,0.0235359991590182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,1,balanced,0.1679840087890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,1,balanced,0.02510400116443634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,1,balanced,2.225829283396403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,1,balanced,0.21420266230901083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,1,balanced,0.02997333308060964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,1,balanced,0.040181333820025124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,1,balanced,0.2521386742591858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,1,balanced,0.035504000882307686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,1,balanced,0.039861333866914116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,1,balanced,0.33320534229278564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,1,balanced,0.035760000348091125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,1,balanced,0.41913068294525146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,1,balanced,0.03682666768630346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,1,balanced,3.4822187423706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,1,balanced,0.04256533086299896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,1,balanced,0.5815253257751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,1,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,1,balanced,0.04154133299986521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,1,balanced,0.743781328201294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,1,balanced,0.04789866507053375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,1,balanced,0.10393599669138591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,1,balanced,0.10599999626477559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,1,balanced,0.1336426635583242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,1,balanced,1.0694560209910076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,1,balanced,0.07567466795444489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,1,balanced,0.09273599584897359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,1,balanced,0.12825600306193033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,1,balanced,1.4022986094156902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,1,balanced,0.15332800149917603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,1,balanced,6.8557173411051435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,1,balanced,0.2149333357810974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,1,balanced,0.2723306616147359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,1,balanced,1.7282986640930176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,1,balanced,0.386624018351237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,1,balanced,0.5026666720708212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,1,balanced,2.801205317179362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,1,balanced,0.7329493363698324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,1,balanced,0.960922638575236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,1,balanced,1.1916800340016682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,1,balanced,5.495418548583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,1,power_law_1.01,0.07300480008125305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,1,balanced,1.881589412689209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,1,power_law_1.01,0.10005760192871094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,1,power_law_1.01,0.12350080013275147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,1,power_law_1.01,0.19471999406814575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,1,power_law_1.01,0.2006848096847534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,1,power_law_1.01,0.2012160062789917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,1,power_law_1.01,0.20473599433898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,1,power_law_1.01,0.21337599754333497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,1,power_law_1.01,0.21772160530090331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,1,power_law_1.01,0.22421119213104249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,1,balanced,3.7226826349894204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,1,power_law_1.01,0.22174720764160155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,1,power_law_1.01,0.23182079792022706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,1,power_law_1.01,0.23452160358428956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,1,power_law_1.01,0.23754239082336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,1,power_law_1.01,0.25331840515136717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,1,power_law_1.01,0.26679039001464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,1,power_law_1.01,0.2839871883392334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,1,power_law_1.01,0.3158655881881714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,1,power_law_1.01,0.37233920097351075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,1,power_law_1.01,0.02194560021162033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,1,power_law_1.01,0.04677119851112366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,1,power_law_1.01,0.05120639801025391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,1,power_law_1.01,0.4319551944732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,1,power_law_1.01,0.5292736053466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,1,power_law_1.01,0.06225280165672302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,1,power_law_1.01,0.022623999416828154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,1,power_law_1.01,0.05756800174713135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,1,power_law_1.01,0.6990143775939941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,1,power_law_1.01,0.07669119834899903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,1,power_law_1.01,0.11743359565734864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,1,power_law_1.01,0.849459171295166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,1,power_law_1.01,0.022950400412082673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,1,power_law_1.01,0.06179839968681335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,1,power_law_1.01,0.12041599750518799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,1,power_law_1.01,0.02361599951982498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,1,power_law_1.01,0.12190719842910766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,1,power_law_1.01,1.1972415924072266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,1,power_law_1.01,0.09027199745178223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,1,power_law_1.01,0.029267200827598573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,1,power_law_1.01,0.12428159713745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,1,power_law_1.01,1.5445952415466309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,1,power_law_1.01,0.09071999788284302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,1,power_law_1.01,0.12600959539413453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,1,power_law_1.01,2.1873088836669923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,1,power_law_1.01,0.03895680010318756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,1,power_law_1.01,0.09327999949455261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,1,power_law_1.01,0.1292415976524353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,1,power_law_1.01,0.03482879996299744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,1,power_law_1.01,2.910495948791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,1,power_law_1.01,0.12981760501861572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,1,power_law_1.01,0.10680960416793824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,1,power_law_1.01,0.03976320028305054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,1,power_law_1.01,0.13041919469833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,1,power_law_1.01,3.5901695251464845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,1,power_law_1.01,0.09308159947395325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,1,power_law_1.01,0.14069119691848755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,1,power_law_1.01,0.03480319976806641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,1,power_law_1.01,5.652492904663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,1,power_law_1.01,0.14024959802627562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,1,power_law_1.01,0.09363200068473816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,1,power_law_1.01,0.14644479751586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,1,power_law_1.01,0.15850239992141724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,1,power_law_1.01,10.957138824462891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,1,power_law_1.01,0.09442560076713562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,1,power_law_1.01,0.03581439852714539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,1,power_law_1.01,0.16560640335083007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,1,power_law_1.01,0.09774079918861389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,1,power_law_1.2,0.07366399765014649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,1,power_law_1.01,0.17536640167236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,1,power_law_1.01,0.10157439708709717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,1,power_law_1.01,0.041868799924850465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,1,power_law_1.01,0.20134398937225342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,1,power_law_1.01,0.10341119766235352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,1,power_law_1.2,0.09255679845809936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,1,power_law_1.01,0.23069438934326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,1,power_law_1.01,0.038553598523139956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,1,power_law_1.01,0.11008000373840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,1,power_law_1.2,0.13337600231170654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,1,power_law_1.01,0.2893824100494385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,1,power_law_1.01,0.3459455966949463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,1,power_law_1.01,0.1215999960899353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,1,power_law_1.2,0.20412800312042237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,1,power_law_1.01,0.44934401512145994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,1,power_law_1.01,0.040275201201438904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,1,power_law_1.01,0.12384639978408814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,1,power_law_1.2,0.20581119060516356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,1,power_law_1.01,0.5643775939941407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,1,power_law_1.01,0.14350080490112305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,1,power_law_1.01,0.04787200093269348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,1,power_law_1.2,0.20734720230102538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,1,power_law_1.01,0.7783232212066651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,1,power_law_1.01,0.16075520515441893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,1,power_law_1.01,0.9909824371337891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,1,power_law_1.01,0.10217599868774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,1,power_law_1.2,0.20945279598236083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.01,0.18472959995269775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,1,power_law_1.01,1.4158783912658692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,1,power_law_1.2,0.21810560226440429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,1,power_law_1.01,0.10375679731369018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.01,0.22572801113128663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,1,power_law_1.01,1.8473215103149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,1,power_law_1.2,0.2169408082962036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,1,power_law_1.01,0.13212800025939941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.01,0.2638015985488892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,1,power_law_1.2,0.2167423963546753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,1,power_law_1.01,2.2472255706787108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.01,0.3509183883666992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,1,power_law_1.01,0.0735360026359558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,1,power_law_1.2,0.22064640522003173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,1,power_law_1.01,3.506700897216797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.01,0.43505277633666994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.01,0.09162880182266235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,1,power_law_1.2,0.23128960132598878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.01,0.6089471817016602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,1,power_law_1.01,6.920082855224609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,1,power_law_1.2,0.05175039768218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.01,0.12548480033874512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.01,0.7734208106994629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,1,power_law_1.2,0.2392064094543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,1,power_law_1.2,0.061919999122619626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.01,0.15107200145721436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.01,1.1062015533447265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,1,power_law_1.2,0.23870720863342285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,1,power_law_1.2,0.07382400035858154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.01,0.21159679889678956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.01,1.474892807006836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,1,power_law_1.2,0.25246078968048097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,1,power_law_1.2,0.1152575969696045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.01,0.2688447952270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,1,power_law_1.2,0.2606208086013794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.01,1.8226112365722655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.01,0.3829184055328369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,1,power_law_1.2,0.12097920179367065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,1,power_law_1.2,0.28682880401611327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.01,0.5027904033660888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,1,power_law_1.01,2.9515071868896485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,1,power_law_1.2,0.12218879461288452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,1,power_law_1.2,0.32842240333557127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.01,0.729414415359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,1,power_law_1.01,5.457062530517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,1,power_law_1.2,0.12377599477767945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.01,0.9599871635437012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,1,power_law_1.2,0.3832319974899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,1,power_law_1.2,0.12537599802017213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.01,1.1897536277770997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,1,power_law_1.2,0.45331201553344724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,1,power_law_1.01,1.8826559066772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,1,power_law_1.2,0.12887680530548096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,1,power_law_1.01,3.722284698486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,1,power_law_1.2,0.5277952194213867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,1,power_law_1.2,0.13136639595031738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,1,power_law_1.2,0.6894464015960693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,1,power_law_1.2,0.131167995929718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,1,power_law_1.2,0.8358464241027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,1,power_law_1.2,0.13616000413894652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,1,power_law_1.2,1.209113597869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,1,power_law_1.2,0.14327679872512816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,1,power_law_1.2,1.5565376281738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,1,power_law_1.2,0.1442944049835205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,1,power_law_1.2,0.15896960496902465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,1,power_law_1.2,2.2077119827270506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,1,power_law_1.2,0.16679680347442627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,1,power_law_1.2,2.927257537841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,1,power_law_1.2,0.17825280427932738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,1,power_law_1.2,0.2038655996322632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,1,power_law_1.2,3.588819122314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,1,power_law_1.2,0.2335360050201416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,1,power_law_1.2,0.2835776090621948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,1,power_law_1.2,5.67432975769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,1,power_law_1.2,0.3459968090057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,1,power_law_1.2,0.4603072166442871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,1,power_law_1.2,10.942623901367188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,1,power_law_1.2,0.5560575962066651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,1,power_law_1.2,0.778707218170166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,1,power_law_1.2,0.9895999908447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,1,power_law_1.2,1.4206591606140138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,1,power_law_1.2,1.8290048599243165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,1,power_law_1.2,2.2496448516845704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,1,power_law_1.2,3.507769775390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,1,power_law_1.2,6.91229476928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,1,power_law_1.2,0.045388799905776975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,1,power_law_1.2,0.05447040200233459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,1,power_law_1.2,0.06173440217971802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,1,power_law_1.2,0.0907904028892517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,1,power_law_1.2,0.09048320055007934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,1,power_law_1.2,0.022457599639892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,1,power_law_1.2,0.09141119718551635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,1,power_law_1.2,0.023097600042819976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,1,power_law_1.2,0.09217280149459839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,1,power_law_1.2,0.02287999987602234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,1,power_law_1.2,0.09282559752464295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,1,power_law_1.2,0.023731200397014617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,1,power_law_1.2,0.09415040016174317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,1,power_law_1.2,0.0293503999710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,1,power_law_1.2,0.09444479942321778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,1,power_law_1.2,0.038796800374984744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,1,power_law_1.2,0.09773439764976502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,1,power_law_1.2,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,1,power_law_1.2,0.09998080134391785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,1,power_law_1.2,0.03875199854373932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,1,power_law_1.2,0.1034559965133667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,1,power_law_1.2,0.03450239896774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,1,power_law_1.2,0.11290880441665649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,1,power_law_1.2,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,1,power_law_1.2,0.12194559574127198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,1,power_law_1.2,0.04126720130443573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,1,power_law_1.2,0.12666879892349242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,1,power_law_1.2,0.038431999087333676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,1,power_law_1.2,0.1423616051673889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,1,power_law_1.2,0.03969280123710632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,1,power_law_1.2,0.1608191967010498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,1,power_law_1.2,0.04695680141448975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.2,0.1850816011428833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,1,power_law_1.2,0.10222079753875732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.2,0.2265791893005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,1,power_law_1.2,0.10396159887313842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.2,0.2705280065536499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,1,power_law_1.2,0.13163520097732545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.2,0.3529344081878662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,1,power_law_1.2,0.07281280159950257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.2,0.43319039344787597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.2,0.09159680008888245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.2,0.610975980758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.2,0.12583680152893068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.2,0.7704895973205567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.2,0.15168000459671022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.2,1.1224831581115722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.2,0.21256959438323975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.2,1.4735296249389649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.2,0.2686592102050781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.2,1.841926383972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.2,0.3832319974899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,1,power_law_1.2,2.940358352661133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.2,0.49939842224121095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,1,power_law_1.2,5.464051055908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.2,0.7292863845825195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.2,0.9611904144287109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.2,1.1880895614624023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,1,power_law_1.2,1.880953598022461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,1,power_law_1.2,3.7196414947509764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,4,balanced,0.06942399839560191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,4,balanced,0.092031995455424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,4,balanced,0.1418826679388682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,4,balanced,0.2451146642367045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,4,balanced,0.44041065375010174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,4,balanced,0.8331039746602377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,4,balanced,0.8365386327107748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,4,balanced,0.8420373598734537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,4,balanced,0.8446719646453857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,4,balanced,0.8463892936706543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,4,balanced,0.8429599603017172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,4,balanced,0.8528213500976562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,4,balanced,0.850762685139974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,4,balanced,0.8517226378122965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,4,balanced,0.8629546960194906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,4,balanced,0.8647733529408773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,4,balanced,0.8722506364186605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,4,balanced,0.05013333261013031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,4,balanced,0.8887840112050375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,4,balanced,0.913162628809611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,4,balanced,0.06871999800205231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,4,balanced,0.9471147060394287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,4,balanced,0.09520533680915833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,4,balanced,0.148799995581309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,4,balanced,0.9837439854939779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,4,balanced,0.2504426638285319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,4,balanced,0.4453173478444417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,4,balanced,1.0335360368092854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,4,balanced,0.44893864790598553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,4,balanced,0.4487733443578084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,4,balanced,1.0882506370544434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,4,balanced,0.4556586742401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,4,balanced,0.4542880058288574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,4,balanced,0.4578506549199422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,4,balanced,1.3575199445088704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,4,balanced,0.0562720000743866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,4,balanced,0.4598986705144246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,4,balanced,0.05353599786758423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,4,balanced,0.06366399923960368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,4,balanced,0.46288001537323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,4,balanced,0.05167999863624573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,4,balanced,0.08267733454704285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,4,balanced,1.3707146644592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,4,balanced,0.4670240084330241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,4,balanced,0.06472533444563548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,4,balanced,0.11750400066375732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,4,balanced,0.47497065862019855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,4,balanced,0.09358400106430054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,4,balanced,0.17976532379786173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,4,balanced,0.4766240119934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,4,balanced,0.14432000120480856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,4,balanced,0.2977653344472249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,4,balanced,1.9553173383076985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,4,balanced,0.48767467339833576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,4,balanced,0.25149865945180255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,4,balanced,0.30162666241327923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,4,balanced,0.26311999559402466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,4,balanced,0.5079573392868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,4,balanced,0.3040800094604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,4,balanced,0.2752053340276082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,4,balanced,0.30456533034642536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,4,balanced,0.2637386719385783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,4,balanced,0.5236266851425171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,4,balanced,0.3046293258666992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,4,balanced,0.2688586711883545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,4,balanced,0.3076266646385193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,4,balanced,0.5586400032043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,4,balanced,2.172442595163981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,4,balanced,0.2815413276354472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,4,balanced,0.31003199021021527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,4,balanced,0.6021920045216879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,4,balanced,0.29896533489227295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,4,power_law_1.01,0.1476032018661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,4,balanced,0.3142079909642537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,4,balanced,0.6643413305282593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,4,balanced,0.30773333708445233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,4,power_law_1.01,0.2388927936553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,4,balanced,0.3160960078239441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,4,balanced,0.31277332703272503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,4,power_law_1.01,0.135315203666687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,4,balanced,0.32309865951538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,4,power_law_1.01,0.09410560131072998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,4,balanced,0.42367998758951825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,4,balanced,0.7281493345896403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,4,balanced,2.9583679835001626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,4,power_law_1.01,0.19559040069580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,4,balanced,0.327237327893575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,4,power_law_1.01,0.144159996509552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,4,balanced,0.42317867279052734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,4,power_law_1.01,0.32318079471588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,4,balanced,0.9479359785715739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,4,power_law_1.01,0.10230400562286376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,4,balanced,0.3349813222885132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,4,balanced,0.42473065853118896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,4,power_law_1.01,0.41651201248168945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,4,power_law_1.01,0.12825599908828736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,4,balanced,0.34722665945688885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,4,balanced,0.6797440052032471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,4,power_law_1.01,0.6088064193725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,4,power_law_1.01,0.19555200338363649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,4,balanced,0.3609226544698079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,4,balanced,0.7563520272572836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,4,balanced,1.0358826319376628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,4,power_law_1.01,0.6599679946899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,4,power_law_1.01,0.22985599040985108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,4,balanced,0.39369599024454754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,4,balanced,0.6012320121129354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,4,power_law_1.01,0.614739179611206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,4,power_law_1.01,0.3321216106414795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,4,balanced,0.4275039831797282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,4,power_law_1.01,0.6634111881256104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,4,balanced,0.6432213385899862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,4,power_law_1.01,0.3343039989471436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,4,balanced,3.9637012481689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,4,balanced,0.48849066098531085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,4,power_law_1.01,0.7008895874023438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,4,balanced,1.5540639559427898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,4,power_law_1.01,0.35688960552215576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,4,balanced,0.7051733334859213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,4,power_law_1.01,0.7409408092498779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,4,balanced,0.553061326344808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,4,power_law_1.01,0.37348480224609376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,4,power_law_1.01,0.7398335933685303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,4,power_law_1.01,0.3697920083999634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,4,balanced,0.7687946955362955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,4,power_law_1.01,0.7742400169372559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,4,power_law_1.01,0.38971519470214844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,4,balanced,0.7423733075459799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,4,power_law_1.01,0.7555136203765869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,4,power_law_1.01,0.39210240840911864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,4,balanced,1.7333119710286458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,4,balanced,0.8925120035807291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,4,power_law_1.01,0.7731135845184326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,4,power_law_1.01,0.41169281005859376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,4,power_law_1.01,0.8416831970214844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,4,balanced,0.8588586648305258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,4,power_law_1.01,0.4312448024749756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,4,power_law_1.01,0.9065983772277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,4,power_law_1.01,0.440883207321167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,4,balanced,1.0332372983296711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,4,power_law_1.01,0.8869695663452148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,4,power_law_1.01,0.4634751796722412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,4,power_law_1.01,0.9889408111572265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,4,power_law_1.01,0.5045695781707764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,4,balanced,1.197002649307251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,4,power_law_1.01,1.0209407806396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,4,power_law_1.01,0.5337024211883545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,4,balanced,2.3585227330525718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,4,power_law_1.01,1.1525952339172363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,4,power_law_1.01,0.6004735946655273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,4,balanced,1.7140906651814778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,4,power_law_1.01,1.3355968475341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,4,power_law_1.01,0.6230207920074463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,4,power_law_1.01,1.736288070678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,4,power_law_1.01,0.7296127796173095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,4,power_law_1.01,1.9236415863037108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,4,balanced,8.532058715820312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,4,balanced,1.5497973759969075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,4,power_law_1.01,0.8611328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,4,power_law_1.01,2.589388847351074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,4,power_law_1.01,1.1483712196350098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,4,balanced,1.9406612714131672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,4,power_law_1.01,3.0453439712524415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,4,power_law_1.01,1.287059211730957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,4,power_law_1.01,3.657612609863281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,4,power_law_1.01,1.7269184112548828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,4,balanced,3.2566773096720376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,4,power_law_1.01,2.291622352600098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,4,power_law_1.01,5.460575866699219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,4,balanced,1.8903627395629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,4,power_law_1.01,2.7110015869140627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,4,power_law_1.01,11.293036651611327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,4,balanced,2.6411840120951333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,4,power_law_1.01,4.355596923828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,4,power_law_1.01,7.880883026123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,4,balanced,2.8691625595092773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,4,balanced,3.7861334482828775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,4,balanced,6.34661865234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,4,balanced,5.508576075236003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,4,power_law_1.01,0.08540160059928895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,4,balanced,7.51197878519694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,4,power_law_1.01,0.11291520595550537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,4,power_law_1.01,0.08057600259780884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,4,power_law_1.01,0.04952960014343262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,4,power_law_1.01,0.10496640205383301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,4,power_law_1.01,0.07651839852333069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,4,power_law_1.01,0.05359359979629517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,4,power_law_1.01,0.1379968047142029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,4,power_law_1.01,0.08526080250740051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,4,power_law_1.01,0.18094079494476317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,4,power_law_1.01,0.11555839776992798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,4,power_law_1.01,0.2213184118270874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,4,power_law_1.01,0.14847999811172485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,4,power_law_1.01,0.23556480407714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,4,power_law_1.01,0.18762880563735962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,4,power_law_1.01,0.24588799476623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,4,power_law_1.01,0.2062079906463623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,4,power_law_1.01,0.25834240913391116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,4,power_law_1.01,0.20791680812835694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,4,power_law_1.01,0.2571327924728394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,4,power_law_1.01,0.2171839952468872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,4,power_law_1.01,0.2711040019989014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,4,power_law_1.01,0.21699199676513672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,4,power_law_1.01,0.2714047908782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,4,power_law_1.01,0.21479039192199706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,4,power_law_1.01,0.2871551990509033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,4,power_law_1.01,0.22992000579833985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,4,power_law_1.01,0.2952703952789307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,4,power_law_1.01,0.24061439037322999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,4,power_law_1.01,0.29423999786376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,4,power_law_1.01,0.26721920967102053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,4,power_law_1.01,0.31475200653076174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,4,power_law_1.01,0.25447680950164797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,4,power_law_1.01,0.3322495937347412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,4,power_law_1.01,0.2663935899734497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.01,0.37090559005737306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,4,power_law_1.01,0.537766408920288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.01,0.5796288013458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.01,0.42407679557800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.01,0.429369592666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.01,0.4710400104522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.01,0.443782377243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.01,0.5581632137298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.01,0.5021696090698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.01,0.6725247859954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.01,0.5494527816772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.01,0.8509056091308593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.01,0.7041344165802002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.01,1.1071295738220215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.01,0.828012752532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.01,1.495206356048584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.01,1.2056447982788085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.01,1.8619903564453124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.01,1.5097599983215333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,4,power_law_1.01,2.171718406677246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,4,power_law_1.01,1.836511993408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,4,power_law_1.01,3.4041793823242186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,4,power_law_1.01,2.856492805480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,4,power_law_1.01,6.349772644042969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,4,power_law_1.01,5.394623947143555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,4,power_law_1.2,0.14520319700241088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,4,power_law_1.2,0.2221760034561157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,4,power_law_1.2,0.14030079841613768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,4,power_law_1.2,0.22784640789031982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,4,power_law_1.2,0.2869632005691528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,4,power_law_1.2,0.38229119777679443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,4,power_law_1.2,0.5634111881256103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,4,power_law_1.2,0.6245696067810058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,4,power_law_1.2,0.667622423171997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,4,power_law_1.2,0.6639423847198487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,4,power_law_1.2,0.6710591793060303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,4,power_law_1.2,0.6730944156646729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,4,power_law_1.2,0.08581759929656982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,4,power_law_1.2,0.7148287773132325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,4,power_law_1.2,0.09218559861183166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,4,power_law_1.2,0.7683775901794434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,4,power_law_1.2,0.10645120143890381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,4,power_law_1.2,0.1423359990119934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,4,power_law_1.2,0.7690624237060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,4,power_law_1.2,0.07614079713821412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,4,power_law_1.2,0.08869119882583618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,4,power_law_1.2,0.8379072189331055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,4,power_law_1.2,0.10485759973526002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,4,power_law_1.2,0.13055360317230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,4,power_law_1.2,0.8516223907470704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,4,power_law_1.2,0.1329599976539612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,4,power_law_1.2,0.18806400299072265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,4,power_law_1.2,0.9927295684814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,4,power_law_1.2,0.150873601436615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,4,power_law_1.2,0.22062718868255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,4,power_law_1.2,0.9806143760681152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,4,power_law_1.2,0.2247999906539917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,4,power_law_1.2,0.3181632041931152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,4,power_law_1.2,1.089235210418701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,4,power_law_1.2,0.2329535961151123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,4,power_law_1.2,0.33130879402160646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,4,power_law_1.2,1.0570816040039062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,4,power_law_1.2,0.23878400325775145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,4,power_law_1.2,0.3450943946838379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,4,power_law_1.2,1.2726719856262207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,4,power_law_1.2,0.2471679925918579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,4,power_law_1.2,0.36092801094055177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,4,power_law_1.2,1.4203776359558105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,4,power_law_1.2,0.25404160022735595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,4,power_law_1.2,0.385318398475647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,4,power_law_1.2,1.8317312240600585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,4,power_law_1.2,0.26599678993225095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,4,power_law_1.2,0.3922431945800781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,4,power_law_1.2,2.098988723754883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,4,power_law_1.2,0.2794624090194702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,4,power_law_1.2,0.40953598022460935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,4,power_law_1.2,0.0483711987733841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,4,power_law_1.2,0.2891263961791992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,4,power_law_1.2,2.9515264511108397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,4,power_law_1.2,0.40741119384765623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,4,power_law_1.2,0.07402240037918091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,4,power_law_1.2,0.3031167984008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,4,power_law_1.2,4.089599990844727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,4,power_law_1.2,0.43585920333862305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,4,power_law_1.2,0.05397120118141174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,4,power_law_1.2,0.3019968032836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,4,power_law_1.2,3.9617599487304687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,4,power_law_1.2,0.45409278869628905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,4,power_law_1.2,0.08113920092582702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,4,power_law_1.2,6.5500541687011715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,4,power_law_1.2,0.3278271913528442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,4,power_law_1.2,0.48113279342651366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,4,power_law_1.2,0.10426239967346192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,2,balanced,0.07029866675535838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,4,power_law_1.2,0.3460864067077637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,4,power_law_1.2,12.615679931640624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,4,power_law_1.2,0.5440063953399659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,4,power_law_1.2,0.13370879888534545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,2,balanced,0.09480533003807068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.2,0.37579519748687745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,4,power_law_1.2,0.5530687808990479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,2,balanced,0.14523200194040933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,4,power_law_1.2,0.18731520175933838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.2,0.45810561180114745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,4,power_law_1.2,0.6190656185150146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,2,balanced,0.24417599042256674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,4,power_law_1.2,0.1953727960586548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.2,0.48348159790039064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,2,balanced,0.4434080123901367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,4,power_law_1.2,0.6617728233337402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,4,power_law_1.2,0.19844479560852052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.2,0.6254528045654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,2,balanced,0.8369440237681071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,2,balanced,0.05188799897829691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,4,power_law_1.2,0.8026687622070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,4,power_law_1.2,0.2027008056640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,2,balanced,0.8392480214436849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.2,0.6928895950317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,4,power_law_1.2,0.9213503837585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,2,balanced,0.8392586708068848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,2,balanced,0.06980800131956737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,4,power_law_1.2,0.2120448112487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,2,balanced,0.8416533470153809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.2,0.9066623687744141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,4,power_law_1.2,1.2023103713989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,2,balanced,0.09572800000508626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,4,power_law_1.2,0.21279358863830566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.2,1.119315242767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,4,power_law_1.2,1.4167167663574218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,2,balanced,0.8427519798278809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,2,balanced,0.15201600392659506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,4,power_law_1.2,0.22812800407409667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.2,1.620582389831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,4,power_law_1.2,1.9449472427368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,2,balanced,0.8482346534729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,2,balanced,0.25127466519673664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,4,power_law_1.2,0.2380671977996826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.2,2.0688575744628905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,4,power_law_1.2,2.8027711868286134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,2,balanced,0.851194699605306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,2,balanced,0.4510026772816976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,4,power_law_1.2,0.2811392068862915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,2,balanced,0.854037364323934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,4,power_law_1.2,2.9768512725830076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,4,power_law_1.2,2.456608009338379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,2,balanced,0.4535413185755412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,2,balanced,0.8625760078430176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,4,power_law_1.2,0.2717695951461792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,4,power_law_1.2,3.854777526855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,2,balanced,0.45184532801310223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,4,power_law_1.2,0.2853247880935669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,4,power_law_1.2,4.800307083129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,2,balanced,0.8706346352895101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,4,power_law_1.2,7.4194496154785154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,2,balanced,0.45422931512196857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,4,power_law_1.2,0.5528192043304443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,2,balanced,0.8754186630249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,4,power_law_1.2,9.877900695800781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,2,balanced,0.4583359956741333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.2,0.6473152160644531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,2,balanced,0.8915039698282877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,2,balanced,0.46270934740702313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.2,0.46309762001037597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,2,balanced,0.9177119731903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,2,balanced,0.4654080073038737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.2,0.4840703964233398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,2,balanced,0.9452640215555826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,2,balanced,0.4704586664835612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.2,0.5339136123657227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,2,balanced,0.988912026087443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,2,balanced,0.4792213439941406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.2,0.6374847888946533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,2,balanced,1.0365546544392903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.2,0.7624127864837646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,2,balanced,0.4901440143585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.2,0.961849594116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,2,balanced,1.1184639930725098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,2,balanced,0.4952746629714966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.2,1.335654354095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,2,balanced,0.062309334675470986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,2,balanced,0.5082773367563883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,2,balanced,1.2019306818644206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,2,balanced,0.03555733213822047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.2,1.5709952354431151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,2,balanced,0.06819200019041698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,2,balanced,0.5334133307139078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,2,balanced,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,4,power_law_1.2,1.942790412902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,2,balanced,0.09075199564297994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,2,balanced,1.5150879224141438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,2,balanced,0.5584693352381388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,4,power_law_1.2,2.88470401763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,2,balanced,0.058373332023620605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,2,balanced,0.1283093293507894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,2,balanced,0.6083893378575643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,2,balanced,0.08675733208656311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,2,balanced,1.5808053016662598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,2,balanced,0.19487466414769491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,4,power_law_1.2,6.001580810546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,2,balanced,0.14289066195487976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,2,balanced,0.67194135983785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,2,balanced,0.3193546732266744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,2,balanced,0.24712000290552774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,2,balanced,0.322927991549174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,2,balanced,0.25356799364089966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,2,balanced,0.7720533212025961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,2,balanced,2.4557973543802896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,2,balanced,0.3290133277575175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,2,balanced,0.25933865706125897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,2,balanced,0.3280693292617798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,2,balanced,0.8761866887410482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,2,balanced,0.26715733607610065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,2,balanced,0.32898133993148804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,2,balanced,0.275221327940623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,2,balanced,0.33981867631276447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,2,balanced,0.28885867198308307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,2,balanced,1.17029865582784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,2,balanced,2.5631093978881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,2,balanced,0.3383893171946208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,2,balanced,0.27398399511973065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,2,balanced,0.3398933410644531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,2,balanced,0.2829066713651021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,2,balanced,0.3496319850285848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,2,balanced,1.3300480047861736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,2,balanced,0.29582399129867554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,2,balanced,0.35388267040252686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,2,balanced,0.2986133297284444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,2,balanced,0.3633493185043335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,2,balanced,3.7171414693196616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,2,balanced,0.30197866757710773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,2,balanced,0.3744800090789795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,2,balanced,1.924938678741455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,2,balanced,0.3125866651535034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,2,balanced,0.4011146624883016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,2,balanced,0.39773333072662354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,2,balanced,0.42962666352589923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,2,balanced,0.388592004776001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,2,balanced,0.48413864771525067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,2,balanced,0.7316213448842367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,2,balanced,2.3732800483703613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,2,balanced,0.5176106691360474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,2,balanced,0.8024266560872396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,2,balanced,0.6155466636021932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,2,balanced,4.846197446187337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,2,balanced,0.6380746761957804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,2,balanced,0.7116959889729818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,2,balanced,0.6776586373647054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,2,balanced,3.0332746505737305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,2,balanced,0.9969813028971354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,2,balanced,0.7588906288146973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,2,balanced,1.192735989888509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,2,balanced,0.8305013179779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,2,balanced,1.6970879236857097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,2,balanced,1.3956106503804524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,2,balanced,4.548255920410156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,2,balanced,2.154202620188395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,2,balanced,10.441642761230469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,2,balanced,1.5281920433044434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,2,balanced,2.759653409322103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,2,balanced,2.0888586044311523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,2,balanced,8.705925623575846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,2,balanced,4.090053240458171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,2,balanced,2.962165196736654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,2,balanced,7.929050445556641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,2,balanced,5.8528798421223955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,2,power_law_1.01,0.09278720021247863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,2,power_law_1.01,0.14188799858093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,2,power_law_1.01,0.11302399635314941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,2,power_law_1.01,0.2097599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,2,power_law_1.01,0.3159231901168823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,2,power_law_1.01,0.4015103816986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,2,power_law_1.01,0.5808832168579101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,2,power_law_1.01,0.6266111850738525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,2,power_law_1.01,0.628652811050415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,2,power_law_1.01,0.6533184051513672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,2,power_law_1.01,0.6902400016784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,2,power_law_1.01,0.6900991916656494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,2,power_law_1.01,0.7136384010314941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,2,power_law_1.01,0.7431039810180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,2,power_law_1.01,0.7661952018737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,2,power_law_1.01,0.7767615795135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,2,power_law_1.01,0.8164799690246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,2,power_law_1.01,0.888588809967041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,2,power_law_1.01,0.9133440017700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,2,power_law_1.01,1.0048768043518066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,2,power_law_1.01,1.049894428253174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,2,power_law_1.01,1.214451217651367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,2,power_law_1.01,1.3746944427490235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,2,power_law_1.01,1.7034751892089843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,2,power_law_1.01,1.991801643371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,2,power_law_1.01,2.558527946472168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,2,power_law_1.01,3.3250686645507814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,2,power_law_1.01,3.8233985900878906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,2,power_law_1.01,0.0686847984790802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,2,power_law_1.01,0.06680960059165955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,2,power_law_1.01,6.096300888061523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,2,power_law_1.01,0.09473919868469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,2,power_law_1.01,0.08819199800491333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,2,power_law_1.01,0.08869119882583618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,2,power_law_1.01,11.01304931640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,2,power_law_1.01,0.08123520016670227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,2,power_law_1.01,0.11998080015182495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,2,power_law_1.01,0.11313920021057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,2,power_law_1.01,0.18217600584030152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,2,power_law_1.01,0.1527232050895691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,2,power_law_1.01,0.2338047981262207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,2,power_law_1.01,0.1801151990890503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,2,power_law_1.01,0.3262784004211426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,2,power_law_1.01,0.2375744104385376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,2,power_law_1.01,0.3400896072387695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,2,power_law_1.01,0.2531840085983276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,2,power_law_1.01,0.35598719120025635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,2,power_law_1.01,0.2646912097930908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,2,power_law_1.01,0.3723520040512085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,2,power_law_1.01,0.2774912118911743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,2,power_law_1.01,0.3850048065185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,2,power_law_1.01,0.28696959018707274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,2,power_law_1.01,0.39387519359588624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,2,power_law_1.01,0.29361279010772706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,2,power_law_1.01,0.4033088207244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,2,power_law_1.01,0.29747838973999025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,2,power_law_1.01,0.02839680016040802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,2,power_law_1.01,0.4242239952087402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,2,power_law_1.01,0.2973376035690308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,2,power_law_1.01,0.05070080161094666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,2,power_law_1.01,0.43746562004089357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,2,power_law_1.01,0.3186624050140381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,2,power_law_1.01,0.04561919867992401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,2,power_law_1.01,0.4450047969818115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,2,power_law_1.01,0.33011839389801023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,2,power_law_1.01,0.07122560143470764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,2,power_law_1.01,0.4655168056488037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,2,power_law_1.01,0.3519488096237183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,2,power_law_1.01,0.10493439435958862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,2,power_law_1.01,0.5283135890960693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,2,power_law_1.01,0.391430401802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,2,power_law_1.01,0.12933119535446166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,2,power_law_1.01,0.5422272205352783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.01,0.43860478401184083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,2,power_law_1.01,0.1775488018989563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,2,power_law_1.01,0.6324351787567138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.01,0.5070655822753907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,2,power_law_1.01,0.19329919815063476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,2,power_law_1.01,0.6917439937591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.01,0.5474112033843994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,2,power_law_1.01,0.19516160488128662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,2,power_law_1.01,0.8199423789978028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.01,0.6827648162841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,2,power_law_1.01,0.2033535957336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,2,power_law_1.01,0.9452992439270019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.01,0.8051072120666504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,2,power_law_1.01,0.21515519618988038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,2,power_law_1.01,1.248863983154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.01,1.0529919624328614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,2,power_law_1.01,0.2128767967224121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,2,power_law_1.01,1.5635071754455567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.01,1.366105556488037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,2,power_law_1.01,0.2256256103515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,2,power_law_1.01,2.0782848358154298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.01,1.843724822998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,2,power_law_1.01,0.23157119750976562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,2,power_law_1.01,2.7895744323730467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,2,power_law_1.01,0.23213438987731932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.01,2.3821760177612306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,2,power_law_1.01,3.2022785186767577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,2,power_law_1.01,0.24072959423065185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,2,power_law_1.01,3.0250368118286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,2,power_law_1.01,5.0083263397216795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,2,power_law_1.01,0.2515455961227417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,2,power_law_1.01,4.401657485961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,2,power_law_1.01,0.26863999366760255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,2,power_law_1.01,9.503616333007812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.01,0.28383359909057615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,2,power_law_1.01,8.552569580078124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.01,0.6358848094940186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.01,0.7185279846191406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.01,0.4824831962585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.01,0.5626944065093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.01,0.6500095844268798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.01,0.7828224182128907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.01,1.0668031692504882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.01,1.3259712219238282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,2,power_law_1.01,1.5952128410339355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,2,power_law_1.01,2.422220802307129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,2,power_law_1.01,4.555500793457031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,2,power_law_1.2,0.09383040070533752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,2,power_law_1.2,0.13827840089797974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,2,power_law_1.2,0.09863680005073547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,2,power_law_1.2,0.19225599765777587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,2,power_law_1.2,0.2688704013824463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,2,power_law_1.2,0.35834240913391113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,2,power_law_1.2,0.562879991531372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,2,power_law_1.2,0.6011903762817383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,2,power_law_1.2,0.6260928153991699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,2,power_law_1.2,0.6378623962402343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,2,power_law_1.2,0.6851136207580566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,2,power_law_1.2,0.7091328144073487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,2,power_law_1.2,0.06852480173110961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,2,power_law_1.2,0.7180992126464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,2,power_law_1.2,0.7542975902557373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,2,power_law_1.2,0.09305599927902222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,2,power_law_1.2,0.7785471916198731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,2,power_law_1.2,0.08047360181808472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,2,power_law_1.2,0.7865344047546386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,2,power_law_1.2,0.11783679723739623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,2,power_law_1.2,0.8707263946533204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,2,power_law_1.2,0.17103359699249268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,2,power_law_1.2,0.9492992401123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,2,power_law_1.2,0.21723520755767822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,2,power_law_1.2,0.9451711654663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,2,power_law_1.2,0.3164416074752808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,2,power_law_1.2,1.061075210571289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,2,power_law_1.2,0.33139839172363283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,2,power_law_1.2,1.0965375900268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,2,power_law_1.2,0.3542975902557373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,2,power_law_1.2,1.2522751808166503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,2,power_law_1.2,0.35932800769805906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,2,power_law_1.2,1.4311103820800781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,2,power_law_1.2,0.37689599990844724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,2,power_law_1.2,1.814246368408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,2,power_law_1.2,0.38835198879241944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,2,power_law_1.2,2.1253631591796873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,2,power_law_1.2,0.4005887985229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,2,power_law_1.2,2.8481536865234376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,2,power_law_1.2,0.06687999963760376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,2,power_law_1.2,0.4216320037841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,2,power_law_1.2,3.5019454956054688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,2,power_law_1.2,0.08701440095901489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,2,power_law_1.2,0.4497663974761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,2,power_law_1.2,4.042854309082031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,2,power_law_1.2,0.07948799729347229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,2,power_law_1.2,0.4534783840179443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,2,power_law_1.2,6.162675094604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,2,power_law_1.2,0.11094399690628051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,2,power_law_1.2,0.4860799789428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,2,power_law_1.2,0.028595200181007384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,2,power_law_1.2,0.1394495964050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,2,power_law_1.2,11.775462341308593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,2,power_law_1.2,0.5456768035888672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,2,power_law_1.2,0.050374400615692136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,2,power_law_1.2,0.17176320552825927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,2,power_law_1.2,0.5751808166503907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,2,power_law_1.2,0.04596480131149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,2,power_law_1.2,0.23413119316101075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,2,power_law_1.2,0.6705023765563964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,2,power_law_1.2,0.0663424015045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,2,power_law_1.2,0.2533567905426025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,2,power_law_1.2,0.7251327991485595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,2,power_law_1.2,0.09415040016174317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,2,power_law_1.2,0.26428799629211425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,2,power_law_1.2,0.8713088035583496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,2,power_law_1.2,0.11832959651947021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,2,power_law_1.2,0.2690047979354858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,2,power_law_1.2,1.0168255805969237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,2,power_law_1.2,0.178438401222229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,2,power_law_1.2,0.2836927890777588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,2,power_law_1.2,1.345088005065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,2,power_law_1.2,0.1859392046928406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,2,power_law_1.2,0.2858175992965698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,2,power_law_1.2,1.6498367309570312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,2,power_law_1.2,0.19533439874649047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,2,power_law_1.2,0.298304009437561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,2,power_law_1.2,2.237676811218262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,2,power_law_1.2,0.195360004901886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,2,power_law_1.2,0.3135231971740723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,2,power_law_1.2,2.9633600234985353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,2,power_law_1.2,0.21966080665588378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,2,power_law_1.2,0.33660159111022947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,2,power_law_1.2,3.499123382568359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,2,power_law_1.2,0.21023359298706054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,2,power_law_1.2,0.3433536052703857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,2,power_law_1.2,5.238636779785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,2,power_law_1.2,0.22126080989837646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,2,power_law_1.2,0.3605504035949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,2,power_law_1.2,9.315507507324218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,2,power_law_1.2,0.23620479106903075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,2,power_law_1.2,0.40814719200134275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,2,power_law_1.2,0.24122240543365478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.2,0.4546175956726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,2,power_law_1.2,0.24360959529876708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.2,0.5548031806945801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,2,power_law_1.2,0.26485118865966795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.2,0.5588543891906739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,2,power_law_1.2,0.2799232006072998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.2,0.6923903942108154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.2,0.29900801181793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.2,0.8543744087219238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.2,0.6872447967529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.2,1.1355008125305175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.2,0.7453440189361572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.2,1.3706111907958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.2,0.5134399890899658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.2,2.050227165222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.2,0.5648191928863525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.2,2.5498687744140627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.2,0.7212416172027588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,2,power_law_1.2,3.034611129760742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.2,0.8399552345275879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,2,power_law_1.2,4.810291290283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.2,1.113043212890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.2,1.390988826751709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,2,power_law_1.2,9.234841918945312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,2,power_law_1.2,1.5867839813232423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,2,power_law_1.2,2.5481599807739257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,2,power_law_1.2,4.915052795410157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,1,balanced,0.07028799752394359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,1,balanced,0.09529067079226176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,1,balanced,0.14730667074521384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,1,balanced,0.24632000923156738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,1,balanced,0.4464319944381714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,1,balanced,0.8419520060221354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,1,balanced,0.8530453046162924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,1,balanced,0.8440852959950765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,1,balanced,0.8504532972971598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,1,balanced,0.8521386782328287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,1,balanced,0.8572586377461752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,1,balanced,0.8674560387929281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,1,balanced,0.8708906968434652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,1,balanced,0.8854453563690186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,1,balanced,0.8938079675038656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,1,balanced,0.8983786900838217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,1,balanced,0.92795197168986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,1,balanced,0.9686559836069742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,1,balanced,1.0163733164469402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,1,balanced,1.0992586612701416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,1,balanced,1.160373369852702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,1,balanced,1.3137760162353516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,1,balanced,1.4657707214355469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,1,balanced,1.8828907012939453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,1,balanced,0.05611733098824819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,1,balanced,2.155285358428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,1,balanced,0.07926400005817413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,1,balanced,0.07446933289368947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,1,balanced,0.10301333665847778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,1,balanced,0.08834667007128398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,1,balanced,0.15742400288581848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,1,balanced,0.11321600278218587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,1,balanced,0.2673119902610779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,1,balanced,3.121232032775879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,1,balanced,0.16116266449292502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,1,balanced,0.46061865488688153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,1,balanced,0.24733332792917886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,1,balanced,0.4692853291829427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,1,balanced,0.40883731842041016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,1,balanced,0.4709920088450114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,1,balanced,0.41514134407043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,1,balanced,0.48204267024993896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,1,balanced,3.8109814325968423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,1,balanced,0.4142773151397705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,1,balanced,0.47858667373657227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,1,balanced,0.4167199929555257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,1,balanced,0.02334933231274287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,1,balanced,0.48898132642110187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,1,balanced,0.4174933433532715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,1,balanced,0.025648000339667004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,1,balanced,0.4944640000661214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,1,balanced,0.42563732465108234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,1,balanced,0.041296000281969704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,1,balanced,0.4280426502227783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,1,balanced,0.5018293460210165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,1,balanced,0.05624000231424967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,1,balanced,4.815061251322429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,1,balanced,0.43592000007629395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,1,balanced,0.5084853172302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,1,balanced,0.08704533179601033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,1,balanced,0.44300798575083417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,1,balanced,0.5172373453776041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,1,balanced,0.14877333243687949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,1,balanced,0.4519786834716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,1,balanced,0.5312266747156779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,1,balanced,0.15278933445612589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,1,balanced,0.4590826829274495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,1,balanced,0.5492159922917684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,1,balanced,0.15793599685033163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,1,balanced,0.47815465927124023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,1,balanced,0.5933119853337606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,1,balanced,0.16306133071581522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,1,balanced,0.5188373327255249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,1,balanced,0.6243199904759725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,1,balanced,0.16687466700871786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,1,balanced,0.5539199908574423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,1,balanced,0.7106506824493408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,1,balanced,0.1753386656443278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,1,balanced,0.6318560043970743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,1,balanced,7.0688050587972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,1,balanced,0.7949333190917969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,1,balanced,0.18357867002487183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,1,balanced,0.721402645111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,1,balanced,0.19239999850591025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,1,balanced,1.032101313273112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,1,balanced,0.2105706731478373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,1,balanced,0.891157309214274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,1,balanced,0.19483200709025064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,1,balanced,1.1612640221913655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,1,balanced,0.20404267311096191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,1,balanced,1.0594773292541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,1,balanced,0.22495466470718384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,1,balanced,1.5916479428609211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,1,balanced,0.22247999906539917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,1,balanced,1.5454400380452473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,1,balanced,0.24672534068425497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,1,balanced,1.915119965871175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,1,balanced,0.25543999671936035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,1,power_law_1.01,0.06899840235710145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,1,balanced,1.8517279624938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,1,balanced,0.2791786591211955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,1,power_law_1.01,0.09367039799690247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,1,power_law_1.01,0.10858240127563476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,1,balanced,0.5575893322626749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,1,balanced,2.798528035481771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,1,power_law_1.01,0.19811199903488158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,1,balanced,2.751162528991699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,1,balanced,0.6355786720911661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,1,power_law_1.01,0.3033663988113403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,1,balanced,13.600964864095053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,1,power_law_1.01,0.4118015766143799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,1,balanced,0.5235840082168579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,1,power_law_1.01,0.5801280021667481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,1,balanced,3.5325921376546225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,1,power_law_1.01,0.6365183830261231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,1,balanced,3.5176000595092773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,1,power_law_1.01,0.6369215965270996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,1,balanced,0.5781279802322388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,1,power_law_1.01,0.6716928005218505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,1,power_law_1.01,0.6857279777526856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,1,power_law_1.01,0.7131199836730957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,1,power_law_1.01,0.7185984134674073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,1,balanced,4.3917280832926435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,1,power_law_1.01,0.05114240050315857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,1,balanced,0.950272003809611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,1,power_law_1.01,0.7537280082702636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,1,power_law_1.01,0.07197440266609192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,1,balanced,4.434624036153157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,1,power_law_1.01,0.7846335887908935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,1,power_law_1.01,0.0775551974773407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,1,power_law_1.01,0.8072447776794434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,1,power_law_1.01,0.13203200101852416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,1,power_law_1.01,0.8409343719482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,1,power_law_1.01,0.18524160385131835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,1,power_law_1.01,0.9399744033813476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,1,power_law_1.01,0.2418816089630127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,1,balanced,1.0864160060882568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,1,power_law_1.01,0.9791935920715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,1,power_law_1.01,0.3272128105163574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,1,power_law_1.01,1.1118911743164062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,1,power_law_1.01,0.3541440010070801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,1,power_law_1.01,1.1722240447998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,1,power_law_1.01,0.37556478977203367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,1,power_law_1.01,1.3804736137390137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,1,power_law_1.01,0.37331199645996094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,1,balanced,6.657040278116862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,1,power_law_1.01,1.5986432075500487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,1,power_law_1.01,0.40464000701904296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,1,power_law_1.01,2.027801513671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,1,power_law_1.01,0.41283202171325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,1,balanced,6.74897575378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,1,balanced,1.4479039510091145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,1,power_law_1.01,2.422105598449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,1,power_law_1.01,0.4164224147796631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,1,power_law_1.01,3.2565567016601564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,1,power_law_1.01,0.4377471923828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,1,power_law_1.01,4.122598266601562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,1,power_law_1.01,0.45530238151550295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,1,power_law_1.01,0.07369599938392639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,1,power_law_1.01,4.988716888427734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,1,power_law_1.01,0.4703680038452148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,1,power_law_1.01,0.08313599824905396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,1,power_law_1.01,0.508358383178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,1,power_law_1.01,7.617574310302734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,1,power_law_1.01,0.09477760195732117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,1,power_law_1.01,0.5699520111083984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,1,power_law_1.01,0.14111360311508178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,1,power_law_1.01,0.6218815803527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,1,power_law_1.01,14.089862060546874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,1,power_law_1.01,0.022188800573349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,1,power_law_1.01,0.186080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,1,power_law_1.01,0.7421567916870118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,1,balanced,2.0965654055277505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,1,power_law_1.01,0.23370881080627443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,1,power_law_1.01,0.027980801463127137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,1,power_law_1.01,0.8315008163452149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,1,power_law_1.01,0.3100352048873901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,1,power_law_1.01,0.033958399295806886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,1,power_law_1.01,1.0460224151611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,1,power_law_1.01,0.33203198909759524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,1,balanced,12.790426890055338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,1,power_law_1.01,0.05103999972343445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,1,power_law_1.01,1.2628416061401366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,1,power_law_1.01,0.33378560543060304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,1,power_law_1.01,0.08634240031242371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,1,balanced,12.853691101074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,1,power_law_1.01,1.6502656936645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,1,power_law_1.01,0.34851200580596925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,1,power_law_1.01,0.12383999824523925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,1,power_law_1.01,2.049728012084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,1,power_law_1.01,0.3619136095046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,1,power_law_1.01,0.15892479419708253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,1,power_law_1.01,2.823276710510254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,1,power_law_1.01,0.3702143907546997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,1,power_law_1.01,0.17896319627761842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,1,power_law_1.01,3.617510223388672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,1,power_law_1.01,0.3802815914154053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,1,power_law_1.01,0.18110079765319825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,1,power_law_1.01,0.3917376041412354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,1,power_law_1.01,4.407142257690429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,1,power_law_1.01,0.1847551941871643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,1,power_law_1.01,0.40110077857971194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,1,power_law_1.01,6.700543975830078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,1,power_law_1.01,0.19267200231552123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,1,power_law_1.01,0.4202239990234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,1,power_law_1.01,12.85546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,1,power_law_1.01,0.20714879035949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,1,power_law_1.01,0.44890241622924804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,1,balanced,4.122991879781087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,1,power_law_1.01,0.2215104103088379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,1,power_law_1.01,0.513267183303833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,1,power_law_1.01,0.23695359230041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.01,0.5540800094604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,1,power_law_1.01,0.23042559623718262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.01,0.6699391841888428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.01,0.7235455989837647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,1,power_law_1.01,0.24511361122131348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.01,0.9366847991943359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,1,power_law_1.01,0.26489601135253904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.01,1.1308927536010742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,1,power_law_1.01,0.26506240367889405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.01,1.528940773010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.01,0.2943295955657959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.01,1.959596824645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.01,0.3108223915100098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.01,2.7427391052246093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.01,0.3501055955886841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.01,3.619027328491211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.01,0.7638463973999023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,1,power_law_1.01,4.375347137451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.01,0.8661503791809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,1,power_law_1.01,6.650956726074218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.01,0.6649792194366455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,1,power_law_1.01,12.842265319824218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.01,0.7848896026611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.01,1.0225664138793946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.01,1.2862719535827636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,1,power_law_1.01,1.5217344284057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,1,power_law_1.01,2.2725759506225587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,1,power_law_1.01,4.378726577758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,1,power_law_1.2,0.06822400093078614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,1,power_law_1.2,0.09356799721717834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,1,power_law_1.2,0.11224960088729859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,1,power_law_1.2,0.17840640544891356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,1,power_law_1.2,0.257804799079895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,1,power_law_1.2,0.36394240856170657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,1,power_law_1.2,0.5608960151672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,1,power_law_1.2,0.6077760219573974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,1,power_law_1.2,0.6322303771972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,1,power_law_1.2,0.6534592151641846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,1,power_law_1.2,0.6784768104553223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,1,power_law_1.2,0.720198392868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,1,power_law_1.2,0.7216832160949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,1,power_law_1.2,0.7603392124176025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,1,power_law_1.2,0.8140159606933594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,1,power_law_1.2,0.8405055999755859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,1,power_law_1.2,0.8878975868225097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,1,power_law_1.2,0.9859904289245606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,1,power_law_1.2,1.0308032035827637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,1,power_law_1.2,1.1684800148010255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,1,power_law_1.2,1.2105600357055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,1,power_law_1.2,1.413651180267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,1,power_law_1.2,1.6739904403686523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,1,power_law_1.2,2.069593620300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,1,power_law_1.2,2.4854015350341796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,1,power_law_1.2,3.340435028076172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,1,power_law_1.2,4.1976318359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,1,power_law_1.2,5.050252914428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,1,power_law_1.2,7.721228790283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,1,power_law_1.2,14.14586181640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,1,power_law_1.2,0.052147197723388675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,1,power_law_1.2,0.07194240093231201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,1,power_law_1.2,0.08361600041389465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,1,power_law_1.2,0.12428799867630005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,1,power_law_1.2,0.169868803024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,1,power_law_1.2,0.22122879028320314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,1,power_law_1.2,0.07336959838867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,1,power_law_1.2,0.3192960023880005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,1,power_law_1.2,0.08340479731559754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,1,power_law_1.2,0.3484479904174805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,1,power_law_1.2,0.09575039744377137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,1,power_law_1.2,0.3637248039245605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,1,power_law_1.2,0.13303040266036986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,1,power_law_1.2,0.022732800245285033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,1,power_law_1.2,0.37553279399871825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,1,power_law_1.2,0.16561919450759888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,1,power_law_1.2,0.40236802101135255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,1,power_law_1.2,0.02813439965248108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,1,power_law_1.2,0.2162303924560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,1,power_law_1.2,0.4238272190093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,1,power_law_1.2,0.033055999875068666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,1,power_law_1.2,0.29736320972442626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,1,power_law_1.2,0.4311359882354736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,1,power_law_1.2,0.05384320020675659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,1,power_law_1.2,0.31863040924072267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,1,power_law_1.2,0.4456768035888672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,1,power_law_1.2,0.08353279829025269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,1,power_law_1.2,0.3374527931213379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,1,power_law_1.2,0.4716159820556641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,1,power_law_1.2,0.11418240070343018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,1,power_law_1.2,0.33025920391082764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,1,power_law_1.2,0.4791168212890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,1,power_law_1.2,0.149017596244812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,1,power_law_1.2,0.36520960330963137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,1,power_law_1.2,0.5352511882781983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,1,power_law_1.2,0.1764224052429199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,1,power_law_1.2,0.36856319904327395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,1,power_law_1.2,0.603763198852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,1,power_law_1.2,0.17858560085296632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,1,power_law_1.2,0.3799360036849976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,1,power_law_1.2,0.6437312126159668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,1,power_law_1.2,0.1882367968559265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,1,power_law_1.2,0.39523839950561523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,1,power_law_1.2,0.7546112060546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,1,power_law_1.2,0.20492160320281982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,1,power_law_1.2,0.4214144229888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,4,balanced,0.02975466599067052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,4,balanced,0.03373866776625315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,1,power_law_1.2,0.8640000343322753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,1,power_law_1.2,0.21308159828186035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,1,power_law_1.2,0.4311679840087891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,4,balanced,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,1,power_law_1.2,0.2214143991470337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,1,power_law_1.2,1.0820223808288574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,4,balanced,0.031290667752424874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,1,power_law_1.2,0.45116801261901857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,4,balanced,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,1,power_law_1.2,0.2448064088821411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,4,balanced,0.032431999842325844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,1,power_law_1.2,1.2873855590820313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,1,power_law_1.2,0.5207808017730713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,4,balanced,0.038704000413417816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,4,balanced,0.05353599786758423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,1,power_law_1.2,0.23978879451751708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,1,power_law_1.2,1.671379280090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.2,0.567526388168335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,4,balanced,0.05795733133951823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,4,balanced,0.07859200239181519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,1,power_law_1.2,0.25160961151123046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,1,power_law_1.2,2.0745471954345702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.2,0.6822720050811768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,4,balanced,0.06003733476003011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,1,power_law_1.2,0.2779328107833862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,4,balanced,0.07840000092983246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.2,0.7532544136047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,1,power_law_1.2,2.8450368881225585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,1,power_law_1.2,0.277894401550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,4,balanced,0.06033066908518473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,4,balanced,0.07910400132338206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.2,0.9500736236572266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,1,power_law_1.2,3.657958221435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.2,0.29728639125823975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,4,balanced,0.0606826643149058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,4,balanced,0.0784800002972285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.2,1.1358336448669433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.2,0.330348801612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,1,power_law_1.2,4.442963027954102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,4,balanced,0.059845333298047386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,4,balanced,0.07913599908351898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.2,1.5295680046081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.2,0.3677248001098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,4,balanced,0.06065600117047628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,1,power_law_1.2,6.7122047424316404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,4,balanced,0.07766399780909221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,4,balanced,0.061050668358802795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.2,1.9799488067626954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.2,0.7922175884246826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,4,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,4,balanced,0.080485333998998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,4,balanced,0.06434666613737743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,1,power_law_1.2,12.831283569335938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,4,balanced,0.06402133405208588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.2,2.739795112609863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.2,0.9411328315734864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,4,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,4,balanced,0.08038400113582611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.2,3.6523326873779296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,4,balanced,0.06542400022347768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.2,0.7015552043914794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,1,power_law_1.2,4.4084735870361325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,4,balanced,0.0461706668138504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,4,balanced,0.08322133123874664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,4,balanced,0.07283733288447063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.2,0.7969151973724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,1,power_law_1.2,6.680364990234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,4,balanced,0.08268266419569652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,4,balanced,0.07295999924341838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,4,balanced,0.057999998331069946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.2,1.0408448219299316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,4,balanced,0.08694932858149211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,4,balanced,0.07634133100509644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,1,power_law_1.2,12.760038757324219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.2,1.3126079559326171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,4,balanced,0.0766293356815974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,4,balanced,0.08830933769543965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,4,balanced,0.08218666911125183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,4,balanced,0.01937599976857503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,1,power_law_1.2,1.537388801574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,4,balanced,0.07055999835332234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,4,balanced,0.09650133053461711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,4,balanced,0.08799999952316284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,4,balanced,0.021301334102948506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,4,balanced,0.07257066667079926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,1,power_law_1.2,2.31345272064209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,4,balanced,0.09568533301353455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,4,balanced,0.10560533404350281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,4,balanced,0.021573332448800404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,4,balanced,0.07096533477306366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,1,power_law_1.2,4.443123245239258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,4,balanced,0.02348800003528595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,4,balanced,0.10475200414657593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,4,balanced,0.11845333377520244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,4,balanced,0.07039999961853027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,4,balanced,0.028922667105992634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,4,balanced,0.11694399515787761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,4,balanced,0.07301866511503856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,4,balanced,0.14646933476130167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,4,balanced,0.031328000128269196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,4,balanced,0.11688533425331116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,4,balanced,0.0710453341404597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,4,balanced,0.03161599983771642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,4,balanced,0.17730667193730673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,4,balanced,0.033413333197434746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,4,balanced,0.07468266785144806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,4,balanced,0.14081066846847534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,4,balanced,0.03386666625738144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,4,balanced,0.07327466706434886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,4,balanced,0.24399999777475992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,4,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,4,balanced,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,4,balanced,0.15267200271288553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,4,balanced,0.035973332822322845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,4,balanced,0.08076799909273784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,4,balanced,0.03751466671625773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,4,balanced,0.07889066636562347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,4,balanced,0.2940746744473775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,4,balanced,0.03842666745185852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,4,balanced,0.2143253286679586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,4,balanced,0.08205333352088928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,4,balanced,0.04090133309364319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,4,balanced,0.08691199620564778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,4,balanced,0.042410666743914284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,4,balanced,0.09321600198745728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,4,balanced,0.04423466821511587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,4,balanced,0.4227253198623657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,4,balanced,0.25062400102615356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,4,balanced,0.10192533334096272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,4,balanced,0.05277866621812185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,4,balanced,0.055871998270352684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,4,balanced,0.11602133512496948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,4,balanced,0.08784000078837077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,4,balanced,0.10790933171908061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,4,balanced,0.3394560019175212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,4,balanced,0.15785599748293558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,4,balanced,0.5495626529057821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,4,balanced,0.0999840001265208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,4,balanced,0.18190399805704752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,4,balanced,0.1074133316675822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,4,balanced,0.25011199712753296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,4,balanced,0.42467200756073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,4,balanced,0.11828266580899556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,4,balanced,0.6771093209584554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,4,balanced,0.31803200642267865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,4,balanced,0.1883359948794047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,4,balanced,0.20629332462946573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,4,balanced,0.5151946544647217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,4,balanced,0.45140798886617023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,4,balanced,0.29315199454625446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,4,balanced,1.036847988764445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,4,balanced,0.5743306477864584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,4,balanced,0.3811359802881877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,4,balanced,0.7870879968007406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,4,balanced,0.6950186888376871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,4,balanced,0.46991999944051105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,4,balanced,1.061797300974528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,4,balanced,2.0115787188212075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,4,balanced,0.7359413305918375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,4,balanced,1.4872852961222331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,4,power_law_1.01,0.036057600378990175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,4,power_law_1.01,0.032358399033546446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,4,power_law_1.01,0.03450239896774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,4,power_law_1.01,0.0419840008020401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,4,power_law_1.01,0.04775039851665497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,4,power_law_1.01,0.06929919719696045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,4,power_law_1.01,0.07342720031738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,4,power_law_1.01,0.07179520130157471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,4,power_law_1.01,0.07253119945526124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,4,power_law_1.01,0.07426559925079346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,4,power_law_1.01,0.0744383990764618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,4,balanced,2.030453364054362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,4,power_law_1.01,0.07626240253448487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,4,power_law_1.01,0.07880319952964783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,4,power_law_1.01,0.08065919876098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,4,power_law_1.01,0.08665599822998046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,4,power_law_1.01,0.08837760090827942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,4,balanced,1.4524693489074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,4,power_law_1.01,0.0921280026435852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,4,power_law_1.01,0.10841599702835084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,4,power_law_1.01,0.11861120462417603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,4,power_law_1.01,0.13576960563659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,4,power_law_1.01,0.14128639698028564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,4,power_law_1.01,0.16283520460128784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,4,power_law_1.01,0.20503039360046388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,4,power_law_1.01,0.26368639469146726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,4,power_law_1.01,0.3164671897888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,4,power_law_1.01,0.40855040550231936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,4,power_law_1.01,0.5234240055084228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,4,power_law_1.01,0.6725247859954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,4,power_law_1.01,1.0479488372802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,4,power_law_1.01,1.8300031661987304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,4,power_law_1.01,0.03546879887580871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,4,power_law_1.01,0.031251201033592226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,4,power_law_1.01,0.032332798838615416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,4,power_law_1.01,0.04575999975204468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,4,power_law_1.01,0.038022398948669434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,4,power_law_1.01,0.04017280042171478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,4,power_law_1.01,0.04035199880599975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,4,power_law_1.01,0.04434559941291809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,4,power_law_1.01,0.05255039930343628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,4,power_law_1.01,0.05278080105781555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,4,power_law_1.01,0.05601279735565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,4,power_law_1.01,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,4,power_law_1.01,0.05671679973602295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,4,power_law_1.01,0.06986879706382751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,4,power_law_1.01,0.05748479962348938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,4,power_law_1.01,0.06991360187530518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,4,power_law_1.01,0.05777279734611511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,4,power_law_1.01,0.0694271981716156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,4,power_law_1.01,0.06083199977874756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,4,power_law_1.01,0.07108479738235474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,4,power_law_1.01,0.06299520134925843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,4,power_law_1.01,0.07200639843940734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,4,power_law_1.01,0.01825920045375824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,4,power_law_1.01,0.06452479958534241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,4,power_law_1.01,0.07307519912719726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,4,power_law_1.01,0.06925439834594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,4,power_law_1.01,0.018483200669288637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,4,power_law_1.01,0.07512959837913513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,4,power_law_1.01,0.07336320281028748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,4,power_law_1.01,0.019526399672031403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,4,power_law_1.01,0.07467520236968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,4,power_law_1.01,0.07572479844093323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,4,power_law_1.01,0.022118400037288665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,4,power_law_1.01,0.07667199969291687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,4,power_law_1.01,0.07909759879112244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,4,power_law_1.01,0.02333440035581589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,4,power_law_1.01,0.08062080144882203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,4,power_law_1.01,0.0870464026927948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,4,power_law_1.01,0.026118400692939758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,4,power_law_1.01,0.0833728015422821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,4,power_law_1.01,0.10751359462738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,4,power_law_1.01,0.026796799898147584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,4,power_law_1.01,0.08822399973869324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,4,power_law_1.01,0.12319359779357911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,4,power_law_1.01,0.028377598524093627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,4,power_law_1.01,0.09757440090179444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,4,power_law_1.01,0.13226239681243895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,4,power_law_1.01,0.030195200443267824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,4,power_law_1.01,0.11015679836273193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,4,power_law_1.01,0.17704960107803344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,4,power_law_1.01,0.030943998694419862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,4,power_law_1.01,0.12956160306930542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,4,power_law_1.01,0.21402881145477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,4,power_law_1.01,0.03075839877128601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,4,power_law_1.01,0.14689279794692994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,4,power_law_1.01,0.27569921016693116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,4,power_law_1.01,0.0326335996389389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,4,power_law_1.01,0.19359359741210938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,4,power_law_1.01,0.3681983947753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,4,power_law_1.01,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,4,power_law_1.01,0.22316160202026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,4,power_law_1.01,0.5656383991241455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,4,power_law_1.01,0.03356800079345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,4,power_law_1.01,0.28694400787353513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,4,power_law_1.01,0.6773695945739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,4,power_law_1.01,0.036185601353645326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,4,power_law_1.01,0.4028223991394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,4,power_law_1.01,0.8609536170959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,4,power_law_1.01,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,4,power_law_1.01,0.5710783958435058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,4,power_law_1.01,0.041407999396324155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,4,power_law_1.01,1.3390591621398926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,4,power_law_1.01,0.6781824111938477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,4,power_law_1.01,0.04796159863471985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,4,power_law_1.01,2.5383424758911133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,4,power_law_1.01,0.07429760098457336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,4,power_law_1.01,0.8432319641113282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,4,power_law_1.01,0.0949184000492096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,4,power_law_1.01,1.2487296104431151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,4,power_law_1.01,0.07546240091323853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,4,power_law_1.01,0.09157760143280029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,4,power_law_1.01,2.6995264053344727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,4,power_law_1.01,0.10804480314254761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,4,power_law_1.01,0.13924479484558105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,4,power_law_1.01,0.16914559602737428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,4,power_law_1.01,0.235315203666687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,4,power_law_1.01,0.31007359027862547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,4,power_law_1.01,0.38253440856933596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,4,power_law_1.01,0.5622975826263428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,4,power_law_1.01,1.1245439529418946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,4,power_law_1.2,0.036671999096870425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,4,power_law_1.2,0.03108479976654053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,4,power_law_1.2,0.033766400814056394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,4,power_law_1.2,0.04072319865226746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,4,power_law_1.2,0.04619520008563995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,4,power_law_1.2,0.03470079898834229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,4,power_law_1.2,0.07148159742355346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,4,power_law_1.2,0.07233920097351074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,4,power_law_1.2,0.02956160008907318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,4,power_law_1.2,0.07427840232849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,4,power_law_1.2,0.032902398705482484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,4,power_law_1.2,0.07514879703521729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,4,power_law_1.2,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,4,power_law_1.2,0.07706239819526672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,4,power_law_1.2,0.0385343998670578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,4,power_law_1.2,0.07194240093231201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,4,power_law_1.2,0.053439998626708986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,4,power_law_1.2,0.08083840012550354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,4,power_law_1.2,0.056569600105285646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,4,power_law_1.2,0.08058879971504211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,4,power_law_1.2,0.056492799520492555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,4,power_law_1.2,0.07943040132522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,4,power_law_1.2,0.057715201377868654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,4,power_law_1.2,0.08647040128707886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,4,power_law_1.2,0.05826560258865356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,4,power_law_1.2,0.046105599403381346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,4,power_law_1.2,0.0893887996673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,4,power_law_1.2,0.05950719714164734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,4,power_law_1.2,0.09980159997940063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,4,power_law_1.2,0.06290559768676758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,4,power_law_1.2,0.03953279852867127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,4,power_law_1.2,0.10922880172729492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,4,power_law_1.2,0.06640639901161194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,4,power_law_1.2,0.04465279877185822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,4,power_law_1.2,0.12121599912643433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,4,power_law_1.2,0.06835839748382569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,4,power_law_1.2,0.05089920163154602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,4,power_law_1.2,0.13701119422912597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,4,power_law_1.2,0.07379199862480164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,4,power_law_1.2,0.05607039928436279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,4,power_law_1.2,0.13938560485839843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,4,power_law_1.2,0.07589759826660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,4,power_law_1.2,0.018783999979496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,4,power_law_1.2,0.07100800275802613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,4,power_law_1.2,0.18892799615859984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,4,power_law_1.2,0.07996159791946411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,4,power_law_1.2,0.07356799840927124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,4,power_law_1.2,0.018367999792099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,4,power_law_1.2,0.08895999789237977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,4,power_law_1.2,0.21854081153869628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,4,power_law_1.2,0.07049599885940552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,4,power_law_1.2,0.10788480043411255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,4,power_law_1.2,0.020768000185489653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,4,power_law_1.2,0.2742784023284912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,4,power_law_1.2,0.07327359914779663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,4,power_law_1.2,0.12384639978408814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,4,power_law_1.2,0.021510399878025055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,4,power_law_1.2,0.3566015958786011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,4,power_law_1.2,0.07019519805908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,4,power_law_1.2,0.13215359449386596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,4,power_law_1.2,0.023500800132751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,4,power_law_1.2,0.4803008079528809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,4,power_law_1.2,0.0735040009021759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,4,power_law_1.2,0.17825280427932738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,4,power_law_1.2,0.026796799898147584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,4,power_law_1.2,0.7174399852752685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,4,power_law_1.2,0.07783039808273315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,4,power_law_1.2,0.2473599910736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,4,power_law_1.2,0.02895359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,4,power_law_1.2,0.9028672218322754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,4,power_law_1.2,0.07672960162162781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,4,power_law_1.2,0.3386176109313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,4,power_law_1.2,0.029292801022529603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,4,power_law_1.2,1.2399871826171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,4,power_law_1.2,0.0776639997959137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,4,power_law_1.2,0.37813119888305663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,4,power_law_1.2,0.03173120021820068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,4,power_law_1.2,0.0832319974899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,4,power_law_1.2,2.60830078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,4,power_law_1.2,0.7037248134613037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,4,power_law_1.2,0.03141759932041168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,4,power_law_1.2,0.08336640000343323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,4,power_law_1.2,0.7737664222717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,2,balanced,0.02941333254178365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,4,power_law_1.2,0.0336896002292633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,4,power_law_1.2,0.09088640213012696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,4,power_law_1.2,1.0886528015136718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,2,balanced,0.031328000128269196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,4,power_law_1.2,0.03432320058345795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,4,power_law_1.2,0.09998080134391785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,2,balanced,0.03533866753180822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,4,power_law_1.2,1.7983423233032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,4,power_law_1.2,0.035283198952674864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,4,power_law_1.2,0.10948480367660522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,2,balanced,0.05097599824269613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,4,power_law_1.2,0.03463039994239807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,4,power_law_1.2,3.2963905334472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,4,power_law_1.2,0.13942400217056275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,2,balanced,0.07964266836643219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,4,power_law_1.2,0.03694080114364624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,4,power_law_1.2,0.1532863974571228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,2,balanced,0.08002666632334392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,4,power_law_1.2,0.039417600631713866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,4,power_law_1.2,0.19288320541381837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,2,balanced,0.0800799975792567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,4,power_law_1.2,0.04230400025844574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,4,power_law_1.2,0.2320319890975952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,2,balanced,0.08323200047016144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,4,power_law_1.2,0.049132800102233885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,2,balanced,0.0828959991534551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,4,power_law_1.2,0.304800009727478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,4,power_law_1.2,0.0788096010684967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,2,balanced,0.08222400148709615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,4,power_law_1.2,0.10008319616317748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,4,power_law_1.2,0.41805438995361327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,2,balanced,0.0827893316745758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,4,power_law_1.2,0.07993599772453308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,2,balanced,0.08549867073694865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,4,power_law_1.2,0.5899007797241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,4,power_law_1.2,0.09651200175285339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,2,balanced,0.08675199747085571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,4,power_law_1.2,0.10855040550231934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,4,power_law_1.2,0.8204992294311524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,2,balanced,0.08867200215657552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,4,power_law_1.2,0.14981119632720946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,2,balanced,0.09295466542243958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,4,power_law_1.2,1.0983296394348145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,4,power_law_1.2,0.18180480003356933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,2,balanced,0.09480533003807068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,4,power_law_1.2,0.2520576000213623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,4,power_law_1.2,1.573964786529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,2,balanced,0.09947733084360759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,4,power_law_1.2,0.32974720001220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,2,balanced,0.10776533683141072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,4,power_law_1.2,3.9186817169189454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,4,power_law_1.2,0.3768512010574341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,2,balanced,0.11623467008272807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,4,power_law_1.2,0.6550655841827393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,2,balanced,0.13538133104642233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,4,power_law_1.2,1.282367992401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,2,balanced,0.14286399881045023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,2,balanced,0.18262932697931925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,2,balanced,0.21688000361124674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,2,balanced,0.2978559931119283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,2,balanced,0.3614293336868286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,2,balanced,0.5069493452707926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,2,balanced,0.658517320950826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,2,balanced,0.8058559894561768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,2,balanced,0.043920000394185386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,2,balanced,0.03374933451414108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,2,balanced,0.04762133459250132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,2,balanced,0.0540533314148585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,2,balanced,0.03194666653871536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,2,balanced,0.0681279997030894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,2,balanced,0.03578133384386698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,2,balanced,0.04172799984614054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,2,balanced,0.08940266569455464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,2,balanced,0.06065600117047628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,2,balanced,0.08598400155703227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,2,balanced,0.06258666515350342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,2,balanced,0.08825066685676575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,2,balanced,0.064410666624705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,2,balanced,0.017103999853134155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,2,balanced,1.2470346291859944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,2,balanced,0.08344533046086629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,2,balanced,0.06424533327420552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,2,balanced,0.020101333657900494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,2,balanced,0.0849173367023468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,2,balanced,0.06599999964237213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,2,balanced,0.020986666282018025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,2,balanced,0.08658132950464885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,2,balanced,0.06723199784755707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,2,balanced,0.02346666653951009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,2,balanced,0.0881813367207845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,2,balanced,0.06720533470312755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,2,balanced,0.029285334050655365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,2,balanced,0.09098133444786072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,2,balanced,0.07038400073846181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,2,balanced,0.03125333289305369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,2,balanced,0.08915733297665913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,2,balanced,0.07044800122578938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,2,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,2,balanced,0.09327999750773112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,2,balanced,0.0749066670735677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,2,balanced,0.031370667119820915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,2,balanced,0.10030933221181233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,2,balanced,0.08221866687138875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,2,balanced,0.03209066639343897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,2,balanced,0.09922666351000468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,2,balanced,0.08295999964078267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,2,balanced,0.03332266708215078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,2,balanced,0.10500799616177876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,2,balanced,0.08624533812204997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,2,balanced,0.03575466573238373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,2,balanced,0.1183519959449768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,2,balanced,0.0974720021088918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,2,balanced,0.035349334279696144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,2,balanced,0.13191999991734824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,2,balanced,0.10910933216412862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,2,balanced,0.03604800005753835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,2,balanced,0.15125333269437155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,2,balanced,0.1313706636428833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,2,balanced,0.039594667653242745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,2,balanced,0.042037333051363625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,2,balanced,0.1763520042101542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,2,balanced,0.14917332927385965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,2,balanced,0.04370133578777313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,2,balanced,0.0521066685517629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,2,balanced,0.2498826583226522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,2,balanced,0.19723733266194662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,2,balanced,0.05389333268006643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,2,balanced,0.055919999877611794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,2,balanced,0.3001226584116618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,2,balanced,0.23428267240524292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,2,balanced,0.07915733257929485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,2,balanced,0.10483200351397197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,2,balanced,2.43721596399943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,2,balanced,0.4280639886856079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,2,balanced,0.343178669611613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,2,balanced,0.10498666763305664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,2,balanced,0.5564800103505453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,2,balanced,0.11321600278218587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,2,balanced,0.4485066731770833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,2,balanced,0.1843306620915731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,2,balanced,0.7980693181355795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,2,balanced,0.657696008682251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,2,balanced,0.19479467471440634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,2,balanced,1.0347466468811035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,2,balanced,0.273690660794576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,2,balanced,0.8668639659881592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,2,balanced,0.35594133536020917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,2,balanced,1.2762133280436199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,2,balanced,1.0681493282318115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,2,power_law_1.01,0.03386879861354828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,2,power_law_1.01,0.031200000643730165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,2,power_law_1.01,0.03496319949626923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,2,balanced,0.43669335047403973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,2,power_law_1.01,0.04297600090503693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,2,power_law_1.01,0.049439999461174014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,2,power_law_1.01,0.0332863986492157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,2,power_law_1.01,0.0704576015472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,2,power_law_1.01,0.02990719974040985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,2,power_law_1.01,0.07194240093231201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,2,balanced,1.970586617787679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,2,power_law_1.01,0.03384959995746613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,2,power_law_1.01,0.07800959944725036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,2,power_law_1.01,0.03917439877986908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,2,balanced,1.6755199432373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,2,power_law_1.01,0.07943040132522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,2,power_law_1.01,0.04325119853019714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,2,power_law_1.01,0.08128640055656433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,2,power_law_1.01,0.05537279844284058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,2,power_law_1.01,0.08018559813499451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,2,power_law_1.01,0.05826560258865356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,2,power_law_1.01,0.08472319841384887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,2,power_law_1.01,0.0593280017375946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,2,power_law_1.01,0.08632320165634155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,2,balanced,0.6801866690317789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,2,power_law_1.01,0.06117759943008423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,2,power_law_1.01,0.08944000005722046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,2,power_law_1.01,0.06178560256958008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,2,power_law_1.01,0.09311360120773315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,2,power_law_1.01,0.06537600159645081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,2,power_law_1.01,0.09559040069580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,2,power_law_1.01,0.06982399821281433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,2,power_law_1.01,0.1
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,2,power_law_1.01,0.0722432017326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,2,power_law_1.01,0.10654079914093018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,2,power_law_1.01,0.07368959784507752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,2,power_law_1.01,0.12309759855270386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,2,power_law_1.01,0.08135679960250855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,2,power_law_1.01,0.14364800453186036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,2,power_law_1.01,0.08580480217933655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,2,power_law_1.01,0.04720639884471893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,2,power_law_1.01,0.15556479692459108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,2,power_law_1.01,0.09009919762611389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,2,power_law_1.01,0.045286399126052854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,2,power_law_1.01,0.20124800205230714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,2,power_law_1.01,0.10003199577331542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,2,power_law_1.01,0.050297600030899045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,2,power_law_1.01,0.2500351905822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,2,power_law_1.01,0.11360000371932984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,2,power_law_1.01,0.058483201265335086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,2,power_law_1.01,0.32810239791870116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,2,power_law_1.01,0.13814400434494017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,2,power_law_1.01,0.06451839804649354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,2,balanced,3.8206027348836265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,2,power_law_1.01,0.16675200462341308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,2,power_law_1.01,0.402950382232666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,2,power_law_1.01,0.08126720190048217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,2,power_law_1.01,0.21422719955444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,2,power_law_1.01,0.5685760021209717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,2,balanced,3.1999947230021157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,2,power_law_1.01,0.08313599824905396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,2,power_law_1.01,0.2693376064300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,2,power_law_1.01,0.7106368064880371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,2,power_law_1.01,0.08261759877204895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,2,power_law_1.01,0.38860158920288085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,2,power_law_1.01,0.9124863624572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,2,power_law_1.01,0.08621439933776856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,2,power_law_1.01,0.016595199704170227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,2,power_law_1.01,0.5021823883056641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,2,power_law_1.01,0.08538240194320679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,2,power_law_1.01,1.3419648170471192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,2,balanced,1.3348852793375652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,2,power_law_1.01,0.01738879978656769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,2,power_law_1.01,0.7079999923706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,2,power_law_1.01,0.0880128026008606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,2,power_law_1.01,2.7440511703491213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,2,power_law_1.01,0.018700799345970152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,2,power_law_1.01,0.9645695686340332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,2,power_law_1.01,0.08999040126800537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,2,power_law_1.01,0.02162559926509857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,2,power_law_1.01,1.0992256164550782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,2,power_law_1.01,0.09226239919662475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,2,power_law_1.01,0.022342400252819063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,2,power_law_1.01,1.8052288055419923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,2,power_law_1.01,0.09438080191612244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,2,power_law_1.01,0.025830399990081788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,2,power_law_1.01,0.10004479885101318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,2,power_law_1.01,3.681350326538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,2,power_law_1.01,0.02693760097026825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,2,power_law_1.01,0.10426880121231079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,2,power_law_1.01,0.02873600125312805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,2,power_law_1.01,0.11297279596328735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,2,power_law_1.01,0.030739200115203858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,2,power_law_1.01,0.12199039459228515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,2,power_law_1.01,0.03192960023880005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,2,power_law_1.01,0.14225280284881592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,2,power_law_1.01,0.030963200330734252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,2,power_law_1.01,0.17458560466766357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,2,power_law_1.01,0.03320319950580597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,2,power_law_1.01,0.20675199031829833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,2,power_law_1.01,0.03418239951133728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,2,power_law_1.01,0.26822400093078613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,2,power_law_1.01,0.03656960129737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,2,power_law_1.01,0.32159359455108644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,2,power_law_1.01,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,2,power_law_1.01,0.4758080005645752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,2,power_law_1.01,0.04066559970378876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,2,power_law_1.01,0.6176640033721924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,2,power_law_1.01,0.04333440065383911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,2,power_law_1.01,0.8743743896484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,2,power_law_1.01,0.048198398947715757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,2,power_law_1.01,1.1336576461791992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,2,power_law_1.01,0.054476797580718994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,2,power_law_1.01,1.4399935722351074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,2,power_law_1.01,0.06613759994506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,2,power_law_1.01,2.083513641357422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,2,power_law_1.01,0.08257279992103576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,2,power_law_1.01,0.09655680060386658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,2,power_law_1.01,4.024345779418946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,2,power_law_1.01,0.111244797706604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,2,power_law_1.01,0.13814400434494017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,2,power_law_1.01,0.1726591944694519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,2,power_law_1.01,0.23548800945281984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,2,power_law_1.01,0.31244161128997805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,2,power_law_1.01,0.37149438858032224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,2,power_law_1.01,0.5797887802124023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,2,power_law_1.01,1.148921585083008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,2,power_law_1.2,0.03335039913654327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,2,power_law_1.2,0.030163198709487915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,2,power_law_1.2,0.03434880077838898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,2,power_law_1.2,0.041766399145126344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,2,power_law_1.2,0.04615040123462677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,2,power_law_1.2,0.07052159905433655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,2,power_law_1.2,0.07385600209236146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,2,power_law_1.2,0.07995520234107971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,2,power_law_1.2,0.07982720136642456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,2,power_law_1.2,0.08121600151062011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,2,power_law_1.2,0.08277119994163513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,2,power_law_1.2,0.08481280207633972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,2,power_law_1.2,0.08511360287666321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,2,power_law_1.2,0.0901311993598938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,2,power_law_1.2,0.0939136028289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,2,power_law_1.2,0.09900799989700318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,2,power_law_1.2,0.10014079809188843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,2,power_law_1.2,0.11143679618835449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,2,power_law_1.2,0.12721279859542847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,2,power_law_1.2,0.15754239559173583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,2,power_law_1.2,0.16104960441589355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,2,power_law_1.2,0.2016832113265991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,2,power_law_1.2,0.26257278919219973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,2,power_law_1.2,0.34369280338287356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,2,power_law_1.2,0.4479680061340332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,2,power_law_1.2,0.6237311840057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,2,power_law_1.2,0.7714752197265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,2,power_law_1.2,0.9777471542358398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,2,power_law_1.2,1.5750847816467286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,2,power_law_1.2,3.38671989440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,2,power_law_1.2,0.0477183997631073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,2,power_law_1.2,0.033190399408340454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,2,power_law_1.2,0.04470399916172028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,2,power_law_1.2,0.030105599761009218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,2,power_law_1.2,0.04888319969177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,2,power_law_1.2,0.03367680013179779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,2,power_law_1.2,0.05617920160293579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,2,power_law_1.2,0.0392767995595932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,2,power_law_1.2,0.061510401964187625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,2,power_law_1.2,0.016595199704170227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,2,power_law_1.2,0.04035840034484863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,2,power_law_1.2,0.0813759982585907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,2,power_law_1.2,0.05761920213699341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,2,power_law_1.2,0.017235200107097625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,2,power_law_1.2,0.08371840119361877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,2,power_law_1.2,0.05811200141906738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,2,power_law_1.2,0.01812479943037033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,2,power_law_1.2,0.08446080088615418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,1,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,2,power_law_1.2,0.060736000537872314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,2,power_law_1.2,0.020473599433898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,2,power_law_1.2,0.08647680282592773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,2,power_law_1.2,0.06304640173912049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,1,balanced,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,2,power_law_1.2,0.02255360037088394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,2,power_law_1.2,0.08552320003509521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,2,power_law_1.2,0.06335359811782837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,1,balanced,0.03893866638342539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,2,power_law_1.2,0.08792960047721862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,2,power_law_1.2,0.025574401021003723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,2,power_law_1.2,0.06636800169944763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,1,balanced,0.05585066477457682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,2,power_law_1.2,0.0913919985294342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,2,power_law_1.2,0.02655999958515167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,2,power_law_1.2,0.07175040245056152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,2,power_law_1.2,0.09400320053100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,1,balanced,0.08756267031033833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,2,power_law_1.2,0.027923199534416198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,2,power_law_1.2,0.07366399765014649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,1,balanced,0.08699733018875122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,2,power_law_1.2,0.09671040177345276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,2,power_law_1.2,0.030630400776863097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,1,balanced,0.037690666814645134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,2,power_law_1.2,0.07620480060577392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,1,balanced,0.0886346697807312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,2,power_law_1.2,0.10247679948806762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,2,power_law_1.2,0.03193599879741669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,2,power_law_1.2,0.0826367974281311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,1,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,2,power_law_1.2,0.10762879848480225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,1,balanced,0.038202665746212006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,1,balanced,0.0922986666361491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,2,power_law_1.2,0.031174400448799135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,2,power_law_1.2,0.08789119720458985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,1,balanced,0.047882666190465294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,2,power_law_1.2,0.11383680105209351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,1,balanced,0.0921493371327718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,2,power_law_1.2,0.03187839984893799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,2,power_law_1.2,0.08955519795417785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,1,balanced,0.06817600131034851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,2,power_law_1.2,0.1338368058204651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,1,balanced,0.09237866600354512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,2,power_law_1.2,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,1,balanced,0.06945066650708516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,2,power_law_1.2,0.10168319940567017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,1,balanced,0.09338133533795674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,2,power_law_1.2,0.14466559886932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,2,power_law_1.2,0.03622399866580963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,2,power_law_1.2,0.1149183988571167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,1,balanced,0.07227199772993724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,1,balanced,0.09499733646710713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,2,power_law_1.2,0.1801151990890503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,2,power_law_1.2,0.15125759840011596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,1,balanced,0.07233599821726482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,2,power_law_1.2,0.038771200180053714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,1,balanced,0.09711999694506328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,2,power_law_1.2,0.2128511905670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,1,balanced,0.07381866872310638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,2,power_law_1.2,0.16293120384216309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,2,power_law_1.2,0.04038400053977966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,1,balanced,0.1048906644185384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,2,power_law_1.2,0.27163519859313967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,1,balanced,0.07461333274841309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,2,power_law_1.2,0.22588160037994384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,2,power_law_1.2,0.04330880045890808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,1,balanced,0.10873599847157796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,2,power_law_1.2,0.330297589302063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,1,balanced,0.07870933413505554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,2,power_law_1.2,0.28837759494781495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,2,power_law_1.2,0.04795520007610321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,1,balanced,0.11286399761835735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,1,balanced,0.081386665503184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,2,power_law_1.2,0.4933119773864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,2,power_law_1.2,0.39584639072418215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,2,power_law_1.2,0.6472000122070313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,2,power_law_1.2,0.053651201725006106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,1,balanced,0.11880000432332356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,1,balanced,0.08310399949550629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,2,power_law_1.2,0.524403190612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,2,power_law_1.2,0.9230912208557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,2,power_law_1.2,0.06721919775009155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,1,power_law_1.01,0.03076480031013489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,1,balanced,0.1341653366883596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,1,balanced,0.08885866403579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,2,power_law_1.2,0.8008959770202637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,2,power_law_1.2,1.2678463935852051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,1,power_law_1.01,0.032601600885391234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,2,power_law_1.2,0.08221439719200134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,2,power_law_1.2,1.0316543579101562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,1,balanced,0.13588266571362814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,1,balanced,0.09613866607348125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,2,power_law_1.2,1.5762240409851074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,1,power_law_1.01,0.03697920143604279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,2,power_law_1.2,0.1009600043296814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,1,balanced,0.1760800083478292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,2,power_law_1.2,1.4062848091125488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,1,balanced,0.09852799773216248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,2,power_law_1.2,2.2873855590820313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,1,power_law_1.01,0.04547199904918671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,2,power_law_1.2,0.11448320150375366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,1,balanced,0.1902880072593689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,2,power_law_1.2,2.0645055770874023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,1,balanced,0.10634666681289673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,2,power_law_1.2,4.675315093994141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,1,power_law_1.01,0.05119360089302063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,2,power_law_1.2,0.14647680521011353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,1,balanced,0.1237333317597707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,2,power_law_1.2,4.200761413574218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,1,power_law_1.01,0.07656959891319275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,1,balanced,0.2729439934094747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,2,power_law_1.2,0.17537280321121215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,1,balanced,0.1476693352063497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,1,power_law_1.01,0.07868800163269044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,1,power_law_1.01,0.034483200311660765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,2,power_law_1.2,0.24437758922576905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,1,balanced,0.18519999583562216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,1,power_law_1.01,0.08354560136795045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,1,balanced,0.32337067524592084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,2,power_law_1.2,0.3175872087478638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,1,power_law_1.01,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,1,power_law_1.01,0.08620160222053527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,1,balanced,0.21029333273569742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,2,power_law_1.2,0.37947518825531007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,1,power_law_1.01,0.03692159950733185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,1,power_law_1.01,0.08798080086708068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,1,balanced,0.451749324798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,1,balanced,0.31860800584157306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,1,power_law_1.01,0.04336000084877014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,2,power_law_1.2,0.5981311798095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,1,power_law_1.01,0.09158400297164918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,1,power_law_1.01,0.04971520006656647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,1,balanced,0.38757868607838947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,1,power_law_1.01,0.0936896026134491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,2,power_law_1.2,1.0918720245361329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,1,power_law_1.01,0.06232320070266724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,1,balanced,0.5746773481369019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,1,power_law_1.01,0.09560959935188293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,1,power_law_1.01,0.06551679968833923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,1,balanced,0.5790239969889323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,1,power_law_1.01,0.10139520168304443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,1,power_law_1.01,0.06836479902267456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,1,power_law_1.01,0.10617599487304688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,1,power_law_1.01,0.0707264006137848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,1,power_law_1.01,0.10949120521545411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,1,power_law_1.01,0.0722432017326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,1,balanced,0.8382613658905029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,1,balanced,0.7525333563486735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,1,power_law_1.01,0.07888640165328979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,1,power_law_1.01,0.12328319549560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,1,power_law_1.01,0.08183680176734924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,1,power_law_1.01,0.142739200592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,1,power_law_1.01,0.0857151985168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,1,power_law_1.01,0.15426559448242189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,1,power_law_1.01,0.19016319513320923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,1,power_law_1.01,0.08748800158500672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,1,balanced,1.1172266801198323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,1,power_law_1.01,0.20703999996185302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,1,power_law_1.01,0.09438719749450683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,1,balanced,1.080906629562378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,1,power_law_1.01,0.09895039796829223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,1,power_law_1.01,0.2751039981842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,1,power_law_1.01,0.10780800580978393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,1,power_law_1.01,0.3398207902908325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,1,power_law_1.01,0.129203200340271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,1,power_law_1.01,0.47064957618713377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,1,power_law_1.01,0.15256320238113402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,1,power_law_1.01,0.5981440067291259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,1,balanced,1.4690027236938477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,1,power_law_1.01,0.19553920030593872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,1,power_law_1.01,0.8487296104431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,1,power_law_1.01,0.23200640678405762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,1,power_law_1.01,1.0976832389831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,1,balanced,1.349573294321696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,1,power_law_1.01,0.32154240608215334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,1,power_law_1.01,1.350233554840088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,1,power_law_1.01,0.4073215961456299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,1,power_law_1.01,2.0988224029541014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,1,power_law_1.01,0.5912000179290772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,1,power_law_1.01,0.7604351997375488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,1,power_law_1.01,4.053657531738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,1,balanced,1.825717290242513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,1,power_law_1.01,1.1252736091613769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,1,power_law_1.01,1.4673664093017578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,1,power_law_1.01,1.8137344360351562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,1,power_law_1.01,2.8697023391723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,1,power_law_1.01,5.559782409667969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,1,balanced,2.12389866511027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,1,balanced,2.900218645731608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,1,power_law_1.2,0.029363200068473816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,1,power_law_1.2,0.03249920010566711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,1,power_law_1.2,0.03616639971733093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,1,power_law_1.2,0.04162560105323791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,1,balanced,4.138959884643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,1,power_law_1.2,0.047635200619697574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,1,power_law_1.2,0.07519360184669495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,1,power_law_1.2,0.08116480112075805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,1,power_law_1.2,0.085343998670578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,1,balanced,5.644655863444011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,1,power_law_1.2,0.0893119990825653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,1,power_law_1.2,0.08915200233459472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,1,power_law_1.2,0.0920960009098053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,1,power_law_1.2,0.09512959718704224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,1,power_law_1.2,0.09637759923934937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,1,power_law_1.2,0.10097279548645019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,1,power_law_1.2,0.1078976035118103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,1,power_law_1.2,0.10994559526443481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,1,power_law_1.2,0.12380800247192383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,1,power_law_1.2,0.14138879776000976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,1,power_law_1.2,0.15208959579467773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,1,power_law_1.2,0.19093120098114014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,1,power_law_1.2,0.20923519134521484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,1,power_law_1.2,0.2760960102081299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,1,power_law_1.2,0.33912320137023927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,1,power_law_1.2,0.4716479778289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,1,power_law_1.2,0.5969151973724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,1,power_law_1.2,0.8503168106079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,1,power_law_1.2,1.1056575775146484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,1,power_law_1.2,1.3498240470886231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,1,power_law_1.2,2.097644805908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,1,power_law_1.2,4.038086318969727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,8,balanced,0.08515200018882751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,8,balanced,0.08787199854850769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,8,balanced,0.064410666624705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,8,balanced,0.085807998975118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,8,balanced,0.08638933300971985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,1,power_law_1.2,0.034790399670600894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,8,balanced,0.08795733253161113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,8,balanced,0.06438933312892914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,8,balanced,0.08695466319719951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,8,balanced,0.06358933448791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,8,balanced,0.08718933661778767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,8,balanced,0.06832533578077953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,1,power_law_1.2,0.033004799485206605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,8,balanced,0.08869333068529765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,8,balanced,0.05470933516820272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,8,balanced,0.06651733318964641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,1,power_law_1.2,0.03700479865074158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,8,balanced,0.09149866302808125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,8,balanced,0.06785066425800323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,8,balanced,0.057189335425694786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,8,balanced,0.09200533231099446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,1,power_law_1.2,0.04237439930438995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,8,balanced,0.06859200199445088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,8,balanced,0.09312533338864644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,8,balanced,0.05619733532269796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,8,balanced,0.03363200028737386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,1,power_law_1.2,0.04795520007610321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,8,balanced,0.06913066903750102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,8,balanced,0.10048000017801921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,8,balanced,0.05789333085219065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,8,balanced,0.09956799944241841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,8,balanced,0.05601066847642263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,8,balanced,0.033589333295822144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,8,balanced,0.06854933500289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,1,power_law_1.2,0.06327679753303528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,8,balanced,0.10229333241780598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,8,balanced,0.056832000613212585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,8,balanced,0.03297066688537598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,8,balanced,0.07032000025113423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,8,balanced,0.10771733522415161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,1,power_law_1.2,0.06796159744262695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,8,balanced,0.07233599821726482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,8,balanced,0.10943466424942017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,8,balanced,0.05985066791375478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,8,balanced,0.032602667808532715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,1,power_law_1.2,0.06895999908447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,8,balanced,0.10917866230010986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,8,balanced,0.07206400235493977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,8,balanced,0.058186665177345276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,8,balanced,0.1381600002447764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,8,balanced,0.03368533402681351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,8,balanced,0.07188266515731812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,8,balanced,0.05886933207511902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,1,power_law_1.2,0.07185279726982116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,8,balanced,0.1320799986521403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,8,balanced,0.07311466832955678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,8,balanced,0.05922133227189382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,8,balanced,0.03872533390919367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,8,balanced,0.08282133440176646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,1,power_law_1.2,0.07203840017318726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,8,balanced,0.166703999042511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,8,balanced,0.060378665725390114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,8,balanced,0.08264533181985219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,8,balanced,0.06915733218193054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,1,power_law_1.2,0.0788927972316742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,8,balanced,0.06224533418814341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,8,balanced,0.17645333210627237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,8,balanced,0.08732266227404277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,8,balanced,0.06849599877993266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,8,balanced,0.25116799275080365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,8,balanced,0.09795733292897542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,8,balanced,0.0748586654663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,1,power_law_1.2,0.08218240141868591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,8,balanced,0.0602400004863739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,8,balanced,0.10122133294741313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,8,balanced,0.06840533514817555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,8,balanced,0.28565333286921185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,1,power_law_1.2,0.08646399974822998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,8,balanced,0.0680159976085027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,8,balanced,0.06002666552861532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,8,balanced,0.12546133001645407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,1,power_law_1.2,0.08917120099067688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,8,balanced,0.3951146602630615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,8,balanced,0.0965226689974467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,8,balanced,0.07454933226108551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,1,power_law_1.2,0.0961471974849701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,8,balanced,0.08865066369374593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,8,balanced,0.14321066935857138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,8,balanced,0.07388799885908763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,8,balanced,0.4769279956817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,1,power_law_1.2,0.10101759433746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,8,balanced,0.07445333401362102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,8,balanced,0.1917653282483419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,8,balanced,0.10753066341082256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,1,power_law_1.2,0.1092479944229126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,8,balanced,0.07555200159549713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,8,balanced,0.6789706548055013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,1,power_law_1.2,0.1304128050804138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,8,balanced,0.2270080049832662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,8,balanced,0.0849173367023468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,8,balanced,0.08130133152008057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,1,power_law_1.2,0.1545024037361145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,8,balanced,0.09874133268992107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,8,balanced,0.0939359962940216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,8,balanced,0.30590933561325073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,8,balanced,0.14172266920407614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,1,power_law_1.2,0.19685759544372558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,8,balanced,0.8808426856994629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,8,balanced,0.10314133763313293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,8,balanced,0.16511999567349753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,8,balanced,0.37867732842763263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,8,balanced,0.23267199595769247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,8,balanced,0.11017066240310669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,1,power_law_1.2,0.2335808038711548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,8,balanced,0.29480000336964923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,8,balanced,0.13801067074139914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,1,power_law_1.2,0.3228032112121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,8,balanced,1.089408000310262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,8,balanced,0.5419466495513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,8,balanced,0.4222559928894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,8,balanced,0.16478932897249857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,1,power_law_1.2,0.40950398445129393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,8,balanced,0.532480001449585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,8,balanced,0.6985812981923422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,1,power_law_1.2,0.5869376182556152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,8,balanced,0.212719996770223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,8,balanced,0.7848107020060221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,1,power_law_1.2,0.7627264022827148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,8,balanced,1.6978079477945964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,8,balanced,0.26814399162928265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,1,power_law_1.2,1.119257640838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,8,balanced,1.0238880316416423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,8,balanced,0.8576320012410482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,1,power_law_1.2,1.4623488426208495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,8,power_law_1.01,0.08314239978790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,8,balanced,0.3643999894460042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,8,balanced,1.5166826248168945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,1,power_law_1.2,1.8025152206420898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,8,power_law_1.01,0.08752639889717102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,8,power_law_1.01,0.08463360071182251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,1,power_law_1.2,2.869977569580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,8,balanced,0.46218665440877277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,8,power_law_1.01,0.08495360016822814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,8,balanced,1.3311999638875325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,8,balanced,2.008517265319824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,8,power_law_1.01,0.08554239869117737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,1,power_law_1.2,5.549894332885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,8,power_law_1.01,0.08904320001602173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,8,power_law_1.01,0.10168960094451904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,8,balanced,3.3725706736246743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,8,power_law_1.01,0.1235200047492981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,8,balanced,0.5841439962387085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,8,power_law_1.01,0.14083839654922486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,8,balanced,2.495530605316162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,8,power_law_1.01,0.14823039770126342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,8,power_law_1.01,0.13671040534973145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,8,power_law_1.01,0.06206079721450806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,8,power_law_1.01,0.18356479406356813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,8,power_law_1.01,0.18748799562454224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,8,power_law_1.01,0.06269440054893494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,8,balanced,0.8791893323262533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,8,power_law_1.01,0.06279680132865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,8,power_law_1.01,0.16202239990234374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,8,power_law_1.01,0.06618239879608154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,8,balanced,2.5836426417032876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,8,balanced,3.9502506256103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,8,power_law_1.01,0.20295040607452391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,8,power_law_1.01,0.06580479741096497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,8,power_law_1.01,0.06707839965820313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,8,power_law_1.01,0.2180351972579956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,8,power_law_1.01,0.06968960165977478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,8,power_law_1.01,0.1896064043045044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,8,power_law_1.01,0.06951040029525757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,8,power_law_1.01,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,8,power_law_1.01,0.07396479845046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,8,power_law_1.01,0.05475199818611145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,8,power_law_1.01,0.25440640449523927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,8,power_law_1.01,0.05264000296592712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,8,power_law_1.01,0.07491199970245362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,8,power_law_1.01,0.05457280278205871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,8,power_law_1.01,0.30328960418701173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,8,power_law_1.01,0.07567359805107117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,8,power_law_1.01,0.05464959740638733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,8,power_law_1.01,0.44625282287597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,8,power_law_1.01,0.07900800108909607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,8,balanced,1.765226682027181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,8,power_law_1.01,0.05614079833030701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,8,power_law_1.01,0.05591679811477661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,8,power_law_1.01,0.46102399826049806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,8,power_law_1.01,0.08911359906196595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,8,power_law_1.01,0.05795199871063232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,8,power_law_1.01,0.09281920194625855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,8,power_law_1.01,0.06295040249824524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,8,power_law_1.01,0.6914175987243653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,8,balanced,7.845696131388347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,8,power_law_1.01,0.06414080262184144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,8,power_law_1.01,0.1118016004562378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,8,power_law_1.01,0.8674304008483886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,8,power_law_1.01,0.06706560254096985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,8,power_law_1.01,0.11701120138168335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,8,power_law_1.01,1.299846363067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,8,power_law_1.01,0.06895999908447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,8,power_law_1.01,0.07197440266609192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,8,power_law_1.01,0.12365440130233765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,8,power_law_1.01,1.6077888488769532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,8,power_law_1.01,0.07638400197029113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,8,power_law_1.01,0.07946239709854126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,8,power_law_1.01,0.15805439949035643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,8,power_law_1.01,0.0930176019668579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,8,power_law_1.01,2.314361572265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,8,power_law_1.01,0.10005760192871094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,8,power_law_1.01,0.19912320375442505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,8,power_law_1.01,0.1137727975845337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,8,power_law_1.01,3.5167102813720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,8,power_law_1.01,0.2489856004714966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.01,0.1368384003639221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.01,0.16996480226516725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,8,power_law_1.01,0.29712638854980467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.01,0.21943678855895996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,8,power_law_1.01,0.42691841125488283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,8,power_law_1.01,4.648051071166992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.01,0.316102409362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,8,power_law_1.01,0.5652224063873291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.01,0.3502847909927368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,8,power_law_1.01,0.8179712295532227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,8,power_law_1.01,5.4618175506591795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.01,0.5603583812713623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,8,power_law_1.01,1.0295040130615234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,8,power_law_1.01,12.630534362792968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.01,0.740883207321167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,8,power_law_1.01,1.464134407043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.01,1.1019071578979491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.01,1.412287998199463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,8,power_law_1.01,1.8796159744262695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,8,power_law_1.01,2.0136320114135744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,8,power_law_1.01,2.9871936798095704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,8,power_law_1.01,2.694937515258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,8,power_law_1.01,6.381369781494141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,8,power_law_1.01,4.1109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,8,power_law_1.01,8.059929656982423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,1,8,power_law_1.2,0.08247039914131164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,8,power_law_1.01,0.032543998956680295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,1,8,power_law_1.2,0.08533120155334473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,8,power_law_1.01,0.031142398715019226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,1,8,power_law_1.2,0.08415359854698182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,8,power_law_1.01,0.030687999725341798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,1,8,power_law_1.2,0.08412160277366638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,8,power_law_1.01,0.03190400004386902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,1,8,power_law_1.2,0.08499199748039246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,8,power_law_1.01,0.03278079926967621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,1,8,power_law_1.2,0.08991360068321227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,8,power_law_1.01,0.037868800759315493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,1,8,power_law_1.2,0.10831359624862671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,8,power_law_1.01,0.06709120273590088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,1,8,power_law_1.2,0.12316160202026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,8,power_law_1.01,0.06810879707336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,1,8,power_law_1.2,0.14485759735107423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,8,power_law_1.01,0.06872959733009339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,1,8,power_law_1.2,0.14199039936065674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,1,8,power_law_1.2,0.14697599411010742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,8,power_law_1.01,0.06515200138092041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,1,8,power_law_1.2,0.1697983980178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,8,power_law_1.01,0.05801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,1,8,power_law_1.2,0.1863360047340393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,8,power_law_1.01,0.09338880181312562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,1,8,power_law_1.2,0.15877120494842528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,8,power_law_1.01,0.08736000061035157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,1,8,power_law_1.2,0.06190720200538635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,1,8,power_law_1.2,0.20398080348968506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,1,8,power_law_1.2,0.20422399044036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,8,power_law_1.01,0.07242879867553711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,1,8,power_law_1.2,0.06160640120506287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,1,8,power_law_1.2,0.1806272029876709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,8,power_law_1.01,0.1034432053565979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,1,8,power_law_1.2,0.23941121101379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,1,8,power_law_1.2,0.062118399143219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,1,8,power_law_1.2,0.288102388381958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,8,power_law_1.01,0.08320639729499817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,1,8,power_law_1.2,0.06490240097045899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,1,8,power_law_1.2,0.3913919925689697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,8,power_law_1.01,0.09688320159912109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,1,8,power_law_1.2,0.06562560200691223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,1,8,power_law_1.2,0.47628159523010255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,8,power_law_1.01,0.13841279745101928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,1,8,power_law_1.2,0.6799039840698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,1,8,power_law_1.2,0.06639999747276307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.01,0.16197119951248168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,1,8,power_law_1.2,0.8505663871765137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,1,8,power_law_1.2,0.06938239932060242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,1,8,power_law_1.2,1.3935551643371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.01,0.23018879890441896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,1,8,power_law_1.2,0.06952319741249084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,1,8,power_law_1.2,1.6167423248291015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.01,0.29205119609832764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,1,8,power_law_1.2,2.202681541442871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.01,0.42098560333251955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,1,8,power_law_1.2,0.07441920042037964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,1,8,power_law_1.2,3.555052947998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.01,0.528601598739624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,1,8,power_law_1.2,0.05384320020675659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,1,8,power_law_1.2,0.07693439722061157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,1,8,power_law_1.2,0.03272959887981415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,1,8,power_law_1.2,4.204979324340821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,1,8,power_law_1.2,0.07379199862480164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.01,0.7831744194030762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,1,8,power_law_1.2,0.05482239723205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,1,8,power_law_1.2,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,1,8,power_law_1.2,0.0836031973361969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,1,8,power_law_1.2,5.913983917236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.01,1.018943977355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,1,8,power_law_1.2,0.030483201146125793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,1,8,power_law_1.2,0.05434880256652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,1,8,power_law_1.2,0.08309119939804077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,1,8,power_law_1.2,0.03144319951534271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,1,8,power_law_1.2,12.593408203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.01,1.5168319702148438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,1,8,power_law_1.2,0.055929601192474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,1,8,power_law_1.2,0.09571840167045594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,1,8,power_law_1.2,0.03250559866428375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,1,8,power_law_1.2,0.11704319715499878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,1,8,power_law_1.2,0.03828479945659637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.01,2.007820892333984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,1,8,power_law_1.2,0.05588480234146118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,1,8,power_law_1.2,0.12146559953689576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,1,8,power_law_1.2,0.06718080043792725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,8,power_law_1.01,2.493395233154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,1,8,power_law_1.2,0.12131839990615845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,1,8,power_law_1.2,0.056518399715423585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,1,8,power_law_1.2,0.06778879761695862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,8,power_law_1.01,3.9513599395751955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,1,8,power_law_1.2,0.1570304036140442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,1,8,power_law_1.2,0.057651197910308837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,1,8,power_law_1.2,0.07063040137290955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,1,8,power_law_1.2,0.1956544041633606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,1,8,power_law_1.2,0.06496000289916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,8,power_law_1.01,7.83927001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,1,8,power_law_1.2,0.05873280167579651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,1,8,power_law_1.2,0.24976000785827637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,1,8,power_law_1.2,0.05894399881362915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,1,8,power_law_1.2,0.31162879467010496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,1,8,power_law_1.2,0.09515519738197327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,1,8,power_law_1.2,0.06272640228271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,1,8,power_law_1.2,0.4493567943572998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,1,8,power_law_1.2,0.08883839845657349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,1,8,power_law_1.2,0.5706943988800048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,1,8,power_law_1.2,0.06537600159645081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,1,8,power_law_1.2,0.07262719869613647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,1,8,power_law_1.2,0.7995520114898682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,4,balanced,0.06666133304437001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,4,balanced,0.039690665900707245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,1,8,power_law_1.2,0.06710399985313416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,1,8,power_law_1.2,1.0777983665466309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,1,8,power_law_1.2,0.1060863971710205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,4,balanced,0.03969600051641464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,1,8,power_law_1.2,1.6325439453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,1,8,power_law_1.2,0.07048959732055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,1,8,power_law_1.2,0.08421760201454162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,4,balanced,0.05825600028038025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,1,8,power_law_1.2,0.09840000271797181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,4,balanced,0.08213333288828532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,4,balanced,0.05865600208441416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,1,8,power_law_1.2,2.151020812988281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,1,8,power_law_1.2,0.07362560033798218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,4,balanced,0.08416533470153809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,1,8,power_law_1.2,0.07727360129356384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,1,8,power_law_1.2,0.1391808032989502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,4,balanced,0.06223999957243601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,4,balanced,0.08620267113049825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,1,8,power_law_1.2,2.5845375061035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,4,balanced,0.08412266770998637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,1,8,power_law_1.2,0.09253759980201721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,4,balanced,0.06187200049559275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.2,0.16318080425262452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,4,balanced,0.08460799853006999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,1,8,power_law_1.2,0.09679999947547913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,1,8,power_law_1.2,4.113990402221679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,4,balanced,0.08673066894213359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,4,balanced,0.043978666265805565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,4,balanced,0.0643039991458257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.2,0.23128321170806884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,1,8,power_law_1.2,0.10213119983673095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,4,balanced,0.09074667096138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,4,balanced,0.06409066418806712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,4,balanced,0.08912000060081482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.2,0.2939071893692017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,1,8,power_law_1.2,0.12005759477615356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,1,8,power_law_1.2,7.907981109619141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,4,balanced,0.09005332986513774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,4,balanced,0.06414400041103363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.2,0.42183680534362794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,4,balanced,0.041834667325019836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,4,balanced,0.09041600426038106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.2,0.13822720050811768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,4,balanced,0.06434133152167003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,4,balanced,0.04808533191680908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,4,balanced,0.09091200431187947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.2,0.5309440135955811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,4,balanced,0.06607466439406078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.2,0.17007360458374024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,4,balanced,0.05167999863624573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,4,balanced,0.09707732995351155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,4,balanced,0.06611200173695882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.2,0.787007999420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,4,balanced,0.05045333504676819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,4,balanced,0.10245333115259807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.2,0.23255679607391358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,4,balanced,0.06709866722424825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.2,1.0232768058776855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,4,balanced,0.05186133086681366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,4,balanced,0.10633599758148193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,4,balanced,0.0661653329928716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.2,0.3077183961868286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,4,balanced,0.05212800204753876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.2,1.5154303550720214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,4,balanced,0.10989866654078166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,4,balanced,0.07120533287525177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,4,balanced,0.05226666728655497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.2,0.3865407943725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.2,2.007923126220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,4,balanced,0.1243839959303538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,4,balanced,0.07923200229803722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,4,balanced,0.052111998200416565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.2,0.5984447956085205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,4,balanced,0.08009600142637889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,4,balanced,0.052058666944503784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,4,balanced,0.13726933797200522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,1,8,power_law_1.2,2.499513626098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,4,balanced,0.05276800195376078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,4,balanced,0.08529067039489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,4,balanced,0.1632960041364034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.2,0.7711296081542969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,4,balanced,0.054144000013669334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,1,8,power_law_1.2,3.9476608276367187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,4,balanced,0.05584000051021576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,4,balanced,0.18336000045140585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,4,balanced,0.09425066908200581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.2,1.1683584213256837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,4,balanced,0.06078400214513143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,4,balanced,0.23291200399398804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,4,balanced,0.10193600257237752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.2,1.5400447845458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,4,balanced,0.06879466772079468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,1,8,power_law_1.2,7.8385986328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,4,balanced,0.06841066479682922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,4,balanced,0.12270933389663696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,1,8,power_law_1.2,1.7363519668579102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,4,balanced,0.07072533170382182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,4,balanced,0.27832533915837604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,4,balanced,0.14074666301409403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,4,balanced,0.0812799980243047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,4,balanced,0.0867733359336853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,1,8,power_law_1.2,2.868953514099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,4,balanced,0.1804693341255188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,4,balanced,0.38309331734975177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,4,balanced,0.09913067022959392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,1,8,power_law_1.2,6.223430252075195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,4,balanced,0.11636799573898315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,4,balanced,0.2244053284327189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,4,balanced,0.4881440003712972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,4,balanced,0.14425599575042725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,4,balanced,0.3083786765734355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,4,balanced,0.17587200800577799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,4,balanced,0.6963573296864828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,4,balanced,0.23061867554982504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,4,balanced,0.3869226773579915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,4,balanced,0.2789919972419739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,4,balanced,0.8732799688975016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,4,balanced,0.55130668481191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,4,balanced,0.3862080176671346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,4,balanced,1.0838666756947835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,4,balanced,0.5063146750132242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,4,balanced,0.6875733534495035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,4,balanced,0.6234826644261678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,4,balanced,0.8481653531392416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,4,balanced,1.682458718617757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,4,balanced,0.9478507041931152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,4,balanced,1.3095946311950684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,4,balanced,3.3487412134806314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,4,balanced,2.5539040565490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,4,balanced,1.8687040011088054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,4,balanced,0.023930666347344715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,4,balanced,0.02363733450571696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,4,balanced,0.023498666783173878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,4,balanced,0.025583999852339428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,4,balanced,0.02550400048494339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,4,balanced,0.02717866748571396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,4,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,4,balanced,0.029743999242782593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,4,balanced,0.060959999759991966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,4,balanced,0.06097066899140676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,4,balanced,0.0605973352988561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,4,balanced,0.06271466612815857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,4,balanced,0.0621066689491272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,4,balanced,0.04819199939568838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,4,power_law_1.01,0.0732479989528656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,4,balanced,0.06885333359241486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,4,balanced,0.05824000140031179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,4,power_law_1.01,0.0781823992729187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,4,balanced,0.06516266862551372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,4,power_law_1.01,0.07562879920005798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,4,balanced,0.08474666873613994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,4,power_law_1.01,0.0838591992855072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,4,balanced,0.10098666946093242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,4,power_law_1.01,0.0832256019115448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,4,balanced,0.14070933063824972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,4,power_law_1.01,0.08625919818878174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,4,power_law_1.01,0.09155200123786926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,4,balanced,0.17358932892481485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,4,power_law_1.01,0.09785599708557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,4,balanced,0.24174400170644125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,4,power_law_1.01,0.10927360057830811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,4,power_law_1.01,0.11747200489044189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,4,balanced,0.30171199639638263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,4,power_law_1.01,0.09909120202064514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,4,power_law_1.01,0.10414079427719117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,4,balanced,0.43879465262095135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,4,power_law_1.01,0.10587519407272339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,4,power_law_1.01,0.1214143991470337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,4,balanced,0.5674826701482137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,4,power_law_1.01,0.14379520416259767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,4,power_law_1.01,0.16124160289764405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,4,balanced,0.8317920366923014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,4,power_law_1.01,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,4,power_law_1.01,0.14261759519577027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,4,power_law_1.01,0.049516800045967105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,4,power_law_1.01,0.1914944052696228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,4,power_law_1.01,0.05321599841117859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,4,power_law_1.01,0.20659201145172118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,4,power_law_1.01,0.04314239919185638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,4,power_law_1.01,0.3211008071899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,4,balanced,1.0967520078023274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,4,power_law_1.01,0.05070080161094666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,4,power_law_1.01,0.045542401075363156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,4,power_law_1.01,0.37764480113983157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,4,power_law_1.01,0.05958399772644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,4,power_law_1.01,0.5179967880249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,4,power_law_1.01,0.06045439839363098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,4,power_law_1.01,0.04924800097942352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,4,power_law_1.01,0.5268223762512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,4,power_law_1.01,0.06076800227165222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,4,power_law_1.01,0.049516800045967105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,4,balanced,1.3609600067138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,4,power_law_1.01,0.06166399717330932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,4,power_law_1.01,0.8853759765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,4,power_law_1.01,0.050169599056243894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,4,power_law_1.01,0.06492159962654113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,4,power_law_1.01,0.05138559937477112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,4,power_law_1.01,0.06732159852981567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,4,power_law_1.01,1.5391551971435546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,4,power_law_1.01,0.05177599787712097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,4,power_law_1.01,0.07084800004959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,4,power_law_1.01,2.149068832397461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,4,power_law_1.01,0.05570560097694397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,4,power_law_1.01,0.07059199810028076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,4,power_law_1.01,0.056492799520492555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,4,power_law_1.01,0.07325440049171447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,4,balanced,2.1504480044047036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,4,power_law_1.01,2.2483776092529295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,4,power_law_1.01,0.05864319801330566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,4,power_law_1.01,0.07683839797973632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,4,power_law_1.01,2.579046440124512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,4,power_law_1.01,0.06242560148239136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,4,power_law_1.01,0.061900800466537474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,4,power_law_1.01,0.08056960105895997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,4,power_law_1.01,4.595782470703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,4,power_law_1.01,0.06891520023345947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,4,power_law_1.01,0.09080320000648498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,4,power_law_1.01,0.07857279777526856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,4,power_law_1.01,0.09893119931221009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,4,power_law_1.01,8.86263656616211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,4,power_law_1.01,0.0778432011604309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,4,power_law_1.01,0.10766079425811767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,4,power_law_1.01,0.09062399864196777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,4,power_law_1.01,0.13089280128479003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,4,power_law_1.01,0.09879680275917054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,4,power_law_1.01,0.15749759674072267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,4,power_law_1.01,0.1910591959953308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.01,0.11223039627075196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,4,balanced,4.262949307759603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,4,power_law_1.01,0.23384320735931396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.01,0.15239039659500123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,4,power_law_1.01,0.2731008052825928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.01,0.18846720457077026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,4,power_law_1.01,0.022649599611759184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,4,power_law_1.01,0.42740478515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.01,0.24566400051116943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,4,power_law_1.01,0.5879487991333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,4,power_law_1.01,0.022060799598693847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,4,power_law_1.01,0.021331200003623964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.01,0.30992000102996825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,4,power_law_1.01,0.7321536064147949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,4,power_law_1.01,0.024345600605010988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.01,0.41086721420288086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,4,power_law_1.01,0.024633599817752837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,4,power_law_1.01,1.044927978515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.01,0.46460161209106443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,4,power_law_1.01,0.02531839907169342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,4,power_law_1.01,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,4,power_law_1.01,1.1711872100830079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.01,0.7723711967468262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,4,power_law_1.01,0.029075199365615846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.01,1.0913663864135743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,4,power_law_1.01,1.3766143798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,4,power_law_1.01,0.05925760269165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,4,power_law_1.01,1.474950408935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,4,power_law_1.01,3.139839935302734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,4,power_law_1.01,0.06104959845542908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,4,power_law_1.01,2.6616512298583985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,4,power_law_1.01,4.695539093017578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,4,power_law_1.01,0.05917440056800842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,4,power_law_1.01,4.531289672851562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,4,power_law_1.01,0.06569600105285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,4,power_law_1.01,0.05675519704818725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,4,power_law_1.01,0.04609920084476471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,4,power_law_1.01,0.06377599835395813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,4,power_law_1.01,0.05575039982795715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,4,power_law_1.01,0.06280959844589233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,4,power_law_1.01,0.08220160007476807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.01,0.09813759922981262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.01,0.1390720009803772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,4,power_law_1.2,0.07729920148849487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,4,power_law_1.2,0.057068800926208495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,4,power_law_1.2,0.047577598690986635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.01,0.1713536024093628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,4,power_law_1.2,0.0839743971824646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,4,power_law_1.2,0.04853120148181915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.01,0.23936638832092286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,4,power_law_1.2,0.04591360092163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,4,power_law_1.2,0.08024320006370544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,4,power_law_1.2,0.05509120225906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.01,0.2980736017227173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,4,power_law_1.2,0.043833601474761966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,4,power_law_1.2,0.0839680016040802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,4,power_law_1.2,0.08387839794158936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.01,0.4351039886474609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,4,power_law_1.2,0.04924159944057464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,4,power_law_1.2,0.05998079776763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,4,power_law_1.2,0.08624640107154846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.01,0.5630271911621094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,4,power_law_1.2,0.06050559878349304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,4,power_law_1.2,0.04898560047149658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,4,power_law_1.2,0.09217280149459839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.01,0.8277824401855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,4,power_law_1.2,0.060787200927734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,4,power_law_1.2,0.10386559963226319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,4,power_law_1.2,0.05009920001029968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,4,power_law_1.2,0.11166080236434936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.01,1.0926655769348144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,4,power_law_1.2,0.06324480175971985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,4,power_law_1.2,0.11294080018997192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,4,power_law_1.2,0.050732797384262084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,4,power_law_1.01,1.356172752380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,4,power_law_1.2,0.06663680076599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,4,power_law_1.2,0.09915519952774048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,4,power_law_1.2,0.052121597528457644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,4,power_law_1.01,2.1442495346069337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,4,power_law_1.2,0.10683519840240478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,4,power_law_1.2,0.06830080151557923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,4,power_law_1.2,0.05591679811477661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,4,power_law_1.2,0.10382720232009887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,4,power_law_1.2,0.06865280270576476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,4,power_law_1.01,4.252358245849609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,4,power_law_1.2,0.05621119737625122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,4,power_law_1.2,0.12362879514694214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,4,power_law_1.2,0.14815360307693481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,4,power_law_1.2,0.06981120109558106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,4,power_law_1.2,0.057734400033950806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,4,power_law_1.2,0.17464319467544556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,4,power_law_1.2,0.15890560150146485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,4,power_law_1.2,0.06104320287704468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,4,power_law_1.2,0.07294719815254211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,4,power_law_1.2,0.1988927960395813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,4,power_law_1.2,0.24022400379180908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,4,power_law_1.2,0.062009602785110474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,4,power_law_1.2,0.07393280267715455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,4,power_law_1.2,0.2841792106628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,4,power_law_1.2,0.06930559873580933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,4,power_law_1.2,0.3792959928512573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,4,power_law_1.2,0.08855680227279664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,4,power_law_1.2,0.49402241706848143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,4,power_law_1.2,0.07783679962158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,4,power_law_1.2,0.09859840273857116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,4,power_law_1.2,0.6666240215301513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,4,power_law_1.2,1.110905647277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,4,power_law_1.2,0.08241279721260071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,4,power_law_1.2,0.0998848021030426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,4,power_law_1.2,1.4063424110412597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,4,power_law_1.2,0.08789759874343872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,4,power_law_1.2,0.11324160099029541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,4,power_law_1.2,0.11568640470504761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,4,power_law_1.2,1.955731201171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,4,power_law_1.2,0.14446719884872436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.2,0.1406592011451721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,4,power_law_1.2,2.356697654724121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,4,power_law_1.2,0.15905280113220216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.2,0.18171520233154298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,4,power_law_1.2,3.266457748413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,4,power_law_1.2,0.203110408782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.2,0.21395199298858641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,4,power_law_1.2,4.639622497558594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,4,power_law_1.2,0.28102400302886965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.2,0.28462719917297363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.2,0.346943998336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,4,power_law_1.2,0.3059776067733765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,4,power_law_1.2,11.224896240234376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.2,0.45367679595947263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,4,power_law_1.2,0.3855808019638062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.2,0.6791999816894532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,4,power_law_1.2,0.584819221496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.2,0.7881408214569092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,4,power_law_1.2,0.8430208206176758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.2,1.4193087577819825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,4,power_law_1.2,1.0794943809509276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,4,power_law_1.2,1.5936256408691407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,4,power_law_1.2,1.5537792205810548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,4,power_law_1.2,2.6822847366333007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,4,power_law_1.2,1.528326416015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,4,power_law_1.2,4.740134429931641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,4,power_law_1.2,3.536511993408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,4,power_law_1.2,5.63683853149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,4,power_law_1.2,0.023443199694156647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,4,power_law_1.2,0.021939200162887574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,4,power_law_1.2,0.02144639939069748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,4,power_law_1.2,0.02502399981021881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,4,power_law_1.2,0.02508159875869751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,4,power_law_1.2,0.0255295991897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,4,power_law_1.2,0.03573119938373566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,4,power_law_1.2,0.02874239981174469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,4,power_law_1.2,0.059059202671051025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,4,power_law_1.2,0.06112639904022217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,4,power_law_1.2,0.059411197900772095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,4,power_law_1.2,0.06449919939041138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,4,power_law_1.2,0.057094401121139525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,4,power_law_1.2,0.04620800018310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,4,power_law_1.2,0.06382079720497132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,2,balanced,0.04419200122356415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,4,power_law_1.2,0.05546879768371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,4,power_law_1.2,0.06298239827156067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,2,balanced,0.05904000004132589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,2,balanced,0.08161599934101105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,4,power_law_1.2,0.08257279992103576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,2,balanced,0.08385599652926128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,2,balanced,0.08531733353932698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.2,0.09848319888114929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,2,balanced,0.08453866839408875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.2,0.13779840469360352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,2,balanced,0.0858026643594106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.2,0.17136000394821166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,2,balanced,0.08937066793441772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,2,balanced,0.09025599559148152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.2,0.23947520256042482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.2,0.2982719898223877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,2,balanced,0.08867200215657552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,2,balanced,0.03182400017976761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,2,balanced,0.09005332986513774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.2,0.4352255821228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,2,balanced,0.08861333131790161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,2,balanced,0.03748800108830134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.2,0.5630015850067138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,2,balanced,0.05624533196290334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,2,balanced,0.08966933687527974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,2,balanced,0.03581333408753077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.2,0.8296064376831055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,2,balanced,0.060720001657803856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,2,balanced,0.039488000174363456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,2,balanced,0.0937493344148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.2,1.0929408073425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,2,balanced,0.04594666759173075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,2,balanced,0.059631998340288796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,2,balanced,0.10950932900110881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,2,balanced,0.04877333343029022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,4,power_law_1.2,1.3556096076965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,2,balanced,0.04756266872088114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,2,balanced,0.06142933170000712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,2,balanced,0.11214400331179301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,4,power_law_1.2,2.1595840454101562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,2,balanced,0.04846400022506714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,2,balanced,0.06251733501752217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,2,balanced,0.05099200208981832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,2,balanced,0.10326400399208069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,2,balanced,0.021573332448800404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,2,balanced,0.049695998430252075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,2,balanced,0.06208533545335134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,2,balanced,0.04976533353328705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,2,balanced,0.062218666076660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,4,power_law_1.2,4.258246231079101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,2,balanced,0.13193600376447043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,2,balanced,0.04987200101216634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,2,balanced,0.06407999992370605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,2,balanced,0.05151999990145365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,2,balanced,0.06574399769306183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,2,balanced,0.1309279998143514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,2,balanced,0.054058666030565895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,2,balanced,0.022805333137512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,2,balanced,0.06844800213972728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,2,balanced,0.053930665055910744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,2,balanced,0.15241600076357523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,2,balanced,0.062261333068211876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,2,balanced,0.02327999969323476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,2,balanced,0.06816000243028005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,2,balanced,0.18655999501546225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,2,balanced,0.0681333343187968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,2,balanced,0.021568000316619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,2,balanced,0.07055999835332234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,2,balanced,0.06817066669464111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,2,balanced,0.07154666880766551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,2,balanced,0.2367039918899536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,2,balanced,0.025237334271272022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,2,balanced,0.08342400193214417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,2,balanced,0.0784853349129359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,2,balanced,0.08991466959317525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,2,balanced,0.07900799810886383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,2,balanced,0.025424001117547352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,2,balanced,0.28962133328119916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,2,balanced,0.10455999771753947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,2,balanced,0.08085333307584126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,2,balanced,0.025546667476495106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,2,balanced,0.13216533263524374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,2,balanced,0.09525866309801738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,2,balanced,0.02733866622050603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,2,balanced,0.39511998494466144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,2,balanced,0.1651946703592936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,2,balanced,0.10257066289583842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,2,balanced,0.02959999938805898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,2,balanced,0.19883733987808228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,2,balanced,0.02739199995994568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,2,balanced,0.4875306685765584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,2,balanced,0.12095466256141663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,2,balanced,0.029487999776999157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,2,balanced,0.26075200239817303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,2,power_law_1.01,0.05246080160140991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,2,balanced,0.14332266648610434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,2,balanced,0.0582239975531896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,2,balanced,0.1821333368619283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,2,balanced,0.32470399141311646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,2,balanced,0.6995200316111246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,2,power_law_1.01,0.05570560097694397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,2,balanced,0.05801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,2,balanced,0.05876799921194712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,2,balanced,0.23261332511901855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,2,power_law_1.01,0.06644480228424073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,2,balanced,0.04635733366012573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,2,balanced,0.45263465245564777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,2,power_law_1.01,0.07753599882125854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,2,balanced,0.04366933306058248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,2,power_law_1.01,0.08357120156288148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,2,balanced,0.31563733021418255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,2,balanced,0.8986720244089762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,2,balanced,0.046869332591692604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,2,power_law_1.01,0.0852735996246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,2,power_law_1.01,0.08900480270385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,2,balanced,0.05971199770768484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,2,balanced,0.5854560136795044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,2,power_law_1.01,0.09156479835510253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,2,balanced,0.395248015721639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,2,balanced,0.06683733562628429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,2,power_law_1.01,0.09407359957695008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,2,power_law_1.01,0.09026560187339783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,2,power_law_1.01,0.035078400373458864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,2,balanced,0.08825066685676575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,2,balanced,1.1131892999013264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,2,power_law_1.01,0.09002879858016968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,2,balanced,0.7193066279093424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,2,balanced,0.5571253299713135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,2,power_law_1.01,0.095769602060318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,2,power_law_1.01,0.04140160083770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,2,balanced,0.10822400450706482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,2,power_law_1.01,0.09654399752616882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,2,power_law_1.01,0.04620159864425659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,2,power_law_1.01,0.1091007947921753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,2,balanced,0.1467626690864563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,2,power_law_1.01,0.05864319801330566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,2,power_law_1.01,0.11997439861297607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,2,power_law_1.01,0.05942400097846985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,2,power_law_1.01,0.13114880323410033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,2,balanced,0.7188906669616699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,2,balanced,0.18121600151062012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,2,power_law_1.01,0.06074240207672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,2,power_law_1.01,0.13114880323410033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,2,power_law_1.01,0.061382400989532473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,2,power_law_1.01,0.16017279624938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,2,balanced,1.7149920463562012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,2,balanced,1.0914133389790852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,2,power_law_1.01,0.06287999749183655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,2,power_law_1.01,0.17500159740447999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,2,balanced,0.2570026715596517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,2,power_law_1.01,0.06535040140151978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,2,power_law_1.01,0.20038399696350098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,2,power_law_1.01,0.036396801471710205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,2,power_law_1.01,0.06555520296096802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,2,balanced,0.8809706370035807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,2,power_law_1.01,0.2596800088882446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,2,power_law_1.01,0.0667519986629486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,2,balanced,0.32865599791208905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,2,power_law_1.01,0.34873600006103517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,2,power_law_1.01,0.038841599225997926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,2,power_law_1.01,0.07085440158843995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,2,power_law_1.01,0.3953344106674194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,2,power_law_1.01,0.07363839745521546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,2,power_law_1.01,0.6903872013092041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,2,power_law_1.01,0.03795199990272522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,2,power_law_1.01,0.07160320281982421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,2,power_law_1.01,0.7374591827392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,2,power_law_1.01,0.04656000137329101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,2,balanced,0.4750346740086873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,2,power_law_1.01,0.0813696026802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,2,power_law_1.01,0.04671359956264496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,2,power_law_1.01,1.1524800300598144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,2,balanced,1.3594667116800945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,2,power_law_1.01,0.08399360179901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,2,power_law_1.01,0.048307201266288756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,2,power_law_1.01,1.3576064109802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,2,power_law_1.01,0.047916799783706665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,2,power_law_1.01,0.08967679738998413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,2,balanced,2.1142346064249673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,2,power_law_1.01,0.048044800758361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,2,power_law_1.01,0.10910719633102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,2,power_law_1.01,1.707993507385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,2,balanced,0.623141328493754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,2,power_law_1.01,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,2,power_law_1.01,0.11544320583343506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,2,balanced,3.3397865295410156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,2,power_law_1.01,2.5249664306640627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,2,power_law_1.01,0.05136640071868896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,2,power_law_1.01,0.15063680410385133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,2,power_law_1.01,0.05487359762191772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,2,power_law_1.01,0.18968960046768188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,2,power_law_1.01,5.203961563110352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,2,power_law_1.01,0.05786240100860596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,2,power_law_1.01,0.24896640777587892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,2,power_law_1.01,0.06130560040473938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,2,power_law_1.01,0.30453760623931886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,2,balanced,0.768730640411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,2,power_law_1.01,0.0642624020576477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,2,power_law_1.01,0.44111361503601076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,2,power_law_1.01,0.0726527988910675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,2,power_law_1.01,0.5761983871459961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,2,power_law_1.01,0.07409279942512512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,2,power_law_1.01,0.6715328216552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,2,power_law_1.01,0.07715200185775757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,2,balanced,2.632304032643636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,2,power_law_1.01,1.0941056251525878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,2,power_law_1.01,0.093612802028656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.01,0.10555520057678222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,2,power_law_1.01,1.281049633026123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.01,0.12005759477615356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,2,balanced,1.2120213508605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,2,power_law_1.01,1.7531072616577148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.01,0.15961600542068483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.01,0.1966591954231262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,2,power_law_1.01,3.9946048736572264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.01,0.2656383991241455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.01,0.32598400115966797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.01,0.4217343807220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.01,0.6346176147460938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.01,0.8322688102722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,2,power_law_1.01,0.8945856094360352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,2,balanced,2.393727938334147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,2,power_law_1.01,1.5865856170654298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,2,power_law_1.01,3.1197376251220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,2,power_law_1.01,0.02030719965696335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,2,power_law_1.01,0.02099200040102005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,2,power_law_1.01,0.020665599405765532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,2,power_law_1.01,0.02213120013475418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,2,power_law_1.01,0.02412160038948059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,2,power_law_1.01,0.025536000728607178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,2,power_law_1.01,0.024857600033283234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,2,power_law_1.01,0.025548800826072693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,2,power_law_1.01,0.02797439992427826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,2,power_law_1.01,0.02709760069847107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,2,power_law_1.01,0.027750399708747864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,2,power_law_1.01,0.0561024010181427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,2,power_law_1.01,0.056595200300216676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,2,power_law_1.01,0.05797119736671448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,2,power_law_1.01,0.045151999592781066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,2,power_law_1.01,0.043007999658584595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,2,power_law_1.01,0.04455040097236633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,2,power_law_1.01,0.05767679810523987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.01,0.06517760157585144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.01,0.08592000007629394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.01,0.10754560232162476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.01,0.14581120014190674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.01,0.17932159900665284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.01,0.2537600040435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.01,0.325433611869812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.01,0.4742784023284912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.01,0.6194560050964355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,2,power_law_1.01,0.7692543983459472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,2,power_law_1.01,1.2140159606933594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,2,power_law_1.01,2.391091156005859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,2,power_law_1.2,0.05381760001182556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,2,power_law_1.2,0.038227200508117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,2,power_law_1.2,0.037811198830604555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,2,power_law_1.2,0.03803519904613495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,2,power_law_1.2,0.05681920051574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,2,power_law_1.2,0.03863680064678192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,2,power_law_1.2,0.06392319798469544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,2,power_law_1.2,0.058406400680541995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,2,power_law_1.2,0.08007680177688599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,2,power_law_1.2,0.08261759877204895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,2,power_law_1.2,0.058483201265335086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,2,power_law_1.2,0.037574398517608645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,2,power_law_1.2,0.0845632016658783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,2,power_law_1.2,0.06044800281524658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,1,balanced,0.034688000877698265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,2,power_law_1.2,0.08979840278625488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,2,power_law_1.2,0.061350399255752565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,2,power_law_1.2,0.039750400185585025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,2,power_law_1.2,0.020556800067424774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,2,power_law_1.2,0.06227200031280518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,2,power_law_1.2,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,1,balanced,0.05426666637261709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,2,power_law_1.2,0.04739840030670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,1,balanced,0.08060800035794576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,2,power_law_1.2,0.08956800103187561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,2,power_law_1.2,0.06451839804649354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,2,power_law_1.2,0.020729599893093108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,1,balanced,0.08259200056393941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,2,power_law_1.2,0.06435199975967407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,2,power_law_1.2,0.09287679791450501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,2,power_law_1.2,0.047660800814628604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,2,power_law_1.2,0.020735999941825865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,1,balanced,0.0842186709245046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,2,power_law_1.2,0.0665727972984314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,2,power_law_1.2,0.09102720022201538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,2,power_law_1.2,0.02101760059595108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,2,power_law_1.2,0.04787200093269348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,2,power_law_1.2,0.06966400146484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,1,balanced,0.08448533217112224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,2,power_law_1.2,0.09562240242958069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,2,power_law_1.2,0.07427200078964233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,2,power_law_1.2,0.025388801097869874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,2,power_law_1.2,0.04870400130748749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,1,balanced,0.0862559974193573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,2,power_law_1.2,0.09721599817276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,2,power_law_1.2,0.07196800112724304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,2,power_law_1.2,0.025676798820495606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,1,balanced,0.03181333343187968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,1,balanced,0.08689600229263306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,2,power_law_1.2,0.10246399641036988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,2,power_law_1.2,0.0491456001996994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,2,power_law_1.2,0.08177279829978942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,1,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,1,balanced,0.08690133690834045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,2,power_law_1.2,0.024774399399757386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,2,power_law_1.2,0.1188480019569397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,2,power_law_1.2,0.05203840136528015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,2,power_law_1.2,0.08547840118408204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,1,balanced,0.08690133690834045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,2,power_law_1.2,0.025145599246025087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,1,balanced,0.035589332381884255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,2,power_law_1.2,0.09041919708251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,2,power_law_1.2,0.12922879457473754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,2,power_law_1.2,0.05225600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,1,balanced,0.05528533458709717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,1,balanced,0.08874666690826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,2,power_law_1.2,0.02863999903202057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,2,power_law_1.2,0.1035647988319397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,1,balanced,0.05648533503214518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,2,power_law_1.2,0.12918399572372435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,2,power_law_1.2,0.05594239830970764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,1,balanced,0.07293866574764252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,1,balanced,0.08852266271909077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,1,balanced,0.06124266485373179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,2,power_law_1.2,0.1476415991783142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,2,power_law_1.2,0.11930880546569825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,2,power_law_1.2,0.02675839960575104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,2,power_law_1.2,0.057811200618743896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,1,balanced,0.07441600163777669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,1,balanced,0.09127466877301534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,2,power_law_1.2,0.17834880352020263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,1,balanced,0.060592000683148704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,2,power_law_1.2,0.1600383996963501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,1,balanced,0.0726986676454544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,2,power_law_1.2,0.05982080101966858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,2,power_law_1.2,0.02768639922142029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,2,power_law_1.2,0.2248960018157959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,1,balanced,0.06205866734186808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,1,balanced,0.09377599755922954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,1,balanced,0.07039999961853027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,2,power_law_1.2,0.0643775999546051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,2,power_law_1.2,0.18443520069122316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,2,power_law_1.2,0.23267199993133544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,2,power_law_1.2,0.0559935986995697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,1,balanced,0.08000533282756805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,1,balanced,0.06418133278687795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,1,balanced,0.10550399621327718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,2,power_law_1.2,0.07536640167236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,2,power_law_1.2,0.24662399291992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,2,power_law_1.2,0.30995199680328367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,2,power_law_1.2,0.056492799520492555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,1,balanced,0.07099199791749318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,1,balanced,0.06299733122189839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,1,balanced,0.1074079970518748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,2,power_law_1.2,0.0770687997341156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,2,power_law_1.2,0.31153919696807864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,1,balanced,0.07258133093516032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,2,power_law_1.2,0.42661118507385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,2,power_law_1.2,0.05783680081367493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,1,balanced,0.06454933186372121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,2,power_law_1.2,0.08330879807472229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,1,balanced,0.11258133252461751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,2,power_law_1.2,0.4215231895446777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,1,balanced,0.07313600182533264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,2,power_law_1.2,0.5219711780548095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,2,power_law_1.2,0.6735551834106446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,2,power_law_1.2,0.09664000272750854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,2,power_law_1.2,0.045286399126052854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,1,balanced,0.06541866560777028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,1,balanced,0.07306666672229767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,1,balanced,0.12548800309499106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,2,power_law_1.2,0.7447999954223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,2,power_law_1.2,0.8144703865051269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.2,0.10852479934692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,2,power_law_1.2,0.04297600090503693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,1,balanced,0.06497066716353099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,2,power_law_1.2,1.0023743629455566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,1,balanced,0.1361120045185089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,1,balanced,0.07121066749095917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,2,power_law_1.2,1.0283136367797852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.2,0.13175040483474731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,2,power_law_1.2,0.044409599900245664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,1,balanced,0.06716266771157582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,2,power_law_1.2,1.3031871795654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,1,balanced,0.07356266677379608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.2,0.1652799963951111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,2,power_law_1.2,1.5018560409545898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,1,balanced,0.1600053310394287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,2,power_law_1.2,0.0573311984539032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,1,balanced,0.06782400111357371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.2,0.1965376019477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,1,balanced,0.07230933507283528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,2,power_law_1.2,1.9548416137695312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,2,power_law_1.2,1.767308807373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,1,balanced,0.19147199392318726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,1,balanced,0.07229333122571309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.2,0.06520320177078247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.2,0.28183679580688475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,1,balanced,0.07592533528804779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,2,power_law_1.2,2.950508880615234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,2,power_law_1.2,3.6397247314453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,1,balanced,0.0846720039844513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,1,balanced,0.24023467302322388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.2,0.3883455991744995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.2,0.08655359745025634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,1,balanced,0.07463466624418895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,1,balanced,0.08284266789754231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,2,power_law_1.2,5.890655899047852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,1,balanced,0.3037173350652059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.2,0.490169620513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,1,balanced,0.0786186655362447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.2,0.10727039575576783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,1,balanced,0.08490133285522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.2,0.6608191967010498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,1,balanced,0.10900800426801045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.2,0.1460479974746704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,1,balanced,0.4143199920654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,1,balanced,0.11212799946467082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.2,0.9628735542297363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.2,0.17945599555969238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,1,balanced,0.10318932930628459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,1,balanced,0.14013333121935526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,1,balanced,0.5180266698201498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,2,power_law_1.2,1.206272029876709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.2,0.2530751943588257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,1,balanced,0.16738667090733847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,2,power_law_1.2,1.7904064178466796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,1,balanced,0.11115200320879619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.2,0.3258368015289307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,1,balanced,0.22207466761271158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,2,power_law_1.2,3.020582389831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,1,balanced,0.13926933209101358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.2,0.47325439453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,1,balanced,0.7349706490834554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,1,balanced,0.279968003431956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.2,0.6203711986541748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,1,balanced,0.17117865880330405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,1,balanced,0.3909440040588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,2,power_law_1.2,0.7697535991668701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,1,balanced,0.22329066197077432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,1,balanced,0.951690673828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,1,balanced,0.5046399831771851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,2,power_law_1.2,1.2128128051757812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,1,balanced,0.2708853284517924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,2,power_law_1.2,2.395552062988281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,1,balanced,0.3807520071665446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,1,balanced,0.7226186593373617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,1,balanced,1.1694347063700359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,1,balanced,0.4882773160934448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,1,balanced,0.9541973272959391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,1,balanced,0.6978027025858561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,1,balanced,1.8225706418355305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,1,balanced,1.1793920199076335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,1,balanced,0.9107093016306559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,1,balanced,1.1174293359120686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,1,balanced,1.8598133722941081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,1,balanced,0.019413333386182785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,1,balanced,3.6182241439819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,1,balanced,1.737925370534261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,1,balanced,0.019434666881958645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,1,balanced,0.019434666881958645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,1,balanced,0.019167999426523846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,1,balanced,0.019189332922299702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,1,balanced,0.020314666132132213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,1,balanced,3.683429400126139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,1,balanced,0.021151999632517498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,1,balanced,0.021935999393463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,1,balanced,0.02537599951028824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,1,balanced,3.400202751159668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,1,balanced,0.029311999678611755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,1,balanced,0.025290665527184803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,1,balanced,0.026799999177455902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,1,balanced,0.029626667499542236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,1,balanced,0.029631999631722767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,1,balanced,0.03180799881617228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,1,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,1,balanced,0.035877334574858345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,1,balanced,0.047877331574757896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,1,balanced,0.04984533290068308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,1,balanced,0.06514666477839152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,1,balanced,0.07739733159542084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,1,balanced,0.09918399651845296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,1,balanced,0.12360533078511556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,1,balanced,0.16787733634312949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,1,balanced,0.21060800552368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,1,balanced,0.297541340192159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,1,power_law_1.01,0.03818880021572113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,1,power_law_1.01,0.047539201378822324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,1,power_law_1.01,0.05454720258712768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,1,power_law_1.01,0.08092799782752991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,1,balanced,0.38397332032521564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,1,power_law_1.01,0.08175359964370728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,1,power_law_1.01,0.08379520177841186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,1,power_law_1.01,0.08642560243606567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,1,power_law_1.01,0.0864960014820099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,1,power_law_1.01,0.0877951979637146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,1,balanced,0.4727253516515096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,1,power_law_1.01,0.08835840225219727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,1,power_law_1.01,0.08698239922523499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,1,power_law_1.01,0.09217280149459839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,1,power_law_1.01,0.03134720027446747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,1,power_law_1.01,0.09410560131072998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,1,power_law_1.01,0.09826560020446777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,1,power_law_1.01,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,1,power_law_1.01,0.10853760242462158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,1,power_law_1.01,0.03621760010719299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,1,balanced,0.7370560169219971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,1,power_law_1.01,0.11239039897918701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,1,power_law_1.01,0.045184001326560974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,1,power_law_1.01,0.05743359923362732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,1,power_law_1.01,0.11811840534210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,1,power_law_1.01,0.058220797777175905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,1,power_law_1.01,0.13797119855880738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,1,power_law_1.01,0.05085440278053284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,1,power_law_1.01,0.05989120006561279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,1,power_law_1.01,0.14021120071411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,1,power_law_1.01,0.05669119954109192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,1,power_law_1.01,0.06162559986114502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,1,power_law_1.01,0.07955200076103211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,1,power_law_1.01,0.17560960054397584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,1,power_law_1.01,0.061427199840545656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,1,power_law_1.01,0.07134720087051391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,1,power_law_1.01,0.19746559858322144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,1,power_law_1.01,0.062489598989486694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,1,power_law_1.01,0.07118719816207886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,1,power_law_1.01,0.26640639305114744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,1,power_law_1.01,0.06361600160598754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,1,power_law_1.01,0.07226880192756653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,1,power_law_1.01,0.31228160858154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,1,power_law_1.01,0.06529920101165772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,1,power_law_1.01,0.07204480171203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,1,power_law_1.01,0.06759039759635925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,1,power_law_1.01,0.43916797637939453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,1,power_law_1.01,0.07106559872627258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,1,power_law_1.01,0.07075200080871583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,1,power_law_1.01,0.5322175979614258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,1,power_law_1.01,0.07190399765968322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,1,power_law_1.01,0.07378559708595275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,1,power_law_1.01,0.7484928131103515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,1,power_law_1.01,0.07310079932212829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,1,power_law_1.01,0.08325759768486023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,1,power_law_1.01,0.9705663681030273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,1,power_law_1.01,0.07619839906692505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,1,power_law_1.01,0.08629119992256165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,1,power_law_1.01,0.01815039962530136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,1,power_law_1.01,1.1855487823486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,1,power_law_1.01,0.07653759717941284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,1,power_law_1.01,0.08869760036468506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,1,balanced,1.4424586296081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,1,power_law_1.01,0.08179839849472045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,1,power_law_1.01,0.10671360492706299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,1,power_law_1.01,1.8456127166748046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,1,power_law_1.01,0.01770240068435669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,1,power_law_1.01,0.09071360230445862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,1,power_law_1.01,0.1164736032485962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,1,power_law_1.01,0.01791999936103821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,1,power_law_1.01,0.09006720185279846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,1,power_law_1.01,0.14597760438919066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,1,power_law_1.01,3.650643157958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,1,power_law_1.01,0.018675200641155243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,1,power_law_1.01,0.09354879856109619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,1,power_law_1.01,0.1804095983505249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,1,power_law_1.01,0.01930239945650101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,1,power_law_1.01,0.11347839832305909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,1,power_law_1.01,0.019359999895095827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,1,power_law_1.01,0.2309312105178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.01,0.12528640031814575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,1,power_law_1.01,0.019833600521087645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,1,power_law_1.01,0.276908802986145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.01,0.16168960332870483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,1,power_law_1.01,0.020588800311088562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.01,0.18484480381011964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,1,power_law_1.01,0.38869121074676516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.01,0.23856639862060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,1,power_law_1.01,0.022892799973487855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,1,power_law_1.01,0.4947519779205322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.01,0.2942528009414673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,1,power_law_1.01,0.02739199995994568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,1,power_law_1.01,0.7027840137481689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.01,0.4086400032043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,1,power_law_1.01,0.02426239997148514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,1,power_law_1.01,0.9099648475646973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.01,0.5271296024322509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,1,power_law_1.01,0.025337600708007814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,1,power_law_1.01,1.1208127975463866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,1,power_law_1.01,0.027539199590682982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.01,0.7426176071166992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,1,power_law_1.01,1.7410816192626952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,1,power_law_1.01,0.0293503999710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.01,0.9715968132019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,1,power_law_1.01,0.03149439990520477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.01,1.197932815551758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,1,power_law_1.01,3.3926017761230467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,1,power_law_1.01,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,1,power_law_1.01,1.869830322265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,1,power_law_1.01,0.035078400373458864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,1,power_law_1.01,3.6798336029052736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,1,power_law_1.01,0.04573439955711365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.01,0.047737601399421695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,1,power_law_1.2,0.036601600050926206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.01,0.06268799901008607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.01,0.07712000012397766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,1,power_law_1.2,0.046137601137161255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.01,0.09768959879875183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,1,power_law_1.2,0.05292159914970398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.01,0.12031359672546386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,1,power_law_1.2,0.08072959780693054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.01,0.16483839750289916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,1,power_law_1.2,0.08182399868965148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.01,0.20912001132965088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,1,power_law_1.2,0.0846783995628357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.01,0.29499518871307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,1,power_law_1.2,0.08728960156440735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.01,0.38156158924102784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,1,power_law_1.2,0.0864575982093811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,1,power_law_1.2,0.08892800211906433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.01,0.4703807830810547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,1,power_law_1.2,0.08787199854850769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,1,power_law_1.01,0.7346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,1,power_law_1.2,0.0886784017086029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,1,power_law_1.01,1.4406656265258788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,1,power_law_1.2,0.09165440201759338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,1,power_law_1.2,0.04447999894618988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,1,power_law_1.2,0.09493759870529175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,1,power_law_1.2,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,1,power_law_1.2,0.053625601530075076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,1,power_law_1.2,0.09717760086059571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,1,power_law_1.2,0.03670400083065033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,1,power_law_1.2,0.05504000186920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,1,power_law_1.2,0.036883199214935304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,1,power_law_1.2,0.07854080200195312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,1,power_law_1.2,0.10661120414733886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,1,power_law_1.2,0.0723136007785797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,1,power_law_1.2,0.05738880038261414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,1,power_law_1.2,0.11122560501098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,1,power_law_1.2,0.07082880139350892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,1,power_law_1.2,0.05941759943962097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,1,power_law_1.2,0.06036480069160462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,1,power_law_1.2,0.07095680236816407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,1,power_law_1.2,0.11493120193481446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,1,power_law_1.2,0.06098560094833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,1,power_law_1.2,0.07228800058364868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,1,power_law_1.2,0.06172159910202026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,1,power_law_1.2,0.1381376028060913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,1,power_law_1.2,0.07091839909553528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,1,power_law_1.2,0.06197119951248169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,1,power_law_1.2,0.14368640184402465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,1,power_law_1.2,0.07055360078811646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,1,power_law_1.2,0.06340479850769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,1,power_law_1.2,0.17391999959945678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,1,power_law_1.2,0.07374719977378845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,1,power_law_1.2,0.19689600467681884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,1,power_law_1.2,0.06603519916534424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,1,power_law_1.2,0.0747648000717163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,1,power_law_1.2,0.0770367980003357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,1,power_law_1.2,0.06835200190544129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,1,power_law_1.2,0.26679039001464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,1,power_law_1.2,0.07934079766273498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,1,power_law_1.2,0.07084800004959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,1,power_law_1.2,0.09015679955482483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,1,power_law_1.2,0.317958402633667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,1,power_law_1.2,0.07334399819374085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,1,power_law_1.2,0.09088000059127807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,1,power_law_1.2,0.43851518630981445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,1,power_law_1.2,0.08280320167541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,1,power_law_1.2,0.5426368236541748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,1,power_law_1.2,0.09864320158958435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,1,power_law_1.2,0.0859391987323761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,1,power_law_1.2,0.7509312152862548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,1,power_law_1.2,0.08930559754371643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,1,power_law_1.2,0.1151360034942627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,1,power_law_1.2,0.9790207862854003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,1,power_law_1.2,1.1899904251098632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,1,power_law_1.2,0.10547840595245361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.2,0.12549760341644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,1,power_law_1.2,0.11715840101242066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,1,power_law_1.2,1.8447744369506835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.2,0.15265920162200927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,1,power_law_1.2,0.14829440116882325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,1,power_law_1.2,3.6557632446289063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.2,0.18097920417785646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,1,power_law_1.2,0.17711360454559327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,1,power_law_1.2,0.23117439746856688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.2,0.23946239948272705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,1,power_law_1.2,0.27847039699554443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.2,0.2945087909698486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,1,power_law_1.2,0.3895103931427002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.2,0.40458240509033205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,1,power_law_1.2,0.4955904006958008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.2,0.5135551929473877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,1,power_law_1.2,0.7004415988922119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,1,power_law_1.2,0.9118911743164062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.2,0.7434048175811767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,1,power_law_1.2,1.1180607795715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.2,0.9658111572265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.2,1.184006404876709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,1,power_law_1.2,1.7466367721557616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,1,power_law_1.2,1.8655616760253906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,1,power_law_1.2,3.399769592285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,1,power_law_1.2,3.671641540527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,1,power_law_1.2,0.01897599995136261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,8,balanced,0.08188800017038982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,8,balanced,0.11901866396268208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,8,balanced,0.11574932932853699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,1,power_law_1.2,0.017849600315093993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,8,balanced,0.12592533230781555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,8,balanced,0.11760532855987549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,1,power_law_1.2,0.01809920072555542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,8,balanced,0.08349866668383281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,8,balanced,0.12523200114568075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,8,balanced,0.11619200309117635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,1,power_law_1.2,0.018739199638366698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,8,balanced,0.04788800080617269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,8,balanced,0.08102400104204814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,8,balanced,0.12307199835777283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,8,balanced,0.12453866998354594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,1,power_law_1.2,0.018963199853897095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,8,balanced,0.0865280032157898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,8,balanced,0.1220746636390686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,8,balanced,0.11876799662907918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,1,power_law_1.2,0.01934719979763031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,8,balanced,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,8,balanced,0.08472533027331035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,8,balanced,0.12283200025558472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,8,balanced,0.1200320025285085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,8,balanced,0.08660800258318584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,1,power_law_1.2,0.01950719952583313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,8,balanced,0.12243200341860454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,8,balanced,0.12108266353607178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,8,balanced,0.04746133089065552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,8,balanced,0.08752533793449402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,8,balanced,0.12383466958999634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,8,balanced,0.1178559958934784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,1,power_law_1.2,0.020684799551963805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,8,balanced,0.048351998130480446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,8,balanced,0.08679999907811482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,8,balanced,0.13196800152460733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,8,balanced,0.12164800365765889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,8,balanced,0.08716799815495808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,1,power_law_1.2,0.022988800704479218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,8,balanced,0.04829333225886027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,8,balanced,0.1312373379866282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,8,balanced,0.11854400237401326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,8,balanced,0.08893332878748576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,1,power_law_1.2,0.02863360047340393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,8,balanced,0.05384533107280731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,8,balanced,0.1357973317305247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,8,balanced,0.11563199758529663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,8,balanced,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,1,power_law_1.2,0.024006399512290954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,8,balanced,0.1320480008920034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,8,balanced,0.07502399881680806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,8,balanced,0.10913599530855815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,1,power_law_1.2,0.02540160119533539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,8,balanced,0.13196800152460733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,8,balanced,0.09306666254997253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,8,balanced,0.06781866649786632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,8,balanced,0.10744000474611919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,1,power_law_1.2,0.027935999631881713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,8,balanced,0.14626666903495789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,8,balanced,0.08943999807039897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,8,balanced,0.09060266613960266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,8,balanced,0.1111893355846405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,1,power_law_1.2,0.029395198822021483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,8,balanced,0.1470186710357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,8,balanced,0.08100266754627228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,8,balanced,0.10219200452168782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,1,power_law_1.2,0.031481599807739256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,8,balanced,0.09884799520174663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,8,balanced,0.14773333072662354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,8,balanced,0.0763626645008723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,8,balanced,0.10018666585286458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,1,power_law_1.2,0.034534400701522826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,8,balanced,0.103301336367925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,8,balanced,0.15590932965278625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,8,balanced,0.09514133135477702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,8,balanced,0.14268799622853598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,1,power_law_1.2,0.034860798716545106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,8,balanced,0.10394133130709331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,8,balanced,0.1773866613705953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,8,balanced,0.14055466651916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,1,power_law_1.2,0.045798400044441225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,8,balanced,0.13236266374588013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,8,balanced,0.10909333825111389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,8,balanced,0.18495466311772665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,8,balanced,0.13371200362841287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.2,0.048332801461219786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,8,balanced,0.10162132978439331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,8,balanced,0.13432000080744425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,8,balanced,0.16407466928164163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,8,balanced,0.21887999773025513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.2,0.06220800280570984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,8,balanced,0.14702933033307394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,8,balanced,0.19993066787719727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,8,power_law_1.01,0.11689599752426147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,8,balanced,0.1339306632677714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,8,balanced,0.26708799600601196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.2,0.07670400142669678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,8,balanced,0.12599466244379678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,8,balanced,0.27268799146016437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.2,0.0974016010761261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,8,balanced,0.1627679963906606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,8,balanced,0.33389333883921307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,8,balanced,0.16198399662971497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,8,power_law_1.01,0.1232640027999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.2,0.12083840370178223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,8,balanced,0.34731733798980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,8,balanced,0.21413866678873697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,8,balanced,0.41285332043965656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,8,balanced,0.2013173302014669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,8,power_law_1.01,0.12424319982528687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.2,0.16393599510192872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,8,balanced,0.4783039887746175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,8,power_law_1.01,0.12088320255279542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,8,balanced,0.2752586603164673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.2,0.20719358921051026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,8,balanced,0.2534773349761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,8,balanced,0.5842666625976562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,8,power_law_1.01,0.12024960517883301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,8,balanced,0.6202613512674967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.2,0.2943295955657959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,8,balanced,0.3908160130182902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,8,balanced,0.3107733329137166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,8,power_law_1.01,0.1289728045463562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,8,balanced,0.7318613529205322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.2,0.38207359313964845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,8,power_law_1.01,0.07947520017623902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,8,balanced,0.5210506518681844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,8,balanced,0.8807573318481445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,8,power_law_1.01,0.1568511962890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,8,balanced,0.4312160015106201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.2,0.4705023765563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,8,balanced,1.0362826983133953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,8,power_law_1.01,0.17463040351867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,8,power_law_1.01,0.08007680177688599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,8,balanced,0.744704008102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,1,power_law_1.2,0.734668779373169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,8,balanced,0.5497226715087891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,8,power_law_1.01,0.21288321018218995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,8,balanced,1.1484053134918213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,8,power_law_1.01,0.07969279885292054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,8,balanced,1.4267733891805012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,1,power_law_1.2,1.4404031753540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,8,balanced,0.9902293682098389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,8,power_law_1.01,0.2422271966934204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,8,power_law_1.01,0.08175359964370728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,8,balanced,0.7815413475036621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,8,power_law_1.01,0.15151360034942626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,8,balanced,1.4625813166300456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,8,power_law_1.01,0.0834496021270752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,8,balanced,1.421738624572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,8,balanced,1.7570026715596516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,8,power_law_1.01,0.18280960321426393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,8,power_law_1.01,0.08597760200500489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,8,balanced,1.0475573539733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,8,power_law_1.01,0.17818880081176758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,8,balanced,1.9182292620340984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,8,power_law_1.01,0.09534720182418824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,8,power_law_1.01,0.09477120041847228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,8,power_law_1.01,0.23188478946685792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,8,power_law_1.01,0.10069760084152221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,8,power_law_1.01,0.2690047979354858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,8,balanced,2.225536028544108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,8,balanced,2.7741705576578775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,8,balanced,2.881135940551758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,8,power_law_1.01,0.10101759433746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,8,balanced,1.2764053344726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,8,power_law_1.01,0.27767679691314695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,8,power_law_1.01,0.11391359567642212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,8,power_law_1.01,0.1021888017654419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,8,balanced,3.8031838734944663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,8,power_law_1.01,0.2795072078704834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,8,power_law_1.01,0.12086399793624877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,8,power_law_1.01,0.11189119815826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,8,power_law_1.01,0.3265919923782349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,8,power_law_1.01,0.1188480019569397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,8,power_law_1.01,0.11006079912185669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,8,power_law_1.01,0.43941121101379393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,8,power_law_1.01,0.1219264030456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,8,balanced,2.0170507431030273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,8,balanced,4.739898681640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,8,power_law_1.01,0.12510720491409302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,8,balanced,4.333370526631673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,8,balanced,5.522314707438151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,8,power_law_1.01,0.6176000118255616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,8,power_law_1.01,0.12142720222473144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,8,power_law_1.01,0.16844160556793214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,8,power_law_1.01,0.7200384140014648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,8,power_law_1.01,0.11985280513763427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,8,power_law_1.01,0.15578240156173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,8,power_law_1.01,0.9659328460693359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,8,balanced,7.574917475382487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,8,power_law_1.01,0.11390719413757325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,8,power_law_1.01,0.15838719606399537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,8,power_law_1.01,1.2964608192443847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,8,balanced,3.9266719818115234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,8,power_law_1.01,0.11489919424057007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,8,power_law_1.01,0.23066239356994628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,8,power_law_1.01,2.1120128631591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,8,power_law_1.01,0.1087231993675232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,8,power_law_1.01,0.2733247995376587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,8,power_law_1.01,2.6129344940185546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,8,power_law_1.01,0.10536960363388062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,8,power_law_1.01,0.37334399223327636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,8,power_law_1.01,4.263475036621093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,8,power_law_1.01,0.09390079975128174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,8,power_law_1.01,0.41539840698242186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,8,power_law_1.01,5.68403205871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,8,power_law_1.01,0.1171839952468872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,8,balanced,15.219210306803385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,8,power_law_1.01,0.6636159896850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,8,power_law_1.01,6.94170913696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,8,power_law_1.01,0.1353983998298645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,8,power_law_1.01,0.8199551582336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,8,power_law_1.01,0.12586239576339722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,8,power_law_1.01,9.74400634765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,8,power_law_1.01,1.2393856048583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,8,power_law_1.01,0.1618239998817444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,8,power_law_1.01,1.6775615692138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,8,power_law_1.01,0.1640064001083374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,8,power_law_1.01,2.383238410949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,8,power_law_1.01,20.598361206054687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,8,power_law_1.01,0.18458240032196044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,8,power_law_1.01,3.1042495727539063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,8,power_law_1.01,0.25575039386749265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.01,0.3318527936935425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,8,power_law_1.01,4.155295944213867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.01,0.40899200439453126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,8,power_law_1.01,5.658137512207031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.01,0.49171199798583987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.01,0.8010047912597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,8,power_law_1.01,12.199584197998046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.01,0.9674624443054199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.01,1.3857215881347655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.01,1.969113540649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.01,2.7749311447143556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.01,3.8302398681640626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,8,power_law_1.01,4.5609790802001955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,8,power_law_1.01,7.504576110839844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,8,power_law_1.01,15.013772583007812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,8,power_law_1.01,0.04668160080909729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,8,power_law_1.01,0.046438398957252505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,8,power_law_1.01,0.04519039988517761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,8,power_law_1.01,0.04632959961891174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,8,power_law_1.01,0.04755840003490448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,8,power_law_1.01,0.052102398872375486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,8,power_law_1.01,0.07361279726028443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,8,power_law_1.01,0.06547840237617493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,8,power_law_1.01,0.08822399973869324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,8,power_law_1.01,0.07852799892425537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,8,power_law_1.01,0.07460479736328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,8,power_law_1.01,0.13307520151138305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,8,power_law_1.01,0.1263167977333069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,8,power_law_1.01,0.09920639991760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,8,power_law_1.01,0.14067200422286988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,8,power_law_1.01,0.12433919906616211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,8,power_law_1.01,0.16108160018920897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,1,8,power_law_1.2,0.08077440261840821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,1,8,power_law_1.2,0.11668479442596436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,8,power_law_1.01,0.21087999343872071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,1,8,power_law_1.2,0.12315520048141479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,1,8,power_law_1.2,0.11539839506149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.01,0.273036789894104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,1,8,power_law_1.2,0.08208640217781067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,1,8,power_law_1.2,0.04688639938831329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,1,8,power_law_1.2,0.12410880327224731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,4,balanced,0.09832533200581868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,4,balanced,0.053930665055910744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,4,balanced,0.04922133187452952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,1,8,power_law_1.2,0.12113920450210572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,1,8,power_law_1.2,0.12235519886016846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.01,0.3884160041809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,1,8,power_law_1.2,0.08001919984817504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,1,8,power_law_1.2,0.047142401337623596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,4,balanced,0.08262933293978374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,4,balanced,0.055786664287249245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,1,8,power_law_1.2,0.12102400064468384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,4,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,1,8,power_law_1.2,0.12969599962234496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,1,8,power_law_1.2,0.1202623963356018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,1,8,power_law_1.2,0.04538240134716034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,4,balanced,0.07686399916807811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.01,0.5189311981201172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,4,balanced,0.1297813355922699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,4,balanced,0.08205333352088928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,1,8,power_law_1.2,0.12158080339431762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,4,balanced,0.06427200138568878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,1,8,power_law_1.2,0.15648640394210817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,1,8,power_law_1.2,0.08377599716186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,1,8,power_law_1.2,0.04672000110149384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,4,balanced,0.12140799562136333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.01,0.7451263904571533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,4,balanced,0.07980800171693166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,1,8,power_law_1.2,0.12463359832763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,4,balanced,0.06879466772079468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,1,8,power_law_1.2,0.1987264037132263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,1,8,power_law_1.2,0.08422399759292602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,1,8,power_law_1.2,0.04749439954757691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,4,balanced,0.1223466694355011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,4,balanced,0.08211199939250946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.01,0.9903615951538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,1,8,power_law_1.2,0.12047359943389893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,4,balanced,0.06759466727574666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,1,8,power_law_1.2,0.22855041027069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,4,balanced,0.08282666901747386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,1,8,power_law_1.2,0.0522816002368927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,1,8,power_law_1.2,0.08949120044708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,4,balanced,0.12199466427167256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,1,8,power_law_1.2,0.11683199405670167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.01,1.4551936149597169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,4,balanced,0.06924800078074138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,1,8,power_law_1.2,0.24414079189300536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,4,balanced,0.08376000324885051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,1,8,power_law_1.2,0.07292799949645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,4,balanced,0.07030933101971944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.01,1.9174720764160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,1,8,power_law_1.2,0.11043200492858887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,1,8,power_law_1.2,0.09335039854049683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,4,balanced,0.0851200024286906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,1,8,power_law_1.2,0.15934079885482788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,4,balanced,0.12285332878430684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,4,balanced,0.07034666836261749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,1,8,power_law_1.2,0.1784000039100647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,4,balanced,0.08713066577911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,1,8,power_law_1.2,0.06556159853935242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,1,8,power_law_1.2,0.11045759916305542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,4,balanced,0.1258026659488678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,1,8,power_law_1.2,0.18538880348205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.01,2.8780096054077147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,4,balanced,0.06963733335336049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,4,balanced,0.08717333277066548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,1,8,power_law_1.2,0.08826879858970642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,1,8,power_law_1.2,0.0985535979270935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,1,8,power_law_1.2,0.10049279928207397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,4,balanced,0.12734400232632956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,1,8,power_law_1.2,0.2205440044403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,4,balanced,0.07067733506361644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,4,balanced,0.08886933326721191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,1,8,power_law_1.2,0.07932159900665284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.01,3.805894470214844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,1,8,power_law_1.2,0.10458240509033204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,1,8,power_law_1.2,0.09944319725036621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,4,balanced,0.12635200222333273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,1,8,power_law_1.2,0.2878976106643677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,4,balanced,0.0703306645154953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,1,8,power_law_1.2,0.10298880338668823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,4,balanced,0.12989333271980286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,4,balanced,0.08815466364224751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,1,8,power_law_1.2,0.0752128005027771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,8,power_law_1.01,4.734220886230469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,1,8,power_law_1.2,0.12490880489349365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,1,8,power_law_1.2,0.28499200344085696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,4,balanced,0.0720000018676122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,1,8,power_law_1.2,0.10249600410461426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,4,balanced,0.12921599547068277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,4,balanced,0.09295466542243958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,1,8,power_law_1.2,0.13564159870147705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,8,power_law_1.01,7.5868476867675785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,1,8,power_law_1.2,0.14468480348587037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,1,8,power_law_1.2,0.28000640869140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,4,balanced,0.07027733325958252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,1,8,power_law_1.2,0.11046400070190429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,4,balanced,0.13036266962687174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,4,balanced,0.10168000062306722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,8,power_law_1.01,15.149087524414062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,1,8,power_law_1.2,0.131660795211792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,1,8,power_law_1.2,0.3501440048217773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,1,8,power_law_1.2,0.11866879463195801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,4,balanced,0.0747626672188441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,1,8,power_law_1.2,0.11268479824066162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,4,balanced,0.13693867127100626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,4,balanced,0.1019040048122406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,1,8,power_law_1.2,0.09997439980506898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,1,8,power_law_1.2,0.4510848045349121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,1,8,power_law_1.2,0.16152960062026978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,4,balanced,0.08426666259765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,1,8,power_law_1.2,0.12803200483322144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,4,balanced,0.14697066942850748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,4,balanced,0.10508267084757487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,1,8,power_law_1.2,0.14328320026397706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,4,balanced,0.14988266428311667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,1,8,power_law_1.2,0.6536320209503174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,1,8,power_law_1.2,0.15140479803085327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,4,balanced,0.08268799881140391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,1,8,power_law_1.2,0.16609280109405516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,4,balanced,0.14644799629847208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,4,balanced,0.13432533542315164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,1,8,power_law_1.2,0.12425600290298462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,1,8,power_law_1.2,0.7050432205200196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,1,8,power_law_1.2,0.17612799406051635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,4,balanced,0.08894933263460796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,1,8,power_law_1.2,0.17140480279922485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,4,balanced,0.19203199942906699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,1,8,power_law_1.2,1.0457792282104492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,4,balanced,0.13612799843152365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,1,8,power_law_1.2,0.1616960048675537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,4,balanced,0.10335466265678406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,1,8,power_law_1.2,1.374847984313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,1,8,power_law_1.2,0.2376703977584839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,1,8,power_law_1.2,0.15946880578994752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,4,balanced,0.1787466605504354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,4,balanced,0.1707786719004313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,1,8,power_law_1.2,0.2106623888015747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,1,8,power_law_1.2,2.217670440673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,4,balanced,0.10758933424949646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.2,0.3149696111679077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,1,8,power_law_1.2,0.21400320529937744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.2,0.2737407922744751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,4,balanced,0.24572267134984335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,1,8,power_law_1.2,2.666988754272461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,4,balanced,0.19413334131240845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.2,0.426470422744751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,4,balanced,0.1241439978281657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.2,0.38766720294952395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,1,8,power_law_1.2,0.27407360076904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,1,8,power_law_1.2,4.050239944458008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,4,balanced,0.26310400168100995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,4,balanced,0.14617066582043967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.2,0.5534080028533935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,4,balanced,0.2558506727218628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.2,0.5196671962738038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,1,8,power_law_1.2,0.3337088108062744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,1,8,power_law_1.2,5.923283386230469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,4,balanced,0.33905065059661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.2,0.7439743995666503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,4,balanced,0.18609066804250082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.2,0.7838912010192871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,1,8,power_law_1.2,7.574540710449218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,4,balanced,0.30869332949320477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,1,8,power_law_1.2,0.438643217086792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.2,0.9902015686035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,4,balanced,0.4038613239924113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,4,balanced,0.22664000590642294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.2,0.9082752227783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.2,1.4553279876708984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,1,8,power_law_1.2,9.229926300048827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.2,1.4590335845947267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,1,8,power_law_1.2,0.6547584056854248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,4,balanced,0.43389864762624103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,4,balanced,0.2990666627883911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,4,balanced,0.567354679107666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.2,1.9181312561035155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.2,1.978278350830078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,1,8,power_law_1.2,0.8241087913513183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,1,8,power_law_1.2,24.66324462890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,4,balanced,0.5533759991327921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.2,2.8764671325683593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.2,2.8265151977539062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,1,8,power_law_1.2,1.280627155303955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,4,balanced,0.3818560043970744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,4,balanced,0.7494346300760905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.2,3.8060672760009764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.2,3.8207233428955076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,1,8,power_law_1.2,1.598259162902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,4,balanced,0.789189338684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,1,8,power_law_1.2,2.3994943618774416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,1,8,power_law_1.2,4.733523178100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,1,8,power_law_1.2,4.768582534790039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,4,balanced,0.527184009552002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,4,balanced,1.0597600142161052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,1,8,power_law_1.2,3.3535232543945312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,4,balanced,1.0326080322265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,1,8,power_law_1.2,6.196115112304687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,1,8,power_law_1.2,7.594560241699218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,1,8,power_law_1.2,4.1415550231933596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,4,balanced,0.6945813496907552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,1,8,power_law_1.2,15.197407531738282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,1,8,power_law_1.2,15.206329345703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,1,8,power_law_1.2,6.120857620239258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,4,balanced,1.4266187349955242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,4,balanced,1.2638506889343262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,4,balanced,0.8405653635660807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,1,8,power_law_1.2,12.494931030273438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,4,balanced,1.8084640502929688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,4,balanced,1.980826695760091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,4,balanced,1.3252320289611816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,4,balanced,2.7968320846557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,4,balanced,3.8612213134765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,4,balanced,2.6521387100219727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,4,balanced,5.509093602498372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,4,balanced,0.02736533433198929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,4,balanced,0.0272533322374026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,4,balanced,0.025450666745503742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,4,balanced,0.03143466760714849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,4,balanced,0.03189333279927572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,4,balanced,0.03396799912055334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,4,balanced,0.04841599861780802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,4,balanced,0.03945599993069967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,4,balanced,0.08243733147780101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,4,balanced,0.08085866769154866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,4,balanced,0.08179200192292531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,4,balanced,0.08636266986529033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,4,balanced,0.08169599870840709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,4,balanced,0.06227200229962667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,4,balanced,0.09146133065223694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,4,balanced,0.07467199862003326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,4,balanced,0.09539199868837993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,4,balanced,0.11974400281906128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,4,balanced,0.14502933621406555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,4,balanced,0.20363734165827432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,4,balanced,0.2592639923095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,4,balanced,0.36139198144276935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,4,balanced,0.47044801712036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,4,balanced,0.6803733507792155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,4,balanced,0.8888426621754965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,4,balanced,1.3069813251495361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,4,power_law_1.01,0.12383999824523925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,4,balanced,1.727952003479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,4,power_law_1.01,0.12773120403289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,4,power_law_1.01,0.09740800261497498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,4,power_law_1.01,0.11973760128021241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,4,power_law_1.01,0.11995520591735839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,4,balanced,2.1490186055501304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,4,power_law_1.01,0.1253119945526123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,4,power_law_1.01,0.13999999761581422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,4,power_law_1.01,0.12937599420547485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,4,power_law_1.01,0.13793280124664306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,4,power_law_1.01,0.1414528012275696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,4,power_law_1.01,0.13238400220870972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,4,power_law_1.01,0.13776639699935914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,4,balanced,3.405061403910319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,4,power_law_1.01,0.14576640129089355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,4,power_law_1.01,0.1543936014175415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,4,power_law_1.01,0.17870080471038818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,4,power_law_1.01,0.20257918834686278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,4,power_law_1.01,0.20118401050567628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,4,power_law_1.01,0.25502719879150393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,4,power_law_1.01,0.06369280219078063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,4,power_law_1.01,0.32870399951934814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,4,power_law_1.01,0.4311488151550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,4,power_law_1.01,0.06804479956626892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,4,power_law_1.01,0.027059200406074523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,4,power_law_1.01,0.058323198556900026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,4,power_law_1.01,0.5708928108215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,4,power_law_1.01,0.8750144004821777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,4,power_law_1.01,0.07608960270881653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,4,power_law_1.01,0.025964799523353576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,4,power_law_1.01,0.8630975723266602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,4,power_law_1.01,0.08030719757080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,4,power_law_1.01,0.06517120003700257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,4,power_law_1.01,1.5373056411743165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,4,power_law_1.01,0.06293759942054748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,4,power_law_1.01,0.025017601251602174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,4,power_law_1.01,0.07982079982757569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,4,balanced,6.747664133707683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,4,power_law_1.01,1.8920896530151368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,4,power_law_1.01,0.03073279857635498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,4,power_law_1.01,0.08220800161361694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,4,power_law_1.01,0.06709120273590088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,4,power_law_1.01,0.03123840093612671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,4,power_law_1.01,0.08750079870223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,4,power_law_1.01,3.126905632019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,4,power_law_1.01,0.03219839930534363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,4,power_law_1.01,0.08792319893836975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,4,power_law_1.01,0.06739839911460876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,4,power_law_1.01,0.04503040015697479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,4,power_law_1.01,4.300755310058594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,4,power_law_1.01,0.09426559805870056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,4,power_law_1.01,0.037427198886871335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,4,power_law_1.01,0.10129280090332031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,4,power_law_1.01,0.0689791977405548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,4,power_law_1.2,0.13244160413742065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,4,power_law_1.01,0.08154879808425904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,4,power_law_1.01,0.08914560079574585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,4,power_law_1.01,5.554739379882813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,4,power_law_1.01,0.06895359754562377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,4,power_law_1.01,0.07939199805259704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,4,power_law_1.01,0.09696000218391418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,4,power_law_1.2,0.07214080095291138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,4,power_law_1.2,0.117740797996521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,4,power_law_1.01,0.08001919984817504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,4,power_law_1.2,0.10645120143890381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,4,power_law_1.01,8.833395385742188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,4,power_law_1.01,0.08562560081481933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,4,power_law_1.01,0.10021760463714599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,4,power_law_1.2,0.06208639740943909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,4,power_law_1.01,0.06914560198783874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,4,power_law_1.2,0.11973760128021241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,4,power_law_1.01,0.07761920094490052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,4,power_law_1.2,0.12017279863357544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,4,power_law_1.2,0.06817280054092408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,4,power_law_1.01,0.060819202661514284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,4,power_law_1.01,0.10539519786834717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,4,power_law_1.2,0.12704000473022461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,4,power_law_1.2,0.07854080200195312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,4,power_law_1.01,0.07031679749488831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,4,power_law_1.01,0.0877951979637146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,4,power_law_1.01,17.321318054199217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,4,power_law_1.2,0.14625920057296754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,4,power_law_1.2,0.07975680232048035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,4,power_law_1.01,0.0726144015789032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,4,power_law_1.01,0.11946879625320435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,4,power_law_1.2,0.12968319654464722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,4,power_law_1.01,0.07127040028572082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,4,power_law_1.01,0.07420160174369812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,4,power_law_1.2,0.13825279474258423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,4,power_law_1.2,0.0841152012348175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,4,power_law_1.01,0.09379839897155762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,4,power_law_1.01,0.12835839986801148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,4,power_law_1.2,0.1436352014541626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,4,power_law_1.01,0.07875840067863464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,4,power_law_1.01,0.1331455945968628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,4,power_law_1.2,0.09018239974975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,4,power_law_1.01,0.11731840372085571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,4,power_law_1.2,0.13423360586166383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,4,power_law_1.01,0.07963520288467407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,4,power_law_1.2,0.1610751986503601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,4,power_law_1.01,0.1950719952583313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.01,0.14241280555725097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,4,power_law_1.01,0.08320000171661376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,4,power_law_1.2,0.08893439769744874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,4,power_law_1.2,0.15166079998016357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,4,power_law_1.01,0.22606079578399657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,4,power_law_1.01,0.09077759981155395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,4,power_law_1.2,0.08956159949302674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,4,power_law_1.2,0.15483520030975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,4,power_law_1.01,0.09567999839782715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,4,power_law_1.2,0.09379199743270875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,4,power_law_1.01,0.2720256090164185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.01,0.20035200119018554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,4,power_law_1.2,0.19575040340423583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,4,power_law_1.2,0.06373119950294495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,4,power_law_1.2,0.18169599771499634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,4,power_law_1.01,0.32009599208831785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,4,power_law_1.01,0.11375999450683594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.01,0.25630080699920654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,4,power_law_1.2,0.22828800678253175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,4,power_law_1.2,0.09157760143280029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,4,power_law_1.01,0.43598079681396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,4,power_law_1.01,0.13621759414672852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,4,power_law_1.2,0.26202239990234377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,4,power_law_1.2,0.057094401121139525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.01,0.3579776048660278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.01,0.1388800024986267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,4,power_law_1.2,0.3520128011703491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,4,power_law_1.01,0.5841919898986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,4,power_law_1.2,0.09790719747543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.01,0.19479680061340332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.01,0.47910399436950685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,4,power_law_1.2,0.5002624034881592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,4,power_law_1.2,0.057011198997497556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,4,power_law_1.01,0.971891212463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,4,power_law_1.2,0.10501760244369507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,4,power_law_1.2,0.5234367847442627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.01,0.2608511924743652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.01,0.6758975982666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,4,power_law_1.01,1.1031935691833497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,4,power_law_1.2,0.9070719718933106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.01,0.33041279315948485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,4,power_law_1.2,0.0662335991859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.01,0.8913984298706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,4,power_law_1.2,0.10831359624862671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.01,0.379040002822876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,4,power_law_1.2,1.0831680297851562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,4,power_law_1.2,0.0667136013507843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,4,power_law_1.2,0.11560319662094116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,4,power_law_1.01,1.900057601928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.01,0.5599999904632569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,4,power_law_1.2,1.5347519874572755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.01,1.3090815544128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,4,power_law_1.2,0.13567999601364136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,4,power_law_1.2,0.06764159798622131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,4,power_law_1.2,1.643129539489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.01,0.8582079887390137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.01,1.729439926147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,4,power_law_1.01,2.5100095748901365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,4,power_law_1.2,0.15344640016555786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,4,power_law_1.2,0.06871039867401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.01,1.1370880126953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,4,power_law_1.2,3.366515350341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,4,power_law_1.01,2.1456960678100585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,4,power_law_1.2,0.17700480222702025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,4,power_law_1.2,0.06852480173110961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,4,power_law_1.01,2.7038400650024412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.01,1.496217632293701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,4,power_law_1.2,4.220204925537109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,4,power_law_1.2,0.22303359508514403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,4,power_law_1.2,0.06570879817008972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,4,power_law_1.01,3.4002750396728514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,4,power_law_1.2,0.071424001455307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,4,power_law_1.2,0.295033597946167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,4,power_law_1.01,3.3073665618896486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,4,power_law_1.01,2.229267120361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,4,power_law_1.2,5.398220825195312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,4,power_law_1.2,0.07197440266609192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,4,power_law_1.2,0.3800575971603394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,4,power_law_1.01,2.7227712631225587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,4,power_law_1.2,0.07930240035057068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,4,power_law_1.01,6.749254608154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,4,power_law_1.2,7.957369232177735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,4,power_law_1.2,0.4685567855834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,4,power_law_1.01,8.91712646484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,4,power_law_1.01,7.751596832275391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,4,power_law_1.2,0.08123520016670227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,4,power_law_1.2,0.59617919921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,4,power_law_1.2,0.08087040185928344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,4,power_law_1.2,1.1521471977233886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,4,power_law_1.2,18.06681671142578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,4,power_law_1.2,0.09648640155792236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,4,power_law_1.2,1.2236160278320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,4,power_law_1.2,0.09603840112686157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,4,power_law_1.2,1.9497535705566407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,4,power_law_1.2,0.1096127986907959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,4,power_law_1.2,2.2174655914306642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,4,power_law_1.2,0.1309183955192566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.2,0.14515199661254882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,4,power_law_1.2,3.1335168838500977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.2,0.18838399648666382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,4,power_law_1.2,4.959052658081054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.2,0.2453376054763794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,4,power_law_1.2,10.425318145751953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.2,0.34279038906097414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.2,0.45296640396118165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.2,0.596563196182251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.2,0.9663616180419922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.2,1.4751935958862306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.2,1.9267135620117188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,4,power_law_1.2,2.125516891479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,4,power_law_1.2,3.7094974517822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,4,power_law_1.2,7.3578239440917965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,4,power_law_1.2,0.027084800601005554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,4,power_law_1.2,0.025811201333999632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,4,power_law_1.2,0.024940800666809083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,4,power_law_1.2,0.03036159873008728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,4,power_law_1.2,0.03153280019760132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,4,power_law_1.2,0.032153600454330446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,4,power_law_1.2,0.04564479887485504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,4,power_law_1.2,0.037868800759315493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,4,power_law_1.2,0.08103039860725403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,4,power_law_1.2,0.07950080037117005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,4,power_law_1.2,0.0796992003917694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,4,power_law_1.2,0.08498560190200806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,4,power_law_1.2,0.07945600152015686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,4,power_law_1.2,0.060864001512527466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,4,power_law_1.2,0.08784639835357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,2,balanced,0.05633600056171417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,4,power_law_1.2,0.07396479845046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,4,power_law_1.2,0.09431679844856262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,2,balanced,0.07999999821186066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,2,balanced,0.12273599704106648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,4,power_law_1.2,0.11683199405670167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,2,balanced,0.12353600064913432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.2,0.1428928017616272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,2,balanced,0.12289067109425862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,2,balanced,0.12341866890589397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.2,0.20127999782562256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,2,balanced,0.12356266379356384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.2,0.25649280548095704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,2,balanced,0.1251040001710256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,2,balanced,0.12572800119717917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.2,0.3579391956329346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,2,balanced,0.12435199817021687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.2,0.46832637786865233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,2,balanced,0.1268160045146942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.2,0.6771008014678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,2,balanced,0.12611200412114462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.2,0.8866623878479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,2,balanced,0.1278986632823944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.2,1.308460807800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,2,balanced,0.1325866679350535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.2,1.724127960205078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,2,balanced,0.1444586714108785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,4,power_law_1.2,2.146611213684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,2,balanced,0.14965867002805075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,2,balanced,0.14621866742769876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,4,power_law_1.2,3.397747039794922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,2,balanced,0.18347734212875366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,2,balanced,0.1844586730003357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,2,balanced,0.228928009668986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,4,power_law_1.2,6.747058868408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,2,balanced,0.26585066318511963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,2,balanced,0.3490826686223348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,2,balanced,0.43534398078918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,2,balanced,0.6133919954299927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,2,balanced,0.7738719781239828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,2,balanced,1.1053333282470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,2,balanced,1.4386666615804036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,2,balanced,0.03583999971548716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,2,balanced,0.054666668176651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,2,balanced,0.02250133454799652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,2,balanced,1.7724266052246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,2,balanced,0.05398400127887726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,2,power_law_1.01,0.07523199915885925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,2,balanced,0.07542400062084198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,2,balanced,0.021717332303524017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,2,balanced,0.023183998962243397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,2,balanced,0.10843732953071594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,2,balanced,0.07589866717656453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,2,power_law_1.01,0.046911999583244324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,2,power_law_1.01,0.06756479740142822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,2,balanced,0.021514666577180225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,2,balanced,0.11085866888364156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,2,balanced,0.0790826678276062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,2,power_law_1.01,0.08056319952011108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,2,balanced,0.02146666745344798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,2,balanced,0.10859733819961548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,2,power_law_1.01,0.11743359565734864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,2,balanced,0.07867733140786488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,2,balanced,0.025306666890780132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,2,balanced,0.1066986620426178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,2,power_law_1.01,0.11538560390472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,2,power_law_1.01,0.055155199766159055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,2,balanced,0.08037866652011871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,2,balanced,0.03141333411137263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,2,power_law_1.01,0.12180479764938354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,2,balanced,0.11156800389289856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,2,balanced,2.798010508219401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,2,power_law_1.01,0.14270080327987672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,2,balanced,0.0817493349313736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,2,balanced,0.027658666173617046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,2,balanced,0.11794666449228923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,2,power_law_1.01,0.12444800138473511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,2,balanced,0.08136533200740814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,2,balanced,0.033786666889985405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,2,balanced,0.10762666662534077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,2,power_law_1.01,0.06002560257911682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,2,power_law_1.01,0.12978559732437134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,2,balanced,0.08096533517042796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,2,power_law_1.01,0.12872960567474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,2,balanced,0.0317493329445521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,2,balanced,0.10436800122261047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,2,balanced,0.08318933347860973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,2,power_law_1.01,0.13222399950027466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,2,power_law_1.01,0.07605760097503662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,2,balanced,0.09885866443316142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,2,balanced,0.03578133384386698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,2,balanced,0.08514133095741272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,2,power_law_1.01,0.06992639899253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,2,power_law_1.01,0.13583999872207642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,2,balanced,0.08682133754094441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,2,balanced,0.10605333248774211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,2,power_law_1.01,0.14117759466171265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,2,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,2,power_law_1.01,0.076883202791214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,2,balanced,0.08554133772850037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,2,power_law_1.01,0.15426559448242189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,2,power_law_1.01,0.0689791977405548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,2,balanced,0.09717333316802979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,2,balanced,0.0399893323580424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,2,power_law_1.01,0.18369920253753663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,2,balanced,0.09960533181826274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,2,balanced,0.09307199716567993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,2,balanced,0.04148799926042557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,2,power_law_1.01,0.0843392014503479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,2,power_law_1.01,0.17998720407485963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,2,power_law_1.01,0.07923840284347534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,2,balanced,0.09291733304659526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,2,balanced,0.09925333658854167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,2,power_law_1.01,0.18420480489730834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,2,balanced,0.06173333525657654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,2,power_law_1.01,0.09615359902381897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,2,power_law_1.01,0.08067200183868409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,2,power_law_1.01,0.214355206489563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,2,balanced,0.09955733021100362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,2,balanced,0.10052800178527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,2,power_law_1.01,0.10519039630889893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,2,power_law_1.01,0.0810368001461029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,2,balanced,0.05413866539796194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,2,power_law_1.01,0.2431488037109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,2,balanced,0.09330667058626811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,2,balanced,0.10540266831715901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,2,power_law_1.01,0.08134400248527526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,2,power_law_1.01,0.10849920511245728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,2,power_law_1.01,0.29847040176391604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,2,balanced,5.572367986043294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,2,balanced,0.14621333281199136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,2,balanced,0.06010666489601135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,2,power_law_1.01,0.10753920078277587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,2,power_law_1.01,0.08539519906044006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,2,balanced,0.12617599964141846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,2,power_law_1.01,0.36897280216217043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,2,power_law_1.01,0.1061568021774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,2,balanced,0.12936000029246011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,2,balanced,0.07374399900436401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,2,power_law_1.01,0.08652799725532531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,2,balanced,0.13573867082595825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,2,power_law_1.01,0.10479999780654907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,2,power_law_1.01,0.5318784236907959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,2,power_law_1.01,0.08912640213966369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,2,balanced,0.09210667014122009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,2,balanced,0.16120533148447672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,2,balanced,0.16636799772580466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,2,power_law_1.01,0.10092799663543701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,2,power_law_1.01,0.09349759817123413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,2,power_law_1.01,0.6652416229248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,2,power_law_1.01,0.1048192024230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,2,balanced,0.12877866625785828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,2,power_law_1.01,0.09778559803962708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,2,power_law_1.01,0.9039615631103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,2,balanced,0.1973066727320353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,2,balanced,0.19542400042215982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,2,power_law_1.01,0.11000959873199463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,2,power_law_1.01,1.3216896057128906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,2,power_law_1.01,0.1057919979095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,2,power_law_1.01,0.11743359565734864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,2,balanced,0.1530026694138845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,2,balanced,0.2611520091692607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,2,balanced,0.26507200797398883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,2,power_law_1.01,0.11239039897918701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,2,power_law_1.01,0.11125119924545288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,2,power_law_1.01,1.4535488128662108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,2,power_law_1.01,0.1323456048965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,2,balanced,0.21387199560801187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,2,power_law_1.01,0.11957759857177734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,2,balanced,0.31992000341415405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,2,power_law_1.01,2.5144895553588866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,2,power_law_1.01,0.1300287961959839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,2,balanced,0.334666649500529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,2,power_law_1.01,0.14375679492950438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,2,power_law_1.01,0.1503999948501587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,2,power_law_1.01,2.8891008377075194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,2,power_law_1.01,0.16168320178985596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,2,balanced,0.2698240081469218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,2,power_law_1.01,0.163481605052948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,2,balanced,0.44493865966796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,2,balanced,0.47089600563049316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.01,0.1837504029273987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,2,power_law_1.01,4.346758270263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,2,power_law_1.01,0.20641920566558838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.01,0.2354048013687134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,2,balanced,0.38598398367563885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.01,0.2774208068847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,2,power_law_1.01,0.25426559448242186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.01,0.3520319938659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,2,power_law_1.01,0.36910080909729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,2,power_law_1.01,9.020333099365235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,2,balanced,0.5637226502100626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,2,balanced,0.6096213261286417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,2,power_law_1.01,0.43239679336547854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.01,0.49706239700317384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,2,power_law_1.01,0.6874752044677734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.01,0.7625599861145019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,2,balanced,0.4989386796951294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.01,0.9159168243408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,2,power_law_1.01,0.7835008144378662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,2,balanced,0.8062613010406494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.01,1.3460415840148925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,2,power_law_1.01,1.104742431640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,2,balanced,0.8823413054148356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.01,1.8009599685668944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,2,power_law_1.01,1.5340607643127442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,2,balanced,0.7330186367034912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,2,power_law_1.01,2.0657920837402344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,2,power_law_1.01,2.019251251220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,2,power_law_1.01,3.3383678436279296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,2,power_law_1.01,2.7599231719970705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,2,balanced,1.0487840175628662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,2,balanced,1.1625280380249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,2,power_law_1.01,6.79925765991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,2,power_law_1.01,5.817382431030273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,2,balanced,0.9636533260345459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,2,balanced,1.290064016977946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,2,balanced,1.432085355122884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,2,balanced,1.189855972925822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,2,balanced,2.020319938659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,2,balanced,2.2692532539367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,2,balanced,1.8842453956604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,2,power_law_1.01,0.02195200026035309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,2,balanced,3.928245226542155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,2,balanced,4.528842608133952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,2,power_law_1.01,0.021452799439430237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,2,power_law_1.01,0.02109439969062805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,2,balanced,3.722106615702311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,2,power_law_1.01,0.020716799795627593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,2,power_law_1.01,0.021568000316619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,2,power_law_1.01,0.02290560007095337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,2,power_law_1.01,0.03059200048446655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,2,power_law_1.01,0.02593280076980591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,2,power_law_1.01,0.03314560055732727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,2,power_law_1.01,0.030476799607276915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,2,power_law_1.01,0.03473919928073883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,2,power_law_1.01,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,2,power_law_1.01,0.039724799990653994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,2,power_law_1.01,0.04078719913959503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,2,power_law_1.01,0.061343997716903687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,2,power_law_1.01,0.052313601970672606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,2,power_law_1.01,0.059462398290634155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,2,power_law_1.01,0.07303040027618408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.01,0.09056640267372132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.01,0.1254271984100342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.01,0.15066239833831788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.01,0.2108288049697876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.01,0.26622719764709474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.01,0.3821120023727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.01,0.49825282096862794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.01,0.7298304080963135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.01,0.9579839706420898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,2,power_law_1.01,1.189356803894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,2,power_law_1.01,1.8797887802124023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,2,power_law_1.01,3.7184513092041014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,2,power_law_1.2,0.0637440025806427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,2,power_law_1.2,0.08023679852485657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,2,power_law_1.2,0.07621759772300721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,2,power_law_1.2,0.11831680536270142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,2,power_law_1.2,0.1204416036605835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,2,power_law_1.2,0.1254207968711853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,2,power_law_1.2,0.14600319862365724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,2,power_law_1.2,0.12505600452423096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,2,power_law_1.2,0.12597119808197021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,2,power_law_1.2,0.1331264019012451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,2,power_law_1.2,0.1351359963417053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,2,power_law_1.2,0.13779840469360352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,2,power_law_1.2,0.140774405002594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,2,power_law_1.2,0.15406719446182252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,2,power_law_1.2,0.1746240019798279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,2,power_law_1.2,0.04856959879398346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,2,power_law_1.2,0.1897663950920105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,2,power_law_1.2,0.05492479801177978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,2,power_law_1.2,0.1795199990272522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,2,power_law_1.2,0.22089600563049316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,2,power_law_1.2,0.05976960062980652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,2,power_law_1.2,0.23427200317382812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,2,power_law_1.2,0.0756928026676178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,2,power_law_1.2,0.2957632064819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,2,power_law_1.2,0.07762560248374939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,2,power_law_1.2,0.3701247930526733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,2,power_law_1.2,0.0794368028640747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,2,power_law_1.2,0.5492159843444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,2,power_law_1.2,0.08332160115242004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,2,power_law_1.2,0.7032896041870117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,2,power_law_1.2,0.0812991976737976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,2,power_law_1.2,1.084275245666504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,2,power_law_1.2,0.08463360071182251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,2,power_law_1.2,0.0877951979637146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,2,power_law_1.2,1.3117823600769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,2,power_law_1.2,0.08772479891777038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,2,power_law_1.2,1.801356887817383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,2,power_law_1.2,0.09198079705238342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,2,power_law_1.2,2.3722560882568358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,2,power_law_1.2,0.09242240190505982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,2,power_law_1.2,0.10060800313949585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,2,power_law_1.2,3.057484817504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,2,power_law_1.2,0.10803200006484985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,2,power_law_1.2,4.524319839477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,2,power_law_1.2,0.11260800361633301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,2,power_law_1.2,0.12015360593795776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,2,power_law_1.2,0.06889600157737732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,2,power_law_1.2,0.02136320024728775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,2,power_law_1.2,0.1471295952796936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,2,power_law_1.2,8.557241821289063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,2,power_law_1.2,0.02094080001115799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,2,power_law_1.2,0.07273600101470948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,2,power_law_1.2,0.17390719652175904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,2,power_law_1.2,0.02136320024728775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,2,power_law_1.2,0.08144000172615051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,2,power_law_1.2,0.2155519962310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,2,power_law_1.2,0.02082560062408447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,2,power_law_1.2,0.020979200303554536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,2,power_law_1.2,0.10261759757995606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,2,power_law_1.2,0.27745919227600097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,2,power_law_1.2,0.022643199563026427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,2,power_law_1.2,0.3665663957595825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,2,power_law_1.2,0.10280959606170655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,2,power_law_1.2,0.029836800694465638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,2,power_law_1.2,0.10800000429153442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,2,power_law_1.2,0.41429758071899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,2,power_law_1.2,0.0265855997800827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,2,power_law_1.2,0.03500800132751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,2,power_law_1.2,0.10314879417419434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,2,power_law_1.2,0.5493951797485351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,2,power_law_1.2,0.030943998694419862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,2,power_law_1.2,0.10752639770507813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,2,power_law_1.2,0.6852159976959229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,2,power_law_1.2,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,2,power_law_1.2,0.10375679731369018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,2,power_law_1.2,1.3030400276184082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,2,power_law_1.2,0.03792639970779419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,2,power_law_1.2,0.09895039796829223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,2,power_law_1.2,0.039155200123786926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,2,power_law_1.2,1.3550784111022949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,2,power_law_1.2,0.04038400053977966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,2,power_law_1.2,0.10075520277023316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,2,power_law_1.2,0.0631168007850647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,2,power_law_1.2,1.8405632019042968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,2,power_law_1.2,0.10918400287628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,2,power_law_1.2,0.052857601642608644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,2,power_law_1.2,0.11074559688568116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,2,power_law_1.2,0.058924800157546996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,2,power_law_1.2,3.0109376907348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,2,power_law_1.2,0.11208959817886352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,2,power_law_1.2,0.072326397895813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,2,power_law_1.2,0.12312959432601929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.2,0.09026560187339783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,2,power_law_1.2,6.016543960571289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.2,0.12448639869689941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,2,power_law_1.2,0.12948479652404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,1,balanced,0.051872000098228455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.2,0.15018240213394166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,2,power_law_1.2,0.1371072053909302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.2,0.21134719848632813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,1,balanced,0.07435200115044911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,2,power_law_1.2,0.15768959522247314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,1,balanced,0.1183573305606842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.2,0.26654720306396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.2,0.19052799940109252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,1,balanced,0.03867733230193456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,1,balanced,0.035562666753927864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,1,balanced,0.12001066406567891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.2,0.38537600040435793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.2,0.26984319686889646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,1,balanced,0.12235732873280843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.2,0.2906752109527588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,1,balanced,0.041946664452552795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.2,0.5011392116546631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,1,balanced,0.12142399946848552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,1,balanced,0.05436266462008158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.2,0.41889281272888185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.2,0.7298687934875489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,1,balanced,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,1,balanced,0.122789333264033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.2,0.48465919494628906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.2,0.958182430267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,1,balanced,0.12315199772516887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,1,balanced,0.07645866771539052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,1,balanced,0.06547733147939046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.2,0.7235455989837647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,1,balanced,0.12293333808581035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,2,power_law_1.2,1.1902015686035157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.2,0.9590975761413574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,1,balanced,0.07969066500663757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,1,balanced,0.06468266745408376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,1,balanced,0.12340799967447917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.2,1.4004480361938476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,1,balanced,0.06623466809590657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,2,power_law_1.2,1.8819135665893554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,1,balanced,0.07854933540026347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,1,balanced,0.12752532958984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.2,1.894246482849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,1,balanced,0.06840000053246816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,1,balanced,0.12798933188120523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,1,balanced,0.08058133224646251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,2,power_law_1.2,2.537606430053711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,1,balanced,0.06849066913127899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,2,power_law_1.2,3.724979019165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,1,balanced,0.13038933277130127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,1,balanced,0.08281066517035167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,1,balanced,0.06677866478761037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,2,power_law_1.2,3.9011775970458986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,1,balanced,0.13335466384887695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,1,balanced,0.06858133276303609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,1,balanced,0.08046933511892955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,1,balanced,0.14327466487884521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,2,power_law_1.2,7.388159942626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,1,balanced,0.1456160048643748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,1,balanced,0.07034666836261749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,1,balanced,0.0821973333756129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,1,balanced,0.07239466905593872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,1,balanced,0.14677332838376364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,1,balanced,0.0828000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,1,balanced,0.07252799967924754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,1,balanced,0.08459200461705525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,1,balanced,0.18080532550811768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,1,balanced,0.07903466622034709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,1,balanced,0.08906666437784831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,1,balanced,0.08628267049789429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,1,balanced,0.1780959963798523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,1,balanced,0.08902399738629659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,1,balanced,0.08738666772842407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,1,balanced,0.09525332848230998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,1,balanced,0.24392000834147134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,1,balanced,0.1037600040435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,1,balanced,0.09338666995366414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,1,balanced,0.10517866412798564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,1,balanced,0.27533332506815594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,1,balanced,0.10964799920717876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,1,balanced,0.1136799951394399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,1,balanced,0.3666293223698934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,1,balanced,0.13712533315022787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,1,balanced,0.4292373259862264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,1,balanced,0.14939733346303305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,1,balanced,0.12267200152079265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,1,balanced,0.18659732739130655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,1,balanced,0.158160001039505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,1,balanced,0.6123786767323812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,1,balanced,0.23228265841801962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,1,balanced,0.18332799275716147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,1,balanced,0.021183999876181286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,1,balanced,0.777786652247111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,1,balanced,0.31006399790445965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,1,balanced,0.24454933404922485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,1,balanced,0.02144533395767212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,1,balanced,0.023002666731675465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,1,balanced,0.3006239930788676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,1,balanced,0.3667413393656413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,1,balanced,0.02310933421055476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,1,balanced,0.0252960001428922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,1,balanced,1.1300960381825764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,1,balanced,0.4219573338826497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,1,balanced,0.029461334149042766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,1,balanced,0.5197226603825887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,1,balanced,0.029487999776999157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,1,balanced,0.02945599953333537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,1,balanced,0.029802667597929638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,1,balanced,0.030400000512599945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,1,balanced,0.5446240107218424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,1,balanced,0.6602986653645834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,1,balanced,1.4885706901550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,1,balanced,0.032144000132878624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,1,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,1,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,1,balanced,0.03565333286921183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,1,balanced,0.7823572953542074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,1,balanced,0.955450693766276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,1,balanced,0.07522666454315186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,1,balanced,0.07732800145943959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,1,balanced,1.837125301361084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,1,balanced,0.08043733239173889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,1,balanced,0.05624533196290334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,1,balanced,0.062319998939832054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,1,balanced,1.025765339533488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,1,balanced,1.256719986597697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,1,balanced,0.08409600456555684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,1,balanced,0.10361066460609436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,1,balanced,0.13903466860453287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,1,balanced,2.916245460510254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,1,balanced,1.2644480069478352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,1,balanced,0.16780267159144083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,1,balanced,1.553701400756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,1,balanced,0.23868799209594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,1,balanced,0.30236266056696576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,1,balanced,1.981386661529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,1,balanced,0.4400000174840291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,1,balanced,2.434783935546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,1,balanced,0.5765386819839478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,1,balanced,5.745706558227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,1,balanced,0.710314671198527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,1,balanced,3.86625067392985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,1,balanced,4.78554121653239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,1,balanced,1.1151893138885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,1,power_law_1.01,0.04991999864578247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,1,power_law_1.01,0.06448000073432922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,1,power_law_1.01,0.08069120049476623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,1,power_law_1.01,0.11670399904251098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,1,power_law_1.01,0.11973760128021241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,1,power_law_1.01,0.12187520265579224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,1,power_law_1.01,0.12611839771270753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,1,power_law_1.01,0.12317440509796143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,1,balanced,2.191802660624186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,1,power_law_1.01,0.12243839502334594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,1,power_law_1.01,0.12470400333404541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,1,power_law_1.01,0.03621760010719299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,1,power_law_1.01,0.12629120349884032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,1,power_law_1.01,0.12816640138626098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,1,power_law_1.01,0.04762240052223206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,1,power_law_1.01,0.13709440231323242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,1,power_law_1.01,0.05102720260620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,1,power_law_1.01,0.13507839441299438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,1,power_law_1.01,0.07571200132369996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,1,power_law_1.01,0.07680000066757202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,1,power_law_1.01,0.14897279739379882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,1,power_law_1.01,0.07908480167388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,1,power_law_1.01,0.0798143982887268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,1,power_law_1.01,0.15628160238265992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,1,power_law_1.01,0.08131200075149536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,1,power_law_1.01,0.15941760540008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,1,power_law_1.01,0.1894592046737671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,1,power_law_1.01,0.08372480273246766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,1,power_law_1.01,0.08547199964523315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,1,power_law_1.01,0.20940160751342773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,1,power_law_1.01,0.08703359961509705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,1,power_law_1.01,0.09108480215072631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,1,power_law_1.01,0.25410559177398684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,1,power_law_1.01,0.09433599710464477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,1,power_law_1.01,0.3093055963516235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,1,power_law_1.01,0.09737600088119507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,1,power_law_1.01,0.4093760013580322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,1,power_law_1.01,0.1087231993675232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,1,power_law_1.01,0.01943040043115616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,1,power_law_1.01,0.03758719861507416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,1,power_law_1.01,0.47226881980895996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,1,power_law_1.01,0.10951679944992065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,1,power_law_1.01,0.04047360122203827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,1,power_law_1.01,0.01996160000562668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,1,power_law_1.01,0.11895040273666382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,1,power_law_1.01,0.0449535995721817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,1,power_law_1.01,0.6648191928863525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,1,power_law_1.01,0.02107519954442978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,1,power_law_1.01,0.1410367965698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,1,power_law_1.01,0.021184000372886657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,1,power_law_1.01,0.06258559823036194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,1,power_law_1.01,0.023839999735355378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,1,power_law_1.01,0.06325119733810425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,1,power_law_1.01,0.15727360248565675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,1,power_law_1.01,0.830624008178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,1,power_law_1.01,0.028569599986076354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,1,power_law_1.01,0.06454399824142457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,1,power_law_1.01,0.2017728090286255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,1,power_law_1.01,1.1742015838623048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,1,power_law_1.01,0.06539520025253295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,1,power_law_1.01,0.028038400411605834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,1,power_law_1.01,0.2396672010421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,1,power_law_1.01,0.06615039706230164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,1,power_law_1.01,0.0282368004322052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,1,power_law_1.01,1.5267200469970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,1,power_law_1.01,0.06622719764709473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,1,power_law_1.01,0.32279040813446047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,1,power_law_1.01,0.027750399708747864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,1,power_law_1.01,0.06773120164871216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,1,power_law_1.01,1.8797183990478517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,1,power_law_1.01,0.3890367984771729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,1,power_law_1.01,0.029139199852943422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,1,power_law_1.01,0.06963199973106385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,1,power_law_1.01,0.07398399710655212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,1,power_law_1.01,0.030873599648475646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,1,power_law_1.01,0.5428607940673829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,1,power_law_1.01,0.07604479789733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,1,power_law_1.01,2.974412727355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,1,power_law_1.01,0.6857728004455567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,1,power_law_1.01,0.030988800525665283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,1,power_law_1.01,0.08006399869918823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,1,power_law_1.01,0.9816320419311524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,1,power_law_1.01,0.0877888023853302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,1,power_law_1.01,0.03221119940280914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,1,power_law_1.01,1.2761152267456055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,1,power_law_1.01,0.08931840062141419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,1,power_law_1.01,5.808224105834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,1,power_law_1.01,0.03477759957313538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,1,power_law_1.01,0.09846400022506714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,1,power_law_1.01,1.5721535682678223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,1,power_law_1.01,0.07408000230789184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,1,power_law_1.01,0.11285120248794556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,1,power_law_1.01,0.07562879920005798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,1,power_law_1.01,2.4547264099121096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.01,0.12906880378723146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,1,power_law_1.01,0.07811840176582337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.01,0.1613759994506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,1,power_law_1.01,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,1,power_law_1.01,4.799103927612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.01,0.18655359745025635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.01,0.0596671998500824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.01,0.24927999973297119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.01,0.08106880187988282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.01,0.3120448112487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.01,0.10166399478912354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.01,0.43323521614074706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.01,0.13610880374908446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.01,0.551910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.01,0.16533759832382203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.01,0.792736005783081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.01,0.23697280883789062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.01,1.0335807800292969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.01,0.30097920894622804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.01,1.2731072425842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.01,0.4375296115875244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,1,power_law_1.01,1.979635238647461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.01,0.5725696086883545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.01,0.7079616069793702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,1,power_law_1.01,3.914022445678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,1,power_law_1.01,1.1117695808410644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,1,power_law_1.01,2.187923240661621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,1,power_law_1.2,0.04984320104122162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,1,power_law_1.2,0.05793280005455017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,1,power_law_1.2,0.07758079767227173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,1,power_law_1.2,0.11656960248947143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,1,power_law_1.2,0.03946239948272705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,1,power_law_1.2,0.11544320583343506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,1,power_law_1.2,0.034944000840187076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,1,power_law_1.2,0.041222399473190306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,1,power_law_1.2,0.12117120027542114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,1,power_law_1.2,0.13010560274124144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,1,power_law_1.2,0.04709759950637817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,1,power_law_1.2,0.043833601474761966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,1,power_law_1.2,0.12272640466690063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,1,power_law_1.2,0.1244928002357483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,1,power_law_1.2,0.06509439945220948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,1,power_law_1.2,0.05000960230827332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,1,power_law_1.2,0.12465280294418335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,1,power_law_1.2,0.06504960060119629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,1,power_law_1.2,0.07589120268821717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,1,power_law_1.2,0.12873599529266358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,1,power_law_1.2,0.06559360027313232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,1,power_law_1.2,0.1314239978790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,1,power_law_1.2,0.07774080038070678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,1,power_law_1.2,0.06553599834442139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,1,power_law_1.2,0.13351680040359498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,1,power_law_1.2,0.07817599773406983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,1,power_law_1.2,0.06648319959640503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,1,power_law_1.2,0.1356608033180237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,1,power_law_1.2,0.07972480058670044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,1,power_law_1.2,0.06628479957580566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,1,power_law_1.2,0.1534656047821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,1,power_law_1.2,0.08133760094642639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,1,power_law_1.2,0.06722559928894042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,1,power_law_1.2,0.15673600435256957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,1,power_law_1.2,0.08206080198287964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,1,power_law_1.2,0.06997759938240052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,1,power_law_1.2,0.15649919509887694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,1,power_law_1.2,0.0840511977672577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,1,power_law_1.2,0.0741823971271515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,1,power_law_1.2,0.18280960321426393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,1,power_law_1.2,0.08513280153274536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,1,power_law_1.2,0.07712640166282654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,1,power_law_1.2,0.2018752098083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,1,power_law_1.2,0.02041600048542023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,1,power_law_1.2,0.09016960263252258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,1,power_law_1.2,0.08076159954071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,1,power_law_1.2,0.2531712055206299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,1,power_law_1.2,0.09194239974021912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,1,power_law_1.2,0.09029120206832886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,1,power_law_1.2,0.3007551908493042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,1,power_law_1.2,0.02035840004682541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,1,power_law_1.2,0.09148160219192505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,1,power_law_1.2,0.09681280255317688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,1,power_law_1.2,0.40627198219299315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,1,power_law_1.2,0.021055999398231506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,1,power_law_1.2,0.10008319616317748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,1,power_law_1.2,0.10703999996185302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,1,power_law_1.2,0.4789567947387695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,1,power_law_1.2,0.021503999829292297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,1,power_law_1.2,0.11276160478591919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,1,power_law_1.2,0.10984959602355956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,1,power_law_1.2,0.6613247871398926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.2,0.13154560327529907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,1,power_law_1.2,0.023999999463558196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,1,power_law_1.2,0.11960959434509277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,1,power_law_1.2,0.8357824325561524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,1,power_law_1.2,0.02803199887275696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.2,0.1595136046409607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,1,power_law_1.2,0.13849600553512573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,1,power_law_1.2,1.171072006225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,1,power_law_1.2,0.027347201108932497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.2,0.18705919981002808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,1,power_law_1.2,0.15789439678192138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,1,power_law_1.2,0.02879360020160675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,1,power_law_1.2,1.5249919891357422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.2,0.24993278980255126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,1,power_law_1.2,0.20112640857696534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,1,power_law_1.2,0.02770560085773468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,1,power_law_1.2,1.874880027770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.2,0.31075201034545896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,1,power_law_1.2,0.02863360047340393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,1,power_law_1.2,2.983750343322754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.2,0.4335616111755371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,1,power_law_1.2,0.2410048007965088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,1,power_law_1.2,0.031123200058937074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,1,power_law_1.2,5.823705673217773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.2,0.5520192146301269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,1,power_law_1.2,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,1,power_law_1.2,0.32302079200744627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,1,power_law_1.2,0.032518398761749265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.2,0.7918079853057861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,1,power_law_1.2,0.39308159351348876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,1,power_law_1.2,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,1,power_law_1.2,0.5425471782684326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.2,1.037292766571045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,1,power_law_1.2,0.0741312026977539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,1,power_law_1.2,0.6841472148895263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.2,1.2767935752868653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,1,power_law_1.2,0.07582719922065735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,1,power_law_1.2,0.07886080145835876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,1,power_law_1.2,0.9792063713073731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,1,power_law_1.2,1.9879487991333007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,1,power_law_1.2,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,1,power_law_1.2,1.2803839683532714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,1,power_law_1.2,3.891846466064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.2,0.060153597593307497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,1,power_law_1.2,1.5703231811523437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.2,0.0814848005771637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.2,0.10226559638977051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,1,power_law_1.2,2.456723213195801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.2,0.13616640567779542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,1,power_law_1.2,4.797983932495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.2,0.16479359865188598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.2,0.23589119911193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.2,0.30090880393981934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.2,0.436736011505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.2,0.57292160987854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.2,0.7059904098510742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,1,power_law_1.2,1.1124544143676758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,1,power_law_1.2,2.1853824615478517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,8,balanced,0.04620266457398733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,8,balanced,0.05239466826121012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,8,balanced,0.0503359983364741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,8,balanced,0.04980800052483877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,8,balanced,0.04990399877230326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,8,balanced,0.07010133564472198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,8,balanced,0.05286400020122528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,8,balanced,0.058042665322621666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,8,balanced,0.06899199883143108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,8,balanced,0.04502933224042257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,8,balanced,0.09340799848238628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,8,balanced,0.08428266644477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,8,balanced,0.09709866841634114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,8,balanced,0.14792533715566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,8,balanced,0.052202666799227394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,8,balanced,0.11885333061218262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,8,balanced,0.1488640010356903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,8,balanced,0.24623999993006387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,8,balanced,0.24770132700602213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,8,balanced,0.44183464845021564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,8,balanced,0.06237333516279856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,8,balanced,0.17595734198888144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,8,balanced,0.2497653365135193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,8,balanced,0.4458186626434326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,8,balanced,0.08705066641171773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,8,balanced,0.2510666648546855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,8,balanced,0.17585599422454834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,8,balanced,0.44520000616709393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,8,balanced,0.251418670018514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,8,balanced,0.44810668627421063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,8,balanced,0.14422399799029031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,8,balanced,0.17725332578023276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,8,balanced,0.44757866859436035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,8,balanced,0.252293328444163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,8,balanced,0.17791465918223062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,8,balanced,0.1388053297996521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,8,balanced,0.446666677792867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,8,balanced,0.258133331934611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,8,balanced,0.17781333128611246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,8,balanced,0.14802666505177817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,8,balanced,0.45104531447092694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,8,balanced,0.2576106588045756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,8,balanced,0.18120000759760538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,8,balanced,0.259552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,8,balanced,0.16517333189646402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,8,balanced,0.45097601413726807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,8,balanced,0.18155733744303384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,8,balanced,0.264085332552592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,8,balanced,0.16383467117945352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,8,balanced,0.180842657883962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,8,balanced,0.26845333973566693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,8,balanced,0.45022400220235187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,8,balanced,0.16742932796478271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,8,balanced,0.18743467330932617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,8,balanced,0.23233066002527872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,8,balanced,0.2711413304011027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,8,balanced,0.4604479869206746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,8,balanced,0.2256586750348409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,8,balanced,0.1909760038057963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,8,balanced,0.46138131618499756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,8,balanced,0.2788640062014262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,8,balanced,0.22845866282780966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,8,balanced,0.46586668491363525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,8,balanced,0.19548799594243368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,8,balanced,0.29292800029118854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,8,balanced,0.3275360067685445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,8,balanced,0.4768426815668742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,8,balanced,0.3319360017776489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,8,balanced,0.20162665843963623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,8,balanced,0.3035306731859843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,8,balanced,0.4929279883702596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,8,balanced,0.3670080105463664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,8,balanced,0.21173866589864096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,8,balanced,0.31378666559855145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,8,balanced,0.3295519948005676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,8,balanced,0.5120480060577393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,8,balanced,0.2203999956448873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,8,balanced,0.33033599456151325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,8,balanced,0.36027733484903973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,8,balanced,0.5328799883524576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,8,balanced,0.2400373419125875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,8,balanced,0.3632213274637858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,8,balanced,0.40861864884694415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,8,balanced,0.5711466471354166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,8,balanced,0.26243199904759723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,8,balanced,0.4147893190383911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,8,balanced,0.3044533332188924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,8,balanced,0.48266132672627765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,8,balanced,0.468234658241272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,8,balanced,0.6349973281224569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,8,power_law_1.01,0.14977920055389404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,8,balanced,0.5566666523615519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,8,balanced,0.3531786600748698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,8,power_law_1.01,0.24142079353332518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,8,balanced,0.5743306477864584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,8,balanced,0.7167572975158691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,8,power_law_1.01,0.15332479476928712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,8,power_law_1.01,0.14615679979324342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,8,power_law_1.01,0.19473279714584352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,8,balanced,0.7018667062123617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,8,balanced,0.4619893232981364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,8,power_law_1.01,0.2699007987976074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,8,power_law_1.01,0.37024641036987305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,8,balanced,0.6732107003529867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,8,power_law_1.01,0.345632004737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,8,balanced,0.7760533491770426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,8,power_law_1.01,0.3562943935394287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,8,power_law_1.01,0.3678335905075073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,8,balanced,0.85807998975118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,8,power_law_1.01,0.3918272018432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,8,balanced,0.537775993347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,8,power_law_1.01,0.39921278953552247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,8,power_law_1.01,0.39400320053100585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,8,power_law_1.01,0.4147712230682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,8,balanced,0.951695998509725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,8,power_law_1.01,0.41217918395996095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,8,balanced,1.0886933008829753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,8,power_law_1.01,0.44124798774719237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,8,power_law_1.01,0.4510528087615967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,8,balanced,1.358415921529134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,8,power_law_1.01,0.5014016151428222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,8,balanced,0.744101365407308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,8,power_law_1.01,0.5083712100982666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,8,power_law_1.01,0.5628736019134521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,8,power_law_1.01,0.5747136116027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,8,power_law_1.01,0.6749311923980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,8,power_law_1.01,0.777894401550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,8,balanced,1.136074701944987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,8,balanced,1.214309295018514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,8,power_law_1.01,0.9075008392333984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,8,power_law_1.01,1.0776448249816895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,8,balanced,1.6452372868855794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,8,power_law_1.01,1.4958975791931153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,8,balanced,0.9539039929707845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,8,power_law_1.01,1.8762815475463868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,8,power_law_1.01,2.38287353515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,8,power_law_1.01,3.230188751220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,8,power_law_1.01,6.111257553100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,8,balanced,1.6627732912699382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,8,balanced,1.48691193262736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,8,balanced,2.1475680669148765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,8,balanced,1.157423973083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,8,power_law_1.01,0.09320319890975952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,8,power_law_1.01,0.1428096055984497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,8,power_law_1.01,0.08487679958343505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,8,power_law_1.01,0.08386560082435608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,8,power_law_1.01,0.10705920457839965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,8,power_law_1.01,0.11049599647521972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,8,power_law_1.01,0.049465599656105044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,8,balanced,2.216591993967692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,8,power_law_1.01,0.13052159547805786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,8,power_law_1.01,0.07931519746780395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,8,power_law_1.01,0.15593600273132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,8,balanced,2.114448070526123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,8,power_law_1.01,0.08724480271339416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,8,power_law_1.01,0.20604801177978516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,8,power_law_1.01,0.07660160064697266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,8,balanced,3.228783925374349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,8,power_law_1.01,0.10641920566558838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,8,power_law_1.01,0.20506880283355713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,8,power_law_1.01,0.12499840259552002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,8,power_law_1.01,0.2106559991836548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,8,power_law_1.01,0.056492799520492555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,8,power_law_1.01,0.21624960899353027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,8,power_law_1.01,0.14748799800872803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,8,balanced,1.728768030802409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,8,power_law_1.01,0.233024001121521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,8,power_law_1.01,0.15047039985656738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,8,power_law_1.01,0.060159999132156375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,8,power_law_1.01,0.2250368118286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,8,power_law_1.01,0.14882559776306153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,8,power_law_1.01,0.23496320247650146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,8,power_law_1.01,0.15612159967422484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,8,power_law_1.01,0.07645440101623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,8,power_law_1.01,0.23647360801696776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,8,power_law_1.01,0.1636672019958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,8,power_law_1.01,0.2545408010482788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,8,power_law_1.01,0.09042559862136841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,8,power_law_1.01,0.15953279733657838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,8,power_law_1.01,0.26031999588012694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,8,power_law_1.01,0.11393920183181763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,8,power_law_1.01,0.1647871971130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,8,power_law_1.01,0.26245760917663574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,8,power_law_1.01,0.11323519945144653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,8,power_law_1.01,0.17730560302734374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,8,power_law_1.01,0.2933439970016479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,8,power_law_1.01,0.12930560111999512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,8,power_law_1.01,0.17928320169448853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,8,power_law_1.01,0.18362879753112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,8,power_law_1.01,0.31052160263061523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,8,power_law_1.01,0.1255295991897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,8,power_law_1.01,0.1976256012916565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,8,power_law_1.01,0.3635263919830322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,8,power_law_1.01,0.1286080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,8,power_law_1.01,0.20539519786834717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,8,power_law_1.01,0.40804481506347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,8,power_law_1.01,0.17374720573425292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.01,0.2271359920501709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,8,power_law_1.01,0.17046400308609008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,8,power_law_1.01,0.49521918296813966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.01,0.2688127994537354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.01,0.2940160036087036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,8,power_law_1.01,0.5485439777374268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,8,power_law_1.01,0.1624768018722534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.01,0.35377280712127684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,8,power_law_1.01,0.26508159637451173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,8,power_law_1.01,0.7237567901611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.01,0.43583998680114744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,8,power_law_1.01,0.27129600048065183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,8,power_law_1.01,0.9238207817077637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.01,0.5883135795593262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,8,power_law_1.01,0.278604793548584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.01,0.741971206665039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,8,power_law_1.01,1.237612819671631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,8,power_law_1.01,0.24800639152526854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.01,1.048300838470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,8,power_law_1.01,1.4867008209228516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.01,0.24895999431610108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.01,1.1868736267089843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,8,power_law_1.01,1.8195392608642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.01,0.26733438968658446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,8,power_law_1.01,1.462662410736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.01,0.27513599395751953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,8,power_law_1.01,2.919001579284668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,8,power_law_1.01,2.4331775665283204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.01,0.31146240234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.01,0.3821887969970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,8,power_law_1.01,5.4362945556640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,8,power_law_1.01,4.372819137573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.01,0.5019199848175049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.01,0.5927807807922363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.01,0.873203182220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.01,1.0779775619506835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,8,power_law_1.01,1.3392448425292969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,8,power_law_1.01,2.0468095779418944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,8,balanced,4.355621337890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,8,power_law_1.01,4.042220687866211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,8,balanced,4.137344042460124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,8,balanced,6.392645517985026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,8,balanced,3.3679892222086587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,8,power_law_1.2,0.14906879663467407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,8,power_law_1.2,0.23518080711364747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,8,power_law_1.2,0.10222079753875732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,8,power_law_1.2,0.14141440391540527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,8,power_law_1.2,0.18650879859924316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,8,power_law_1.2,0.23851521015167237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,8,power_law_1.2,0.3148799896240234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,8,power_law_1.2,0.3398080110549927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,8,power_law_1.2,0.3600703954696655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,8,power_law_1.2,0.3791104078292847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,8,power_law_1.2,0.38353281021118163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,8,power_law_1.2,0.37687680721282957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,8,power_law_1.2,0.3689919948577881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,8,power_law_1.2,0.4041279792785645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,8,power_law_1.2,0.44279041290283205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,8,power_law_1.2,0.4591231822967529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,8,power_law_1.2,0.478547191619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,8,power_law_1.2,0.5523263931274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,8,power_law_1.2,0.5389823913574219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,8,power_law_1.2,0.6054207801818847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,8,power_law_1.2,0.6087999820709229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,8,power_law_1.2,0.7681727886199952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,8,power_law_1.2,0.8704959869384765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,8,power_law_1.2,1.0103232383728027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,8,power_law_1.2,1.2996992111206054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,8,power_law_1.2,1.7505279541015626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,8,power_law_1.2,2.1867904663085938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,8,power_law_1.2,2.6513599395751952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,8,power_law_1.2,4.093753433227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,8,power_law_1.2,7.891033935546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,8,power_law_1.2,0.092985600233078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,8,power_law_1.2,0.08357759714126586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,8,power_law_1.2,0.1331712007522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,8,power_law_1.2,0.10903680324554443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,8,power_law_1.2,0.07255679965019227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,8,power_law_1.2,0.06704000234603882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,8,power_law_1.2,0.10081919431686401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,8,power_law_1.2,0.0807039976119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,8,power_law_1.2,0.11905280351638795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,8,power_law_1.2,0.09690240025520325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,8,power_law_1.2,0.14404480457305907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,8,power_law_1.2,0.1175104022026062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,8,power_law_1.2,0.18511359691619872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,8,power_law_1.2,0.1449023962020874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,8,power_law_1.2,0.19684480428695678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,8,power_law_1.2,0.14540159702301025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,8,power_law_1.2,0.21121280193328856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,8,power_law_1.2,0.14998400211334229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,8,power_law_1.2,0.21000320911407472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,8,power_law_1.2,0.15518720149993898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,8,power_law_1.2,0.04888960123062134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,8,power_law_1.2,0.22072958946228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,8,power_law_1.2,0.16092159748077392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,8,power_law_1.2,0.2225343942642212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,8,power_law_1.2,0.16281599998474122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,8,power_law_1.2,0.07377920150756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,4,balanced,0.045994664231936135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,8,power_law_1.2,0.1565119981765747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,8,power_law_1.2,0.23082880973815917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,8,power_law_1.2,0.044486400485038755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,4,balanced,0.050069332122802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,8,power_law_1.2,0.17440639734268187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,4,balanced,0.053727999329566956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,8,power_law_1.2,0.2546816110610962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,8,power_law_1.2,0.055212801694869994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,4,balanced,0.06799999872843425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,4,balanced,0.053616002202034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,8,power_law_1.2,0.18605439662933348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,4,balanced,0.028922667105992634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,4,balanced,0.09789866209030151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,8,power_law_1.2,0.24664320945739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,8,power_law_1.2,0.18558080196380616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,4,balanced,0.07018133501211803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,4,balanced,0.03162666658560435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,4,balanced,0.1516800026098887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,4,balanced,0.05625066657861074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,8,power_law_1.2,0.07286400198936463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,8,power_law_1.2,0.20502400398254395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,4,balanced,0.044693330923716225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,8,power_law_1.2,0.26803839206695557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,4,balanced,0.24796799818674722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,4,balanced,0.09487467010815938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,4,balanced,0.06098133325576782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,4,balanced,0.06237333516279856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,8,power_law_1.2,0.22183680534362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,4,balanced,0.08541333675384521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,4,balanced,0.25010132789611816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,8,power_law_1.2,0.2827840089797974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,4,balanced,0.14914666612943014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,4,balanced,0.08984532952308655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,4,balanced,0.1399679978688558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,8,power_law_1.2,0.08787199854850769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.2,0.23836801052093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,4,balanced,0.1436853309472402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,4,balanced,0.25385600328445435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,4,balanced,0.245253324508667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,4,balanced,0.12446932991345723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,8,power_law_1.2,0.3119040012359619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.2,0.29367039203643797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,4,balanced,0.15202132860819498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,8,power_law_1.2,0.11133439540863037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,4,balanced,0.2547573248545329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,4,balanced,0.14663466811180115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,4,balanced,0.18479466438293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,4,balanced,0.44154131412506104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.2,0.3229439973831177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,8,power_law_1.2,0.3344831943511963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,4,balanced,0.25657065709431964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,4,balanced,0.15031466881434122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,8,power_law_1.2,0.11237119436264038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,4,balanced,0.25853333870569867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,4,balanced,0.1584160029888153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.2,0.3901119947433472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,4,balanced,0.18737600247065225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,8,power_law_1.2,0.4080832004547119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,4,balanced,0.2609279950459798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,4,balanced,0.44275200366973877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,4,balanced,0.16605866948763529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,8,power_law_1.2,0.12028800249099732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.2,0.4982143878936768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,4,balanced,0.18915732701619467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,8,power_law_1.2,0.43218560218811036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,4,balanced,0.26318933566411334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,4,balanced,0.1671733260154724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,4,balanced,0.1744906703631083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,4,balanced,0.2680746714274089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,4,balanced,0.1906399925549825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.2,0.6513152122497559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,4,balanced,0.44576533635457355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,8,power_law_1.2,0.5386688232421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,4,balanced,0.23185600837071738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,8,power_law_1.2,0.1279744029045105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,4,balanced,0.27505600452423096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,4,balanced,0.19125332434972128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,4,balanced,0.2326293389002482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.2,0.7268159866333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,8,power_law_1.2,0.6189568042755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,4,balanced,0.4490559895833333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,4,balanced,0.2784159978230794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,4,balanced,0.2331626613934835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,4,balanced,0.1963040033976237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,8,power_law_1.2,0.12912000417709352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.2,1.0669695854187011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,8,power_law_1.2,0.7733695983886719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,4,balanced,0.3715200026830037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,4,balanced,0.28682132562001544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,4,balanced,0.19827733437220255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,4,balanced,0.44676800568898517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,8,power_law_1.2,0.16447360515594484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,4,balanced,0.19851199785868326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,4,balanced,0.4137440125147502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,8,power_law_1.2,1.0112192153930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,4,balanced,0.3070080081621806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.2,1.5415679931640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,4,balanced,0.2051253318786621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,4,balanced,0.45235200723012287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,8,power_law_1.2,0.16045440435409547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,4,balanced,0.346288005510966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,4,balanced,0.2097653349240621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,8,power_law_1.2,1.3764096260070802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,4,balanced,0.3202880024909973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,8,power_law_1.2,1.7631935119628905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,8,power_law_1.2,0.15750399827957154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,4,balanced,0.21210666497548422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,4,balanced,0.4551253318786621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,4,balanced,0.3671253522237142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,8,power_law_1.2,1.7851072311401368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,4,balanced,0.2204479972521464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,4,balanced,0.4575306574503581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,8,power_law_1.2,0.2592000007629395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,4,balanced,0.3543146848678589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,8,power_law_1.2,2.7899904251098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,4,balanced,0.23849600553512573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,4,balanced,0.45449598630269367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,8,power_law_1.2,2.075366401672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,4,balanced,0.41441067059834796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,4,balanced,0.46485332647959393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,8,power_law_1.2,0.2676736116409302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,4,balanced,0.4017920096715291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,4,balanced,0.2516053318977356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,8,power_law_1.2,5.222374343872071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,8,power_law_1.2,3.220044708251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,4,balanced,0.4660853147506714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,8,power_law_1.2,0.28765439987182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,4,balanced,0.46167465051015216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,4,balanced,0.4713386694590251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,4,balanced,0.284768005212148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,4,balanced,0.46541865666707355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,8,power_law_1.2,0.25438079833984373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,8,power_law_1.2,6.332147216796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,4,balanced,0.49005333582560223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.2,0.2564480066299438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,4,balanced,0.3232106765111287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,4,balanced,0.5500320196151733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.2,0.2578752040863037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,4,balanced,0.5108000040054321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,4,balanced,0.5336320002873739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.2,0.29159040451049806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,4,balanced,0.37382932504018146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,4,balanced,0.5371413230895996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.2,0.3369856119155884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.2,0.4048192024230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,4,balanced,0.6317386627197266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,4,balanced,0.5648266474405924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,4,balanced,0.6831253369649252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,4,balanced,0.42818665504455566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.2,0.5551487922668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,4,balanced,0.6115253369013468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.2,0.6669695854187012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.2,0.9560192108154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,4,balanced,0.5927733182907104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,4,balanced,0.8201493422190348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.2,1.2456576347351074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,4,balanced,0.6960693200429281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,4,balanced,1.023146629333496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,8,power_law_1.2,1.4575488090515136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,8,power_law_1.2,2.4820480346679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,4,balanced,0.7026560306549072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,8,power_law_1.2,4.739603042602539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,4,balanced,0.7991840044657389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,4,balanced,1.181338628133138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,4,balanced,1.2048319975535076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,4,balanced,0.8905920187632242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,4,balanced,0.9901599884033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,4,balanced,1.4461706479390461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,4,balanced,1.3183680375417073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,4,balanced,1.5956907272338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,4,balanced,1.2880746523539226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,4,balanced,1.829114596048991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,4,balanced,1.428869406382243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,4,power_law_1.01,0.09541760087013244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,4,power_law_1.01,0.1449087977409363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,4,power_law_1.01,0.08972160220146179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,4,power_law_1.01,0.1379263997077942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,4,balanced,1.5847946802775066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,4,power_law_1.01,0.18250240087509156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,4,power_law_1.01,0.25651841163635253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,4,power_law_1.01,0.3400511980056763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,4,power_law_1.01,0.3459775924682617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,4,power_law_1.01,0.36068480014801024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,4,power_law_1.01,0.3714495897293091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,4,balanced,2.3404480616251626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,4,power_law_1.01,0.3696320056915283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,4,power_law_1.01,0.3800640106201172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,4,power_law_1.01,0.3923583984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,4,balanced,1.9829600652058919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,4,power_law_1.01,0.4079103946685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,4,power_law_1.01,0.4290815830230713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,4,power_law_1.01,0.4164224147796631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,4,power_law_1.01,0.4525184154510498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,4,power_law_1.01,0.4914815902709961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,4,balanced,2.7301066716512046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,4,power_law_1.01,0.5014463901519776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,4,power_law_1.01,0.5675712108612061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,4,power_law_1.01,0.5645440101623536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,4,power_law_1.01,0.7023935794830323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,4,power_law_1.01,0.7890944004058837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,4,balanced,2.3943626085917153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,4,power_law_1.01,1.0090815544128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,4,power_law_1.01,1.1691455841064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,4,power_law_1.01,1.6115455627441406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,4,power_law_1.01,1.9002496719360351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,4,power_law_1.01,2.21910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,4,power_law_1.01,3.3053310394287108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,4,power_law_1.01,6.3822273254394535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,4,balanced,2.6537493069966636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,4,balanced,4.631360054016113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,4,power_law_1.01,0.06825600266456604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,4,power_law_1.01,0.0652671992778778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,4,balanced,5.347855885823567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,4,power_law_1.01,0.09407359957695008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,4,power_law_1.01,0.0830847978591919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,4,power_law_1.01,0.07698559761047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,4,power_law_1.01,0.06935039758682252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,4,power_law_1.01,0.08763520121574402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,4,power_law_1.01,0.09208959937095643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,4,power_law_1.01,0.09946240186691284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,4,power_law_1.01,0.1223039984703064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,4,power_law_1.01,0.12312320470809937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,4,balanced,4.654874801635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,4,power_law_1.01,0.15616639852523803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,4,power_law_1.01,0.1535871982574463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,4,power_law_1.01,0.1919167995452881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,4,power_law_1.01,0.15603840351104736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,4,power_law_1.01,0.1630784034729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,4,power_law_1.01,0.20380799770355223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,4,power_law_1.01,0.1657472014427185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,4,power_law_1.01,0.2098623991012573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,4,power_law_1.01,0.1673408031463623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,4,power_law_1.01,0.22263040542602539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,4,power_law_1.01,0.17620480060577393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,4,power_law_1.01,0.2111423969268799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,4,balanced,5.2254025141398115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,4,power_law_1.01,0.18021759986877442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,4,power_law_1.01,0.23095040321350097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,4,power_law_1.01,0.18338559865951537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,4,power_law_1.01,0.22748799324035646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,4,power_law_1.01,0.19262720346450807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,4,power_law_1.01,0.2045759916305542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,4,power_law_1.01,0.24118399620056152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,4,power_law_1.01,0.21155838966369628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,4,power_law_1.01,0.2500672101974487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,4,power_law_1.01,0.2592128038406372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,4,power_law_1.01,0.24023039340972902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,4,power_law_1.01,0.2723072052001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.01,0.26445438861846926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,4,power_law_1.01,0.30248959064483644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.01,0.3172800064086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.01,0.35343360900878906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,4,power_law_1.01,0.32759039402008056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.01,0.43856000900268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,4,power_law_1.01,0.38143999576568605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.01,0.5163263797760009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.01,0.6817152023315429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,4,power_law_1.01,0.43555197715759275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,4,power_law_1.01,0.029145601391792297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.01,0.8678848266601562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,4,power_law_1.01,0.523686408996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,4,power_law_1.01,0.05008640289306641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,4,power_law_1.01,0.644985580444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.01,1.1907456398010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,4,power_law_1.01,0.8053440093994141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,4,power_law_1.01,0.041580799221992495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.01,1.4675071716308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,4,power_law_1.01,0.9996095657348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,4,power_law_1.01,0.054745602607727054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,4,power_law_1.01,1.8843263626098632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,4,power_law_1.01,1.340652847290039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,4,power_law_1.01,0.0689087986946106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,4,power_law_1.01,0.08709759712219238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,4,power_law_1.01,1.7546239852905274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,4,power_law_1.01,2.960211181640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,4,power_law_1.01,2.2041343688964843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,4,power_law_1.01,0.10787839889526367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,4,power_law_1.01,5.216287994384766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,4,power_law_1.01,0.11018240451812744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,4,power_law_1.01,3.2350399017333986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,4,power_law_1.01,0.11831680536270142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,4,power_law_1.01,6.121158218383789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,4,power_law_1.01,0.11793279647827148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,4,power_law_1.01,0.12431999444961547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,4,power_law_1.01,0.13038079738616942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,4,power_law_1.01,0.12926080226898193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,4,power_law_1.01,0.13482240438461304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,4,power_law_1.01,0.1460927963256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,4,power_law_1.01,0.1517632007598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,4,power_law_1.01,0.15048320293426515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,4,power_law_1.01,0.2972543954849243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.01,0.3210495948791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.01,0.2597631931304932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.01,0.26206719875335693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.01,0.3025151968002319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.01,0.349945592880249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.01,0.44136958122253417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.01,0.5498367786407471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.01,0.755673599243164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.01,0.944115161895752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,4,power_law_1.01,1.1036031723022461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,4,power_law_1.01,1.7086719512939452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,4,power_law_1.01,3.29947509765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,4,power_law_1.2,0.09431679844856262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,4,power_law_1.2,0.13596800565719605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,4,power_law_1.2,0.0809984028339386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,4,power_law_1.2,0.12715519666671754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,4,power_law_1.2,0.16649600267410278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,4,power_law_1.2,0.244268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,4,power_law_1.2,0.30853118896484377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,4,power_law_1.2,0.3446208000183105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,4,power_law_1.2,0.35009920597076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,4,power_law_1.2,0.3728895902633667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,4,power_law_1.2,0.38472959995269773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,4,power_law_1.2,0.386028790473938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,4,power_law_1.2,0.38295679092407225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,4,power_law_1.2,0.40149760246276855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,4,power_law_1.2,0.4165823936462402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,4,power_law_1.2,0.43955841064453127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,4,power_law_1.2,0.47583999633789065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,4,power_law_1.2,0.5241983890533447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,4,power_law_1.2,0.5381696224212646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,4,power_law_1.2,0.5959296226501465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,4,power_law_1.2,0.6043647766113281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,4,power_law_1.2,0.7036928176879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,4,power_law_1.2,0.828985595703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,4,power_law_1.2,1.0704895973205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,4,power_law_1.2,1.2624575614929199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,4,power_law_1.2,1.6552064895629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,4,power_law_1.2,0.06643199920654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,4,power_law_1.2,2.0165760040283205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,4,power_law_1.2,0.09072639942169189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,4,power_law_1.2,2.4874496459960938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,4,power_law_1.2,0.062067198753356936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,4,power_law_1.2,0.08794239759445191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,4,power_law_1.2,4.020870590209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,4,power_law_1.2,0.12116479873657227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,4,power_law_1.2,0.14176000356674195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,4,power_law_1.2,7.504672241210938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,4,power_law_1.2,0.18920960426330566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,4,power_law_1.2,0.20145280361175538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,4,power_law_1.2,0.2081536054611206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,4,power_law_1.2,0.06502400040626526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,4,power_law_1.2,0.21025280952453612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,4,power_law_1.2,0.02858240008354187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,4,power_law_1.2,0.212774395942688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,4,power_law_1.2,0.0819263994693756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,4,power_law_1.2,0.2262847900390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,4,power_law_1.2,0.06372479796409607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,4,power_law_1.2,0.04963200092315674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,4,power_law_1.2,0.22717440128326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,4,power_law_1.2,0.0852735996246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,4,power_law_1.2,0.036934399604797365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,4,power_law_1.2,0.24055678844451905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,4,power_law_1.2,0.09736319780349731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,4,power_law_1.2,0.05103359818458557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,4,power_law_1.2,0.2615488052368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,4,power_law_1.2,0.1211583971977234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,4,power_law_1.2,0.061945599317550656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,4,power_law_1.2,0.2624959945678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,2,balanced,0.056128000219662987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,4,power_law_1.2,0.0791104018688202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,4,power_law_1.2,0.1457919955253601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,4,power_law_1.2,0.28286080360412597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,4,power_law_1.2,0.16268800497055053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,4,power_law_1.2,0.1102336049079895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,2,balanced,0.06993600229422252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,4,power_law_1.2,0.30870399475097654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,4,power_law_1.2,0.15780479907989503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,2,balanced,0.09544000029563904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,4,power_law_1.2,0.3409535884857178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,4,power_law_1.2,0.16769280433654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,4,power_law_1.2,0.1082751989364624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,2,balanced,0.1506613294283549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,4,power_law_1.2,0.4106175899505615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,4,power_law_1.2,0.16663680076599122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,4,power_law_1.2,0.11214720010757447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,2,balanced,0.24914133548736572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,4,power_law_1.2,0.4546112060546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,4,power_law_1.2,0.17831039428710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,2,balanced,0.4442559878031413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,4,power_law_1.2,0.11868799924850464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,4,power_law_1.2,0.18016639947891236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,4,power_law_1.2,0.5368576049804688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,2,balanced,0.4445279836654663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,4,power_law_1.2,0.12546559572219848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,2,balanced,0.44709332784016925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,4,power_law_1.2,0.18533120155334473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,4,power_law_1.2,0.655289602279663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,4,power_law_1.2,0.12918399572372435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,2,balanced,0.4497493505477905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,4,power_law_1.2,0.20355839729309083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,4,power_law_1.2,0.13660800457000732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,4,power_law_1.2,0.8653056144714355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,2,balanced,0.4482773145039876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,4,power_law_1.2,0.20837759971618652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,4,power_law_1.2,0.13657599687576294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,2,balanced,0.4533919890721639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,4,power_law_1.2,1.091750431060791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,4,power_law_1.2,0.15349760055541992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,2,balanced,0.45853865146636963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,4,power_law_1.2,0.22298240661621094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,4,power_law_1.2,1.4847552299499511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,2,balanced,0.46271467208862305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,4,power_law_1.2,0.1599679946899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,4,power_law_1.2,0.24883201122283935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,4,power_law_1.2,1.9522367477416993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,4,power_law_1.2,0.15745919942855835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,2,balanced,0.46678932507832843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.2,0.27232000827789304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,4,power_law_1.2,0.321343994140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,2,balanced,0.4825173219045003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.2,0.3380608081817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,4,power_law_1.2,2.1826879501342775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.2,0.3422271966934204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,2,balanced,0.4855733315149943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.2,0.3677504062652588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.2,0.2695487976074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,4,power_law_1.2,3.8874431610107423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,2,balanced,0.49456000328063965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.2,0.4545407772064209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.2,0.27718400955200195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.2,0.5188352108001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,2,balanced,0.5172799825668335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.2,0.7568448066711426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,4,power_law_1.2,6.474291229248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.2,0.31794559955596924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,2,balanced,0.5445333321889242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.2,0.8937024116516114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.2,0.3721920013427734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.2,1.286790370941162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,2,balanced,0.5850720008214315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.2,0.4657599925994873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.2,1.6703424453735352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.2,0.5849984169006348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,2,balanced,0.6311573187510172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,4,power_law_1.2,2.019424057006836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.2,0.7997888088226318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,4,power_law_1.2,3.0070207595825194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.2,1.009119987487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,2,balanced,0.7124746640523275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,4,power_law_1.2,1.2904831886291503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,4,power_law_1.2,6.0604095458984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,4,power_law_1.2,1.874457550048828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,2,balanced,0.8018933137257894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,2,balanced,0.04818666477998098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,4,power_law_1.2,3.7169921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,2,balanced,0.07467733323574066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,2,balanced,0.05215999980767568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,2,balanced,0.07237333556016286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,2,balanced,0.0874079962571462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,2,balanced,0.10274133086204529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,2,balanced,1.0171519915262859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,2,balanced,0.1604373355706533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,2,balanced,0.11026133100191753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,2,balanced,0.25827733675638836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,2,balanced,0.15567466616630554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,2,balanced,0.25937066475550336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,2,balanced,0.23348265886306763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,2,balanced,0.3804853359858195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,2,balanced,0.2613866726557414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,2,balanced,0.3630400101343791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,2,balanced,1.1709280014038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,2,balanced,0.26579199234644574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,2,balanced,0.3610293467839559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,2,balanced,0.26975999275843304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,2,balanced,0.35176531473795575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,2,balanced,0.275546669960022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,2,balanced,0.3521440029144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,2,balanced,0.35344000657399494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,2,balanced,0.27637332677841187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,2,balanced,0.3478399912516276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,2,balanced,0.2794346610705058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,2,balanced,0.35660799344380695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,2,balanced,0.2903680006663005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,2,balanced,0.35166935125986737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,2,balanced,0.298527995745341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,2,balanced,1.6999200185139973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,2,balanced,0.359386682510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,2,balanced,0.3039199908574422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,2,balanced,0.368282675743103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,2,balanced,0.31461334228515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,2,balanced,0.36930131912231445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,2,balanced,0.34193066755930585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,2,balanced,0.3916693528493245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,2,balanced,0.360917329788208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,2,balanced,0.41015466054280597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,2,balanced,0.41012267271677655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,2,balanced,0.44515732924143475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,2,balanced,0.45899732907613117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,2,balanced,0.47565333048502606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,2,balanced,0.02362666775782903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,2,balanced,2.058186690012614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,2,balanced,0.027104000250498455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,2,balanced,0.592192014058431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,2,balanced,0.0397119993964831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,2,balanced,0.5614826679229736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,2,balanced,0.05580266813437144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,2,balanced,0.08635200063387553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,2,balanced,0.1383039951324463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,2,balanced,0.13888532916704813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,2,balanced,0.6958773136138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,2,balanced,0.1420799990495046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,2,balanced,0.6479733387629191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,2,balanced,0.14536000291506448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,2,balanced,0.14797866344451904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,2,balanced,0.15451199809710184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,2,balanced,0.16165866454442343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,2,balanced,0.1673120061556498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,2,balanced,0.9244320392608643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,2,balanced,0.9872213204701742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,2,balanced,0.1639893352985382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,2,balanced,0.17148800690968832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,2,balanced,2.5789705912272134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,2,balanced,0.17730132738749185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,2,balanced,0.19238932927449545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,2,balanced,0.2206719915072123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,2,balanced,1.1198026339213054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,2,balanced,0.2315946618715922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,2,balanced,1.1426986853281658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,2,balanced,0.2964106599489848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,2,balanced,0.26582932472229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,2,balanced,0.36779733498891193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,2,balanced,1.624490737915039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,2,balanced,0.4466506640116374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,2,balanced,1.6288159688313801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,2,balanced,0.5148320198059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,2,power_law_1.01,0.06730239987373351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,2,balanced,3.822906812032064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,2,power_law_1.01,0.09256960153579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,2,power_law_1.01,0.08098559975624084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,2,balanced,2.0422239303588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,2,balanced,0.5664373238881429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,2,power_law_1.01,0.11626880168914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,2,power_law_1.01,0.1894271969795227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,2,balanced,2.123103936513265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,2,power_law_1.01,0.22381439208984374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,2,power_law_1.01,0.3161151885986328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,2,power_law_1.01,0.33927040100097655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,2,power_law_1.01,0.3433216094970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,2,power_law_1.01,0.3648384094238281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,2,power_law_1.01,0.3686464071273804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,2,power_law_1.01,0.3741055965423584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,2,balanced,0.9395733674367269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,2,power_law_1.01,0.38940160274505614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,2,power_law_1.01,0.40709757804870605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,2,power_law_1.01,0.4209023952484131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,2,power_law_1.01,0.42679681777954104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,2,power_law_1.01,0.44958720207214353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,2,balanced,2.528831958770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,2,power_law_1.01,0.5166272163391114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,2,power_law_1.01,0.514079999923706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,2,balanced,2.586554686228434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,2,power_law_1.01,0.5956031799316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,2,power_law_1.01,0.648524808883667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,2,power_law_1.01,0.7649407863616944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,2,power_law_1.01,0.8989376068115235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,2,balanced,1.0719626744588215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,2,power_law_1.01,1.1571840286254882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,2,power_law_1.01,1.354361629486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,2,power_law_1.01,1.824959945678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,2,power_law_1.01,2.3458368301391603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,2,power_law_1.01,2.8195775985717773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,2,power_law_1.01,4.4316864013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,2,power_law_1.01,8.210463714599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,2,power_law_1.01,0.049081599712371825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,2,power_law_1.01,0.0691648006439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,2,balanced,1.4422879219055176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,2,power_law_1.01,0.06487039923667907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,2,balanced,3.8556480407714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,2,power_law_1.01,0.09125120043754578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,2,power_law_1.01,0.12410240173339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,2,power_law_1.01,0.14764800071716308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,2,power_law_1.01,0.1951167941093445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,2,balanced,4.069039980570476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,2,balanced,7.358224232991536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,2,power_law_1.01,0.2119999885559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,2,power_law_1.01,0.2155456066131592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,2,power_law_1.01,0.0890175998210907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,2,power_law_1.01,0.21752960681915284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,2,power_law_1.01,0.23371520042419433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,2,power_law_1.01,0.11155840158462524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,2,power_law_1.01,0.2390144109725952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,2,power_law_1.01,0.10065280199050904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,2,power_law_1.01,0.24506239891052245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,2,power_law_1.01,0.14049919843673705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,2,power_law_1.01,0.2557120084762573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,2,power_law_1.01,0.2663935899734497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,2,power_law_1.01,0.18462719917297363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,2,power_law_1.01,0.2773695945739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,2,power_law_1.01,0.22209279537200927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,2,power_law_1.01,0.2988991975784302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,2,power_law_1.01,0.34028160572052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,2,power_law_1.01,0.280134391784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,2,power_law_1.01,0.36117119789123536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,2,power_law_1.01,0.2964544057846069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,2,power_law_1.01,0.022291199862957002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,2,power_law_1.01,0.43215360641479494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,2,power_law_1.01,0.2881472110748291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,2,balanced,2.084949334462484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,2,power_law_1.01,0.4874879837036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,2,power_law_1.01,0.3061631917953491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,2,power_law_1.01,0.028332799673080444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,2,power_law_1.01,0.6283648014068604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,2,power_law_1.01,0.31135358810424807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,2,power_law_1.01,0.030559998750686646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,2,power_law_1.01,0.04945279955863953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,2,power_law_1.01,0.76593918800354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,2,power_law_1.01,0.3215167999267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,2,power_law_1.01,0.06770560145378113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,2,power_law_1.01,1.052665615081787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,2,power_law_1.01,0.08331519961357117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,2,power_law_1.01,0.3247040033340454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,2,power_law_1.01,1.2930944442749024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,2,power_law_1.01,0.10089600086212158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,2,power_law_1.01,0.3300352096557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,2,power_law_1.01,1.7935808181762696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,2,power_law_1.01,0.112716805934906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,2,power_law_1.01,0.11509120464324951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,2,power_law_1.01,0.33676159381866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,2,power_law_1.01,2.14007682800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,2,power_law_1.01,0.1190335988998413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,2,power_law_1.01,0.34536960124969485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,2,power_law_1.01,2.6508031845092774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,2,power_law_1.01,0.12271360158920289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,2,power_law_1.01,0.3558847904205322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,2,power_law_1.01,0.12872960567474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,2,power_law_1.01,4.149728012084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,2,power_law_1.01,0.3843391895294189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,2,power_law_1.01,0.13260159492492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.01,0.41599359512329104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,2,power_law_1.01,0.13299839496612548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.01,0.4724095821380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,2,power_law_1.01,0.13738880157470704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,2,power_law_1.01,7.9964546203613285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,2,power_law_1.01,0.14431359767913818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.01,0.5314623832702636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,2,balanced,7.471322377522786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,2,power_law_1.01,0.16161279678344725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.01,0.6834496021270752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,2,power_law_1.01,0.18223999738693236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.01,0.1978943943977356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.01,0.8099007606506348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.01,0.24967679977416993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.01,1.095244789123535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.01,0.23709440231323242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.01,0.3025536060333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,2,balanced,8.097546895345053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.01,1.390675163269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.01,0.3656447887420654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.01,0.4547327995300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.01,1.9406848907470704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.01,0.5547008037567138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.01,0.7241087913513183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.01,2.389081573486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.01,0.9320896148681641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,2,power_law_1.01,2.962508773803711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,2,power_law_1.01,1.0865983963012695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,2,power_law_1.01,4.766566467285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,2,power_law_1.01,1.6869119644165038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,2,balanced,4.103647867838542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,2,power_law_1.01,3.221113586425781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,2,power_law_1.01,8.738604736328124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,2,power_law_1.2,0.06759039759635925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,2,power_law_1.2,0.0902400016784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,2,power_law_1.2,0.08288000226020813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,2,power_law_1.2,0.11863679885864258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,2,power_law_1.2,0.16234879493713378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,2,power_law_1.2,0.05061759948730469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,2,power_law_1.2,0.22176640033721923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,2,power_law_1.2,0.06698240041732788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,2,power_law_1.2,0.30891520977020265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,2,power_law_1.2,0.061286401748657224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,2,power_law_1.2,0.32042880058288575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,2,power_law_1.2,0.09278720021247863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,2,power_law_1.2,0.11128319501876831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,2,power_law_1.2,0.3382335901260376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,2,power_law_1.2,0.13810559511184692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,2,power_law_1.2,0.34752640724182127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,2,power_law_1.2,0.19498879909515382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,2,power_law_1.2,0.36896638870239257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,2,power_law_1.2,0.19731839895248413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,2,power_law_1.2,0.3777152061462402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,2,power_law_1.2,0.20707199573516846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,2,power_law_1.2,0.3935807943344116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,2,power_law_1.2,0.21881599426269532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,2,power_law_1.2,0.4204095840454102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,2,power_law_1.2,0.2248768091201782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,2,power_law_1.2,0.4498623847961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,2,power_law_1.2,0.24048640727996826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,2,power_law_1.2,0.45959038734436036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,2,power_law_1.2,0.2473599910736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,2,power_law_1.2,0.47703042030334475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,2,power_law_1.2,0.2579904079437256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,2,power_law_1.2,0.5279744148254395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,2,power_law_1.2,0.2746687889099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,2,power_law_1.2,0.5398464202880859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,2,power_law_1.2,0.28300158977508544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,2,power_law_1.2,0.6228415966033936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,2,power_law_1.2,0.30935680866241455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,2,power_law_1.2,0.6695680141448974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,2,power_law_1.2,0.35345280170440674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,2,power_law_1.2,0.8123647689819335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,2,power_law_1.2,0.37293438911437987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,2,power_law_1.2,0.9439359664916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,2,power_law_1.2,0.43875842094421386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,2,power_law_1.2,1.192249584197998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,2,power_law_1.2,0.5044991970062256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,2,power_law_1.2,0.6443007946014404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,2,power_law_1.2,1.3881471633911133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,2,power_law_1.2,0.7824895858764649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,2,power_law_1.2,1.9787456512451171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,2,power_law_1.2,1.127507209777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,2,power_law_1.2,2.3530624389648436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,2,power_law_1.2,1.3696831703186034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,2,power_law_1.2,2.979046440124512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,2,power_law_1.2,1.8724288940429688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,2,power_law_1.2,4.609337615966797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,2,power_law_1.2,2.420908737182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,2,power_law_1.2,2.7549631118774416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,2,power_law_1.2,8.462329864501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,2,power_law_1.2,4.244249725341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,2,power_law_1.2,8.496876525878907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,1,balanced,0.05748266478379568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,2,power_law_1.2,0.02237440049648285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,2,power_law_1.2,0.09018239974975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,1,balanced,0.0722453345855077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,1,balanced,0.0997173289457957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,2,power_law_1.2,0.027699199318885804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,1,balanced,0.1558613379796346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,2,power_law_1.2,0.11051520109176635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,1,balanced,0.2585600018501282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,2,power_law_1.2,0.02969599962234497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,1,balanced,0.4583626588185628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,2,power_law_1.2,0.0955456018447876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,2,power_law_1.2,0.047014400362968445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,1,balanced,0.45711998144785565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,1,balanced,0.05429333448410034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,1,balanced,0.460970679918925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,2,power_law_1.2,0.12912000417709352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,2,power_law_1.2,0.0659775972366333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,1,balanced,0.47259732087453205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,1,balanced,0.060005332032839455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,2,power_law_1.2,0.07847679853439331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,2,power_law_1.2,0.1747648000717163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,1,balanced,0.47836800416310626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,1,balanced,0.07990399996439616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,2,power_law_1.2,0.10170880556106568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,1,balanced,0.48019198576609295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,2,power_law_1.2,0.19853440523147584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,1,balanced,0.4883093436559041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,1,balanced,0.11682666341463725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,2,power_law_1.2,0.11059199571609497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,2,power_law_1.2,0.28048000335693357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,1,balanced,0.4933706521987915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,1,balanced,0.18385066588719687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,2,power_law_1.2,0.11381759643554687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,2,power_law_1.2,0.287558388710022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,1,balanced,0.499941349029541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,1,balanced,0.28865599632263184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,1,balanced,0.2919306755065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,2,power_law_1.2,0.11860480308532714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,1,balanced,0.5104533433914185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,2,power_law_1.2,0.30028159618377687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,1,balanced,0.2948906620343526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,1,balanced,0.5185226599375407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,2,power_law_1.2,0.12291840314865113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,2,power_law_1.2,0.3054656028747559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,1,balanced,0.2978079915046692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,1,balanced,0.5370986859003702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,2,power_law_1.2,0.13167359828948974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,2,power_law_1.2,0.304422402381897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,1,balanced,0.021429332594076794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,1,balanced,0.3007199962933858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,1,balanced,0.09731733798980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,2,power_law_1.2,0.3212032079696655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,2,power_law_1.2,0.1363968014717102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,1,balanced,0.5769173304239908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,1,balanced,0.3051466743151347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,2,power_law_1.2,0.13371520042419432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,2,power_law_1.2,0.3258048057556152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,1,balanced,0.1111840009689331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,1,balanced,0.141157329082489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,1,balanced,0.31018133958180744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,2,power_law_1.2,0.1398911952972412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,1,balanced,0.6063040097554525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,1,balanced,0.1998293399810791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,1,balanced,0.023589332898457844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,2,power_law_1.2,0.3347071886062622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,1,balanced,0.29922133684158325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,1,balanced,0.3135253389676412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,2,power_law_1.2,0.14439680576324462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,1,balanced,0.4851413170496623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,2,power_law_1.2,0.3583103895187378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,1,balanced,0.3228800098101298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,1,balanced,0.6810932954152426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,1,balanced,0.4814026753107707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,2,power_law_1.2,0.165721595287323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,1,balanced,0.027248000105222065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,1,balanced,0.4776800076166789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,1,balanced,0.3309760093688965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,2,power_law_1.2,0.35244159698486327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,2,power_law_1.2,0.19393279552459716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,1,balanced,0.4861066738764445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,1,balanced,0.3426719903945923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,1,balanced,0.47679467995961505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,1,balanced,0.03733866661787033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.2,0.206278395652771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,1,balanced,0.753706693649292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,1,balanced,0.4878079891204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,1,balanced,0.37489600976308185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,2,power_law_1.2,0.37125120162963865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.2,0.2479167938232422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,1,balanced,0.06052266558011373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,1,balanced,0.4893173376719157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,1,balanced,0.09715732932090759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,1,balanced,0.42127466201782227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,1,balanced,0.4925866524378459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.2,0.24476799964904786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,2,power_law_1.2,0.3998656034469604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,1,balanced,0.10109333197275798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,1,balanced,0.49876264731089276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.2,0.31804800033569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,1,balanced,0.9725173314412435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.2,0.43071999549865725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,1,balanced,0.10337600111961365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,1,balanced,0.4563680092493693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,1,balanced,0.5038880109786987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,1,balanced,0.10622933506965637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.2,0.3823168039321899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.2,0.5009535789489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,1,balanced,0.10997333129247029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,1,balanced,0.5155413150787354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.2,0.4617919921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.2,0.5717440128326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,1,balanced,0.11337600151697795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,1,balanced,0.5401333173116049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,1,balanced,0.5264373222986857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,1,balanced,0.1157919963200887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.2,0.5719871997833252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.2,0.695411205291748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,1,balanced,1.0723146597544353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,1,balanced,0.5645866791407267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,1,balanced,0.11981333295504253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.2,0.8537728309631347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.2,0.781657600402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,1,balanced,0.61407999197642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,1,balanced,0.1204906702041626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,1,balanced,0.5940853357315063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.2,1.1079872131347657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,1,balanced,0.124208003282547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.2,0.9939840316772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,1,balanced,0.12823466459910074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.2,1.3866751670837403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,1,balanced,0.6627466678619385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,2,power_law_1.2,1.1857728004455566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,1,balanced,0.13992533087730408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,1,balanced,0.842853307723999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,1,balanced,1.4909119606018066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.2,1.8937856674194335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,2,power_law_1.2,1.8133440017700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,1,balanced,0.16300266981124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,1,balanced,0.7228960196177164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.2,2.5225856781005858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,1,balanced,0.1727679967880249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,2,power_law_1.2,3.4339710235595704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,1,balanced,0.9639413356781006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,1,balanced,0.8656426270802816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,2,power_law_1.2,3.0003904342651366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,1,balanced,1.8078932762145996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,1,balanced,0.21995733181635538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,2,power_law_1.2,4.470080184936523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,1,balanced,0.24650132656097412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,1,balanced,1.0111626784006755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,1,balanced,1.3761919339497883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,2,power_law_1.2,8.76561279296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,1,balanced,0.3301333387692769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,1,balanced,2.5410240491231284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,1,balanced,0.33850665887196857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,1,balanced,1.5483466784159343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,1,balanced,1.711583932240804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,1,balanced,0.46925334135691327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,1,balanced,1.828938643137614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,1,balanced,3.18175474802653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,1,balanced,2.4769760767618814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,1,balanced,0.599616010983785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,1,balanced,2.6601972579956055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,1,balanced,0.9440213044484457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,1,balanced,3.189039866129557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,1,balanced,3.9162346522013345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,1,balanced,3.4532639185587564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,1,balanced,1.125637372334798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,1,balanced,3.925514539082845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,1,power_law_1.01,0.052223998308181765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,1,power_law_1.01,0.05027840137481689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,1,power_law_1.01,0.06943359971046448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,1,balanced,4.252021471659343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,1,power_law_1.01,0.05633280277252197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,1,power_law_1.01,0.07381119728088378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,1,power_law_1.01,0.06616320013999939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,1,power_law_1.01,0.12892160415649415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,1,power_law_1.01,0.09918720126152039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,1,power_law_1.01,0.18023040294647216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,1,power_law_1.01,0.13696000576019288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,1,power_law_1.01,0.2450239896774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,1,power_law_1.01,0.16929279565811156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,1,power_law_1.01,0.22429440021514893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,1,power_law_1.01,0.3293503999710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,1,balanced,1.3912533124287922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,1,power_law_1.01,0.23279359340667724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,1,power_law_1.01,0.3468415975570679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,1,balanced,5.917946497599284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,1,power_law_1.01,0.359987211227417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,1,power_law_1.01,0.24359679222106934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,1,power_law_1.01,0.24607999324798585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,1,power_law_1.01,0.3686144113540649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,1,power_law_1.01,0.25691521167755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,1,power_law_1.01,0.393721604347229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,1,power_law_1.01,0.2735167980194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,1,power_law_1.01,0.40767998695373536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,1,power_law_1.01,0.2901439905166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,1,power_law_1.01,0.40840959548950195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,1,power_law_1.01,0.30248959064483644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,1,power_law_1.01,0.44036478996276857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,1,power_law_1.01,0.3250432014465332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,1,power_law_1.01,0.44902400970458983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,1,power_law_1.01,0.3467711925506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,1,power_law_1.01,0.47151999473571776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,1,power_law_1.01,0.34698240756988524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,1,power_law_1.01,0.5009984016418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,1,power_law_1.01,0.4150847911834717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,1,power_law_1.01,0.5536191940307618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,1,power_law_1.01,0.4586175918579102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,1,power_law_1.01,0.6165056228637695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,1,power_law_1.01,0.5638336181640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,1,power_law_1.01,0.7221183776855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,1,power_law_1.01,0.644812822341919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,1,balanced,6.045397440592448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,1,power_law_1.01,0.8051839828491211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,1,power_law_1.01,0.8310463905334473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,1,power_law_1.01,1.0042112350463868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,1,power_law_1.01,1.0372096061706544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,1,power_law_1.01,1.1650176048278809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,1,power_law_1.01,1.4001472473144532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,1,power_law_1.01,1.5184703826904298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,1,balanced,6.669200261433919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,1,power_law_1.01,1.751603126525879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,1,power_law_1.01,1.8688255310058595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,1,power_law_1.01,2.4728767395019533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,1,power_law_1.01,2.5532863616943358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,1,power_law_1.01,3.2007102966308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,1,power_law_1.01,3.223539352416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,1,power_law_1.01,3.9228607177734376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,1,power_law_1.01,3.922284698486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,1,balanced,2.2709813117980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,1,power_law_1.01,5.934201431274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,1,power_law_1.01,6.086476898193359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,1,power_law_1.01,11.297824096679687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,1,power_law_1.01,11.66824951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,1,balanced,11.318079630533854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,1,balanced,11.67407480875651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,1,balanced,13.185855865478516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,1,power_law_1.01,0.021536000072956085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,1,power_law_1.01,0.09335039854049683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,1,balanced,4.294298807779948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,1,power_law_1.01,0.10583679676055908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,1,power_law_1.01,0.02280319929122925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,1,power_law_1.01,0.11686400175094605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,1,power_law_1.01,0.1719231963157654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,1,power_law_1.01,0.02625280022621155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,1,power_law_1.01,0.03627519905567169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,1,power_law_1.01,0.22227199077606202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,1,power_law_1.01,0.0584384024143219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,1,power_law_1.01,0.2830080032348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,1,power_law_1.01,0.0822272002696991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,1,power_law_1.01,0.10404479503631592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,1,power_law_1.01,0.36292479038238523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,1,power_law_1.01,0.11136640310287475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,1,power_law_1.01,0.3715519905090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,1,power_law_1.01,0.11384960412979125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,1,power_law_1.01,0.38810880184173585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,1,power_law_1.01,0.12151039838790893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,1,power_law_1.01,0.4043136119842529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,1,power_law_1.01,0.12583039999008178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,1,power_law_1.01,0.4139711856842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,1,power_law_1.01,0.1301632046699524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,1,power_law_1.01,0.4225791931152344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,1,power_law_1.01,0.43184638023376465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,1,power_law_1.01,0.13689600229263305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,1,power_law_1.01,0.4375999927520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,1,power_law_1.01,0.1393407940864563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,1,power_law_1.01,0.4548799991607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,1,power_law_1.01,0.14275840520858765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,1,power_law_1.01,0.45932798385620116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,1,power_law_1.01,0.4843455791473389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,1,power_law_1.01,0.14395519495010375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,1,power_law_1.01,0.5354496002197265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,1,power_law_1.01,0.15674240589141847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.01,0.5932928085327148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,1,power_law_1.01,0.1795904040336609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.01,0.6849855899810791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.01,0.19538559913635253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.01,0.7783872127532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.01,0.2398848056793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.01,0.9767552375793457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.01,0.2795072078704834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.01,1.182323169708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.01,0.37104001045227053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.01,1.5898303985595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.01,0.385862398147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.01,1.9926143646240235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.01,0.5150271892547608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.01,2.7956415176391602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.01,0.6416319847106934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.01,0.9117759704589844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.01,3.6014976501464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.01,1.1714112281799316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,1,power_law_1.01,4.38197135925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,1,power_law_1.01,1.436185646057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,1,power_law_1.01,6.7897789001464846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,1,power_law_1.01,2.2214399337768556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,1,power_law_1.01,4.332992172241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,1,power_law_1.01,13.233433532714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,1,power_law_1.2,0.050246399641036985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,1,power_law_1.2,0.05259519815444946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,1,power_law_1.2,0.05617280006408691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,1,power_law_1.2,0.06528639793395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,1,power_law_1.2,0.06855040192604064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,1,power_law_1.2,0.0977728009223938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,1,power_law_1.2,0.07921919822692872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,1,power_law_1.2,0.1217344045639038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,1,power_law_1.2,0.12309119701385499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,1,power_law_1.2,0.15364480018615723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,1,power_law_1.2,0.21440000534057618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,1,power_law_1.2,0.15942399501800536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,1,power_law_1.2,0.22552320957183838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,1,power_law_1.2,0.22255361080169678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,1,power_law_1.2,0.2367232084274292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,1,power_law_1.2,0.3191103935241699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,1,power_law_1.2,0.2465343952178955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,1,power_law_1.2,0.34162559509277346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,1,power_law_1.2,0.26451199054718016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,1,power_law_1.2,0.35433599948883054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,1,power_law_1.2,0.2771327972412109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,1,power_law_1.2,0.36078081130981443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,1,power_law_1.2,0.28663039207458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,1,power_law_1.2,0.389356803894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,1,power_law_1.2,0.3117120027542114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,1,power_law_1.2,0.4158207893371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,1,power_law_1.2,0.3324352025985718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,1,power_law_1.2,0.09230080246925354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,1,power_law_1.2,0.4229440212249756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,1,power_law_1.2,0.353439998626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,1,power_law_1.2,0.44286718368530276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,1,power_law_1.2,0.10528000593185424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,1,power_law_1.2,0.3615744113922119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,1,power_law_1.2,0.11512320041656494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,1,power_law_1.2,0.4628096103668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,1,power_law_1.2,0.4281023979187012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,1,power_law_1.2,0.161407995223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,1,power_law_1.2,0.4882495880126953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,1,power_law_1.2,0.19582719802856446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,1,power_law_1.2,0.4719679832458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,1,power_law_1.2,0.5116799831390381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,1,power_law_1.2,0.25235838890075685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,1,power_law_1.2,0.5696191787719727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,1,power_law_1.2,0.35185918807983396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,1,power_law_1.2,0.5796607971191406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,1,power_law_1.2,0.6508607864379883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,1,power_law_1.2,0.3625792026519775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,1,power_law_1.2,0.8365056037902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,1,power_law_1.2,0.38689279556274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,1,power_law_1.2,0.6305280208587647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,1,power_law_1.2,0.3952064037322998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,1,power_law_1.2,1.050163173675537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,1,power_law_1.2,0.41536641120910645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,1,power_law_1.2,1.3978303909301757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,1,power_law_1.2,0.7560448169708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,1,power_law_1.2,0.431609582901001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,1,power_law_1.2,1.7577600479125977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,1,power_law_1.2,0.828377628326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,1,power_law_1.2,0.43004159927368163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,1,power_law_1.2,1.0387200355529784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,1,power_law_1.2,0.44720001220703126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,1,power_law_1.2,2.4708351135253905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,1,power_law_1.2,1.1971712112426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,1,power_law_1.2,0.4629183769226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,1,power_law_1.2,0.47667841911315917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,1,power_law_1.2,1.5154239654541015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,1,power_law_1.2,0.021049599349498748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,1,power_law_1.2,3.2167232513427733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,1,power_law_1.2,0.501196813583374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,1,power_law_1.2,1.87589111328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,1,power_law_1.2,0.5603199958801269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,8,balanced,0.025258667767047882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,8,balanced,0.030773334205150604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,1,power_law_1.2,3.931897735595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,1,power_law_1.2,0.02295680046081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.2,0.606982421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,1,power_law_1.2,2.5722623825073243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,8,balanced,0.025221332907676697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.2,0.7137279987335206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,1,power_law_1.2,0.026976001262664796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,8,balanced,0.029631999631722767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,8,balanced,0.025498665869235992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,1,power_law_1.2,3.258099365234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,1,power_law_1.2,6.105190277099609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.2,0.8022975921630859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,8,balanced,0.031343999008337654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,8,balanced,0.03293866664171219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,1,power_law_1.2,0.03612799942493439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,8,balanced,0.051818668842315674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,8,balanced,0.033674667278925575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.2,1.0037759780883788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,1,power_law_1.2,0.05798400044441223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,8,balanced,0.049509331583976746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,8,balanced,0.037808001041412354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,1,power_law_1.2,3.943404769897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,8,balanced,0.04829333225886027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,1,power_law_1.2,0.08015999794006348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.2,1.2064831733703614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,8,balanced,0.039664000272750854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,1,power_law_1.2,11.672441864013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,8,balanced,0.048570667703946434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,8,balanced,0.04084266722202301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,1,power_law_1.2,0.10145280361175538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,8,balanced,0.05000533163547516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.2,1.6035327911376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,8,balanced,0.039520000418027244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,1,power_law_1.2,5.946406555175781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,8,balanced,0.050160000721613564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,1,power_law_1.2,0.11002240180969239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,8,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.2,2.010969543457031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,8,balanced,0.05266133447488149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,8,balanced,0.04016000032424927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,8,balanced,0.05354666709899902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,1,power_law_1.2,0.11609599590301514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,8,balanced,0.03988266736268997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.2,2.8008256912231446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,8,balanced,0.05584533512592316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,8,balanced,0.041536000867684685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,1,power_law_1.2,11.288524627685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,8,balanced,0.05747200051943461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,1,power_law_1.2,0.1193727970123291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.2,3.5895103454589843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,8,balanced,0.04121066629886627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,8,balanced,0.061893333991368614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,8,balanced,0.04168533285458883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,8,balanced,0.06172266602516174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,1,power_law_1.2,0.1260607957839966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,1,power_law_1.2,4.388953781127929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,8,balanced,0.049882665276527405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,8,balanced,0.0629120022058487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,1,power_law_1.2,0.1315392017364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,8,balanced,0.049365331729253135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,8,balanced,0.06630933284759521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,1,power_law_1.2,6.7780609130859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,8,balanced,0.05017599960168203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,1,power_law_1.2,0.13610880374908446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,8,balanced,0.0757066657145818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,8,balanced,0.05504000186920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,1,power_law_1.2,0.14275840520858765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,8,balanced,0.08069866895675659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,1,power_law_1.2,13.1608642578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,1,power_law_1.2,0.1461824059486389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,8,balanced,0.061253334085146584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,8,balanced,0.07941866914431255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,1,power_law_1.2,0.1493824005126953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,1,power_law_1.2,0.16005760431289673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,8,balanced,0.06968533496061961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,1,power_law_1.2,0.1874624013900757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,8,balanced,0.0929813285668691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.2,0.2026815891265869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,8,balanced,0.08186133205890656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.2,0.24728960990905763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.2,0.2874943971633911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.2,0.37828478813171384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,8,balanced,0.10256000359853108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,8,balanced,0.10321600238482158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.2,0.3910720109939575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.2,0.5142975807189941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.2,0.680025577545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,8,balanced,0.12203199664751689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,8,balanced,0.1285706659158071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.2,0.9132160186767578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.2,1.1790592193603515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,1,power_law_1.2,1.4460224151611327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,1,power_law_1.2,2.227622413635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,8,balanced,0.15452800194422403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,8,balanced,0.16305599610010782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,1,power_law_1.2,4.340422439575195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,8,balanced,0.19547200202941895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,8,balanced,0.21068267027537027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,8,balanced,0.01929066702723503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,8,balanced,0.033674667278925575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,8,balanced,0.019999999552965164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,8,balanced,0.03347733368476232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,8,balanced,0.2664480010668437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,8,balanced,0.019738666713237762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,8,balanced,0.033301333586374916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,8,balanced,0.021551998953024547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,8,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,8,balanced,0.023376000424226124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,8,balanced,0.04174399872620901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,8,balanced,0.025263999899228413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,8,balanced,0.043509334325790405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,8,balanced,0.02958400050799052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,8,balanced,0.0454773356517156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,8,balanced,0.028677334388097126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,8,balanced,0.04359999795754751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,8,balanced,0.037578667203585304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,8,balanced,0.04354666670163473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,8,balanced,0.0369759996732076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,8,balanced,0.043791999419530235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,8,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,8,balanced,0.04561600089073181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,8,balanced,0.04633066554864248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,8,balanced,0.24732800324757895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,8,balanced,0.046015997727712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,8,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,8,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,8,balanced,0.045754666129748024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,8,balanced,0.06001600126425425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,8,balanced,0.047781333327293396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,8,balanced,0.3320373296737671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,8,balanced,0.05696000158786774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,8,balanced,0.051957334081331887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,8,balanced,0.053717335065205894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,8,balanced,0.05490666627883911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,8,balanced,0.052005335688591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,8,balanced,0.06235733131567637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,8,balanced,0.052101333936055504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,8,balanced,0.055999999245007835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,8,balanced,0.06726933519045512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,8,balanced,0.06032533446947733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,8,balanced,0.07702399790287018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,8,balanced,0.06381333371003468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,8,balanced,0.29393066962560016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,8,balanced,0.09107733766237895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,8,balanced,0.07046933472156525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,8,balanced,0.10702932874361674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,8,balanced,0.09303466478983562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,8,balanced,0.4126666784286499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,8,balanced,0.1125973363717397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,8,balanced,0.1560479998588562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,8,balanced,0.14205867052078247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,8,balanced,0.18420799573262533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,8,balanced,0.17673067251841226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,8,balanced,0.4360106786092122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,8,balanced,0.2595626711845398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,8,balanced,0.6303466558456421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,8,balanced,0.2355466683705648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,8,balanced,0.3375200033187866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,8,balanced,0.286682665348053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,8,balanced,0.4148266712824504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,8,power_law_1.01,0.0344895988702774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,8,balanced,0.353333314259847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,8,power_law_1.01,0.030105599761009218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,8,power_law_1.01,0.03081600069999695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,8,power_law_1.01,0.02995840013027191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,8,power_law_1.01,0.029318401217460634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,8,power_law_1.01,0.030963200330734252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,8,power_law_1.01,0.03383679986000061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,8,power_law_1.01,0.036262398958206175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,8,power_law_1.01,0.03333120048046112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,8,power_law_1.01,0.03989120125770569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,8,power_law_1.01,0.03736959993839264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,8,power_law_1.01,0.04565759897232056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,8,power_law_1.01,0.0389055997133255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,8,power_law_1.01,0.047302401065826415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,8,power_law_1.01,0.03943679928779602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,8,power_law_1.01,0.04834559857845307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,8,power_law_1.01,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,8,power_law_1.01,0.04852479994297028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,8,power_law_1.01,0.040268799662590025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,8,balanced,0.8023733297983805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,8,power_law_1.01,0.05066239833831787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,8,power_law_1.01,0.042259201407432556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,8,power_law_1.01,0.055244797468185426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,8,power_law_1.01,0.04344319999217987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,8,power_law_1.01,0.04368639886379242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,8,power_law_1.01,0.058297598361968996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,8,power_law_1.01,0.046323201060295104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,8,power_law_1.01,0.0608959972858429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,8,power_law_1.01,0.04935680031776428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,8,power_law_1.01,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,8,power_law_1.01,0.05496320128440857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,8,power_law_1.01,0.05699840188026428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,8,power_law_1.01,0.056966400146484374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,8,power_law_1.01,0.03256320059299469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,8,power_law_1.01,0.06227840185165405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,8,balanced,1.196778694788615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,8,balanced,0.6487626632054647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,8,power_law_1.01,0.05856000185012818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,8,power_law_1.01,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,8,power_law_1.01,0.06298879981040954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,8,power_law_1.01,0.06488320231437683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,8,power_law_1.01,0.0674560010433197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,8,power_law_1.01,0.03727999925613403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,8,power_law_1.01,0.07470719814300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,8,power_law_1.01,0.07081599831581116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,8,power_law_1.01,0.04214400053024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,8,power_law_1.01,0.08538240194320679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,8,power_law_1.01,0.09088640213012696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,8,power_law_1.01,0.04291839897632599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,8,power_law_1.01,0.09260159730911255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,8,power_law_1.01,0.09731199741363525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,8,power_law_1.01,0.04283519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,8,power_law_1.01,0.09508479833602905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,8,power_law_1.01,0.04330880045890808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,8,power_law_1.01,0.11407999992370606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,8,power_law_1.01,0.12471040487289428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,8,power_law_1.01,0.13441280126571656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,8,power_law_1.01,0.04327679872512817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,8,power_law_1.01,0.1561792016029358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,8,balanced,0.5373226801554362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,8,power_law_1.01,0.04408319890499115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,8,power_law_1.01,0.1764032006263733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,8,power_law_1.01,0.2113856077194214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,8,power_law_1.01,0.045388799905776975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,8,power_law_1.01,0.21838080883026123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,8,power_law_1.01,0.2640000104904175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,8,power_law_1.01,0.04653440117835998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,8,power_law_1.01,0.30291199684143066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,8,power_law_1.01,0.048147198557853696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,8,power_law_1.01,0.4027200222015381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,8,power_law_1.01,0.05287039875984192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,8,power_law_1.01,0.39609599113464355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,8,power_law_1.01,0.49979519844055176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,8,power_law_1.01,0.055212801694869994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,8,power_law_1.01,0.45569920539855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,8,power_law_1.01,0.01785600036382675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,8,power_law_1.01,0.0581055998802185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,8,power_law_1.01,0.7815552234649659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,8,power_law_1.01,0.6651904106140136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,8,power_law_1.01,0.06524159908294677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,8,power_law_1.01,0.01886080056428909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,8,power_law_1.01,1.3363327980041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,8,power_law_1.01,0.07070080041885377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,8,power_law_1.01,0.8960063934326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,8,power_law_1.01,0.018911999464035035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,8,power_law_1.01,0.08399360179901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,8,power_law_1.01,0.019167999923229217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,8,power_law_1.01,0.09578880071640014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,8,power_law_1.01,1.9055360794067382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,8,power_law_1.01,0.019545599818229675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,8,power_law_1.01,0.11348479986190796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,8,power_law_1.01,0.02351360023021698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,8,power_law_1.01,0.14362239837646484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,8,power_law_1.01,0.0253248006105423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,8,power_law_1.01,0.025356799364089966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,8,power_law_1.01,0.18172800540924072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,8,power_law_1.01,0.031411200761795044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,8,power_law_1.01,0.25827200412750245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,8,power_law_1.01,0.031564798951148984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,8,power_law_1.01,0.029868799448013305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,8,power_law_1.01,0.3412544012069702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,8,power_law_1.01,0.044915199279785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,8,power_law_1.01,0.45839362144470214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,8,power_law_1.01,0.04444800019264221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,8,power_law_1.01,0.5102464199066162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,8,power_law_1.01,0.04581120014190674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,8,power_law_1.01,0.8052607536315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,8,power_law_1.01,0.050432002544403075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,8,power_law_1.01,0.04991999864578247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,8,power_law_1.01,0.05135999917984009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,8,power_law_1.01,1.727872085571289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,8,power_law_1.01,0.048089599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,8,power_law_1.01,0.0486272007226944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,8,power_law_1.01,0.05672320127487183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,8,power_law_1.01,0.06000000238418579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,8,power_law_1.01,0.07441279888153077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,8,balanced,1.2750399907430012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,8,power_law_1.01,0.08874880075454712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,8,power_law_1.01,0.1115007996559143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,8,power_law_1.01,0.14257279634475709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,8,power_law_1.01,0.19689600467681884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,8,power_law_1.01,0.2545151948928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,8,power_law_1.01,0.3103679895401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,8,balanced,1.0145706335703533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,8,power_law_1.01,0.46211838722229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,8,power_law_1.01,0.9193087577819824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,8,power_law_1.2,0.029286399483680725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,8,power_law_1.2,0.02602880001068115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,8,power_law_1.2,0.031865599751472476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,8,power_law_1.2,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,8,power_law_1.2,0.03850240111351013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,8,power_law_1.2,0.045977601408958436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,8,power_law_1.2,0.045132800936698914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,8,power_law_1.2,0.04922879934310913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,8,power_law_1.2,0.04963200092315674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,8,power_law_1.2,0.03447679877281189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,8,power_law_1.2,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,8,power_law_1.2,0.02924799919128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,8,power_law_1.2,0.056454402208328244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,8,power_law_1.2,0.02964479923248291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,8,power_law_1.2,0.05984640121459961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,8,power_law_1.2,0.03209599852561951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,8,power_law_1.2,0.059552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,8,power_law_1.2,0.03364480137825012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,8,power_law_1.2,0.058336001634597776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,8,power_law_1.2,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,8,power_law_1.2,0.06405119895935059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,8,power_law_1.2,0.03834240138530731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,8,power_law_1.2,0.06618880033493042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,8,power_law_1.2,0.04289279878139496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,8,power_law_1.2,0.03991680145263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,8,power_law_1.2,0.06693120002746582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,8,power_law_1.2,0.040294399857521056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,8,power_law_1.2,0.03292160034179688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,8,power_law_1.2,0.07454079985618592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,8,power_law_1.2,0.03294079899787903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,8,power_law_1.2,0.0411327987909317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,8,power_law_1.2,0.08866559863090515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,8,power_law_1.2,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,8,power_law_1.2,0.04321280121803284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,8,power_law_1.2,0.0945088028907776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,8,power_law_1.2,0.03689599931240082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,8,power_law_1.2,0.042803201079368594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,8,power_law_1.2,0.10325759649276733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,8,power_law_1.2,0.042233601212501526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,8,power_law_1.2,0.04574080109596253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,8,power_law_1.2,0.04273279905319214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,8,power_law_1.2,0.12664320468902587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,8,power_law_1.2,0.042438399791717527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,8,power_law_1.2,0.05023999810218811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,8,power_law_1.2,0.15555839538574218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,8,power_law_1.2,0.04362240135669708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,8,power_law_1.2,0.05586559772491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,8,power_law_1.2,0.2074431896209717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,8,power_law_1.2,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,8,power_law_1.2,0.05787519812583923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,8,power_law_1.2,0.043942400813102724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,8,power_law_1.2,0.2689023971557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,8,power_law_1.2,0.05848960280418396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,8,power_law_1.2,0.04620159864425659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,8,power_law_1.2,0.3560767889022827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,8,power_law_1.2,0.06803839802742004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,8,power_law_1.2,0.04762240052223206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,8,power_law_1.2,0.07802240252494812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,8,power_law_1.2,0.04812160134315491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,8,power_law_1.2,0.4338047981262207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,8,power_law_1.2,0.09196799993515015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,8,power_law_1.2,0.053235197067260744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,8,power_law_1.2,0.5942272186279297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,8,power_law_1.2,0.11439360380172729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,8,power_law_1.2,0.05488640069961548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,8,power_law_1.2,0.06072319746017456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,8,power_law_1.2,0.14497920274734497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,8,power_law_1.2,0.9762944221496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,8,power_law_1.2,0.06689919829368592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,8,power_law_1.2,0.16766719818115233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,8,power_law_1.2,0.07758079767227173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,8,power_law_1.2,0.018668800592422485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,8,power_law_1.2,0.2514944076538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,8,power_law_1.2,2.0111040115356444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,8,power_law_1.2,0.08821120262145996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,8,power_law_1.2,0.30536320209503176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,8,power_law_1.2,0.10081919431686401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,8,power_law_1.2,0.018348799645900728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,8,power_law_1.2,0.5285056114196778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,8,power_law_1.2,0.12711039781570435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,8,power_law_1.2,0.01850239932537079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,8,power_law_1.2,0.6131199836730957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,8,power_law_1.2,0.1577855944633484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,8,power_law_1.2,0.019257600605487823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,8,power_law_1.2,0.22688639163970947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,8,power_law_1.2,0.846406364440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,8,power_law_1.2,0.020339199900627138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,8,power_law_1.2,0.2855999946594238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,8,power_law_1.2,0.02309119999408722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,8,power_law_1.2,1.3171584129333496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,8,power_law_1.2,0.025414401292800905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,8,power_law_1.2,0.35034880638122556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,8,power_law_1.2,0.02526719868183136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,8,power_law_1.2,2.4360319137573243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,8,power_law_1.2,0.6477759838104248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,4,balanced,0.023333333432674408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,8,power_law_1.2,0.03144960105419159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,4,balanced,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,8,power_law_1.2,0.7180928230285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,8,power_law_1.2,0.03134079873561859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,4,balanced,0.02534399926662445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,8,power_law_1.2,0.02881920039653778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,8,power_law_1.2,1.1450559616088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,4,balanced,0.029578665892283123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,4,balanced,0.029919999341169994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,8,power_law_1.2,0.04498560130596161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,4,balanced,0.03133866687615713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,4,balanced,0.03162133445342382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,8,power_law_1.2,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,4,balanced,0.06208533545335134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,8,power_law_1.2,2.7005504608154296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,8,power_law_1.2,0.04254080057144165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,4,balanced,0.035989334185918175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,4,balanced,0.05333333214124044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,4,balanced,0.041509332756201424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,8,power_law_1.2,0.04937599897384644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,4,balanced,0.04165866722663244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,4,balanced,0.05237866441408793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,8,power_law_1.2,0.049344000220298764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,4,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,4,balanced,0.053445334235827126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,8,power_law_1.2,0.052083200216293334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,4,balanced,0.04173333446184794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,4,balanced,0.05460800230503082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,8,power_law_1.2,0.04652799963951111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,4,balanced,0.04189866781234741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,4,balanced,0.053914666175842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,8,power_law_1.2,0.050342398881912234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,4,balanced,0.04190933207670847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,4,balanced,0.05444266895453135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,8,power_law_1.2,0.05952640175819397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,4,balanced,0.04197333256403605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,4,balanced,0.05580799778302511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,8,power_law_1.2,0.06202239990234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,4,balanced,0.045781334241231285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,4,balanced,0.056186666091283165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,8,power_law_1.2,0.08007680177688599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,4,balanced,0.06020266811052958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,8,power_law_1.2,0.09578880071640014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,4,balanced,0.04380266865094503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,4,balanced,0.06577066580454509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,8,power_law_1.2,0.12550400495529174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,4,balanced,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,4,balanced,0.06413866579532623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,4,balanced,0.03601066768169403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,8,power_law_1.2,0.15345920324325563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,4,balanced,0.054154664278030396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,4,balanced,0.017162666966517765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,8,power_law_1.2,0.21095681190490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,4,balanced,0.05323199927806854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,4,balanced,0.067930668592453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,4,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,4,balanced,0.019023999571800232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,8,power_law_1.2,0.2995007991790771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,4,balanced,0.05589333176612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,4,balanced,0.03669866671164831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,4,balanced,0.0199946661790212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,4,balanced,0.070592001080513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,8,power_law_1.2,0.34729599952697754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,4,balanced,0.021327999730904896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,4,balanced,0.041834667325019836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,4,balanced,0.06411199768384297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,4,balanced,0.04816000163555145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,4,balanced,0.023045333723227184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,8,power_law_1.2,0.5298431873321533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,4,balanced,0.025248001019159954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,4,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,4,balanced,0.07456533114115398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,4,balanced,0.026074667771657307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,4,balanced,0.08298133313655853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,4,balanced,0.050245334704717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,4,balanced,0.026816000541051228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,4,balanced,0.04980266590913137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,8,power_law_1.2,1.0732352256774902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,4,balanced,0.08781866232554118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,4,balanced,0.029743999242782593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,4,balanced,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,4,balanced,0.09108266234397888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,4,balanced,0.05034666756788889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,4,balanced,0.0295413335164388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,4,balanced,0.10125866532325745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,4,balanced,0.052000001072883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,4,balanced,0.029114666084448498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,4,balanced,0.03737066686153412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,4,balanced,0.05349866549173991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,4,balanced,0.035749333600203194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,4,balanced,0.09231999516487122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,4,balanced,0.035360001027584076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,4,balanced,0.05407466491063436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,4,balanced,0.0496319979429245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,4,balanced,0.05542933444182078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,4,balanced,0.13148799538612366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,4,balanced,0.04914666712284088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,4,balanced,0.06035733222961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,4,balanced,0.113946666320165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,4,balanced,0.049653331438700356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,4,balanced,0.06213866670926412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,4,balanced,0.06159466505050659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,4,balanced,0.06223999957243601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,4,balanced,0.16075199842453003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,4,balanced,0.06237333516279856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,4,balanced,0.06854933500289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,4,balanced,0.07481066882610321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,4,balanced,0.1346826652685801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,4,balanced,0.07528000076611836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,4,balanced,0.07801066835721333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,4,balanced,0.08519466718037923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,4,balanced,0.20960533618927002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,4,balanced,0.08418666323026021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,4,balanced,0.09949866930643718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,4,balanced,0.17652267217636108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,4,balanced,0.13265066345532736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,4,balanced,0.09354133407274882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,4,balanced,0.25942933559417725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,4,balanced,0.15994667013486227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,4,balanced,0.21158933639526367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,4,balanced,0.142085333665212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,4,balanced,0.21488000949223837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,4,balanced,0.16169066230456033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,4,balanced,0.37642133235931396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,4,balanced,0.2932213346163432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,4,balanced,0.2764853239059448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,4,balanced,0.22794665892918906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,4,balanced,0.38758401075998944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,4,balanced,0.3647199869155884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,4,balanced,0.488597313563029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,4,balanced,0.29345067342122394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,4,balanced,0.5019733508427938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,4,balanced,0.44325868288675946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,4,balanced,0.6097919940948486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,4,balanced,0.3590826590855916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,4,balanced,0.6234826644261678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,4,power_law_1.01,0.026668798923492432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,4,power_law_1.01,0.032876798510551454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,4,power_law_1.01,0.02849920094013214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,4,power_law_1.01,0.028217598795890808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,4,power_law_1.01,0.031379199028015135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,4,balanced,0.6674933433532715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,4,power_law_1.01,0.029395198822021483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,4,power_law_1.01,0.03277440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,4,balanced,0.9485546747843424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,4,power_law_1.01,0.03456639945507049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,4,power_law_1.01,0.03891839981079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,4,power_law_1.01,0.03583999872207642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,4,power_law_1.01,0.04818559885025024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,4,power_law_1.01,0.039315199851989745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,4,power_law_1.01,0.04816640019416809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,4,power_law_1.01,0.040761598944664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,4,power_law_1.01,0.049644801020622256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,4,power_law_1.01,0.039852800965309146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,4,balanced,0.5561973253885905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,4,power_law_1.01,0.051545602083206174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,4,power_law_1.01,0.041094401478767396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,4,power_law_1.01,0.05243520140647888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,4,power_law_1.01,0.0412992000579834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,4,power_law_1.01,0.05548160076141358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,4,power_law_1.01,0.04519680142402649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,4,power_law_1.01,0.05807999968528747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,4,power_law_1.01,0.047391998767852786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,4,power_law_1.01,0.06321280002593994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,4,power_law_1.01,0.049798399209976196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,4,power_law_1.01,0.06517760157585144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,4,power_law_1.01,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,4,power_law_1.01,0.04059520065784454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,4,power_law_1.01,0.06768640279769897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,4,power_law_1.01,0.05677440166473389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,4,balanced,0.966480016708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,4,power_law_1.01,0.0711359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,4,power_law_1.01,0.0597055971622467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,4,power_law_1.01,0.034118399024009705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,4,power_law_1.01,0.07175679802894593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,4,power_law_1.01,0.06356480121612548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,4,power_law_1.01,0.07427200078964233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,4,power_law_1.01,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,4,power_law_1.01,0.07396479845046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,4,power_law_1.01,0.09104639887809754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,4,power_law_1.01,0.03973760008811951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,4,power_law_1.01,0.08841599822044373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,4,power_law_1.01,0.10017280578613282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,4,power_law_1.01,0.04105600118637085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,4,power_law_1.01,0.1026047945022583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,4,power_law_1.01,0.10559359788894654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,4,power_law_1.01,0.0468095988035202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,4,power_law_1.01,0.13787519931793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,4,power_law_1.01,0.11544320583343506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,4,power_law_1.01,0.04812160134315491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,4,power_law_1.01,0.1729151964187622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,4,power_law_1.01,0.14368640184402465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,4,power_law_1.01,0.04888319969177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,4,power_law_1.01,0.23128960132598878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,4,power_law_1.01,0.19230079650878906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,4,power_law_1.01,0.2598207950592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,4,power_law_1.01,0.04929920136928558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,4,power_law_1.01,0.2672384023666382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,4,power_law_1.01,0.37969920635223386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,4,power_law_1.01,0.04978559911251068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,4,power_law_1.01,0.3238719940185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,4,power_law_1.01,0.489408016204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,4,power_law_1.01,0.050963199138641356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,4,power_law_1.01,0.48326401710510253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,4,power_law_1.01,0.053420799970626834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,4,power_law_1.01,0.5451776027679444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,4,power_law_1.01,0.05422719717025757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,4,power_law_1.01,0.6279232025146484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,4,power_law_1.01,0.9095680236816406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,4,power_law_1.01,0.05809280276298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,4,power_law_1.01,0.7803711891174316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,4,power_law_1.01,0.061913597583770755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,4,power_law_1.01,1.680780792236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,4,power_law_1.01,1.1721471786499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,4,power_law_1.01,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,4,power_law_1.01,0.06618239879608154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,4,power_law_1.01,0.07386239767074584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,4,power_law_1.01,2.242681694030762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,4,power_law_1.01,0.08489599823951721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,4,power_law_1.01,0.10339200496673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,4,power_law_1.01,0.11726720333099365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,4,power_law_1.01,0.16369919776916503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,4,power_law_1.01,0.18575359582901002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,4,power_law_1.01,0.26445438861846926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,4,power_law_1.01,0.35710721015930175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,4,power_law_1.01,0.4957119941711426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,4,power_law_1.01,0.6508992195129395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,4,power_law_1.01,0.8839679718017578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,4,power_law_1.01,0.01759999990463257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,4,power_law_1.01,1.405619239807129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,4,power_law_1.01,0.017401599884033205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,4,power_law_1.01,2.423360061645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,4,power_law_1.01,0.018118399381637573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,4,power_law_1.01,0.019436800479888917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,4,power_law_1.01,0.02008959949016571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,4,power_law_1.01,0.02106879949569702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,4,balanced,1.2709440390268962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,4,power_law_1.01,0.022752000391483305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,4,balanced,1.845861275990804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,4,power_law_1.01,0.022988800704479218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,4,power_law_1.01,0.02537600100040436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,4,balanced,1.0900373458862305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,4,power_law_1.01,0.02619520127773285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,4,power_law_1.01,0.02529279887676239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,4,power_law_1.01,0.030700799822807313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,4,power_law_1.01,0.03020159900188446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,4,balanced,1.8918933868408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,4,power_law_1.01,0.029747200012207032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,4,power_law_1.01,0.045465600490570066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,4,power_law_1.01,0.04695680141448975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,4,power_law_1.01,0.04860160052776337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,4,power_law_1.01,0.053337597846984865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,4,power_law_1.01,0.05398399829864502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,4,power_law_1.01,0.058246397972106935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,4,power_law_1.01,0.0637440025806427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,4,power_law_1.01,0.07360640168190002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,4,power_law_1.01,0.0867904007434845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,4,power_law_1.01,0.11447679996490479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,4,power_law_1.01,0.13815679550170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,4,power_law_1.01,0.19570560455322267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,4,power_law_1.01,0.24200959205627443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,4,power_law_1.01,0.2941375970840454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,4,power_law_1.01,0.4385663986206055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,4,power_law_1.01,0.939129638671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,4,power_law_1.2,0.026169601082801818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,4,power_law_1.2,0.03294720053672791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,4,power_law_1.2,0.023494400084018707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,4,power_law_1.2,0.028454399108886717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,4,power_law_1.2,0.02974080145359039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,4,power_law_1.2,0.029606398940086365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,4,power_law_1.2,0.04035840034484863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,4,power_law_1.2,0.032793599367141726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,4,power_law_1.2,0.03287039995193482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,4,power_law_1.2,0.03909760117530823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,4,power_law_1.2,0.03470079898834229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,4,power_law_1.2,0.034195199608802795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,4,power_law_1.2,0.04832000136375427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,4,power_law_1.2,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,4,power_law_1.2,0.04862079918384552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,4,power_law_1.2,0.03553920090198517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,4,power_law_1.2,0.040006399154663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,4,power_law_1.2,0.05127679705619812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,4,power_law_1.2,0.038380798697471616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,4,power_law_1.2,0.04046080112457275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,4,power_law_1.2,0.053235197067260744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,4,power_law_1.2,0.04113920032978058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,4,power_law_1.2,0.054124802350997925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,4,power_law_1.2,0.04132480025291443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,4,power_law_1.2,0.04705280065536499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,4,power_law_1.2,0.05441920161247253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,4,power_law_1.2,0.04174720048904419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,4,power_law_1.2,0.04776319861412048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,4,power_law_1.2,0.05923200249671936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,4,power_law_1.2,0.045535999536514285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,4,power_law_1.2,0.060172802209854125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,4,power_law_1.2,0.04897280037403107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,4,power_law_1.2,0.04738560020923614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,4,power_law_1.2,0.06469119787216186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,4,power_law_1.2,0.048979198932647704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,4,power_law_1.2,0.051020801067352295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,4,power_law_1.2,0.06935679912567139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,4,power_law_1.2,0.0497408002614975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,4,power_law_1.2,0.0715071976184845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,4,power_law_1.2,0.05072640180587769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,4,power_law_1.2,0.051686400175094606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,4,power_law_1.2,0.07208960056304932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,4,power_law_1.2,0.057068800926208495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,4,power_law_1.2,0.0775871992111206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,4,power_law_1.2,0.0541375994682312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,4,power_law_1.2,0.061875200271606444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,4,power_law_1.2,0.09220479726791382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,4,power_law_1.2,0.05504639744758606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,4,power_law_1.2,0.06632959842681885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,4,power_law_1.2,0.10724480152130127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,4,power_law_1.2,0.058284801244735715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,4,power_law_1.2,0.07699840068817139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,4,power_law_1.2,0.10966399908065796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,4,power_law_1.2,0.0639743983745575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,4,power_law_1.2,0.09089919924736023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,4,power_law_1.2,0.14108799695968627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,4,power_law_1.2,0.06635519862174988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,2,balanced,0.03334933271010717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,4,power_law_1.2,0.1088703989982605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,4,power_law_1.2,0.17703039646148683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,2,balanced,0.025248001019159954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,4,power_law_1.2,0.06615679860115051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,4,power_law_1.2,0.016339200735092162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,4,power_law_1.2,0.12011519670486451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,4,power_law_1.2,0.2344127893447876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,2,balanced,0.03162133445342382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,4,power_law_1.2,0.07620480060577392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,4,power_law_1.2,0.1467136025428772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,4,power_law_1.2,0.3115520000457764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,2,balanced,0.025621332228183746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,2,balanced,0.03362133353948593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,4,power_law_1.2,0.08789759874343872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,4,power_law_1.2,0.017183999717235564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,2,balanced,0.03975466638803482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,2,balanced,0.027445333699385326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,4,power_law_1.2,0.4636096000671387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,4,power_law_1.2,0.20505599975585936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,4,power_law_1.2,0.11620479822158813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,4,power_law_1.2,0.018143999576568603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,2,balanced,0.04590400060017904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,2,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,4,power_law_1.2,0.2933056116104126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,4,power_law_1.2,0.6132544040679931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,4,power_law_1.2,0.1156991958618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,4,power_law_1.2,0.018611200153827667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,4,power_law_1.2,0.1726912021636963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,2,balanced,0.0655680000782013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,2,balanced,0.04766400158405304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,4,power_law_1.2,0.3770240068435669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,4,power_law_1.2,0.01940480023622513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,4,power_law_1.2,0.5681407928466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,2,power_law_1.01,0.025203201174736022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,2,balanced,0.055162668228149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,2,balanced,0.04807466765244802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,4,power_law_1.2,0.021324799954891206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,4,power_law_1.2,0.20920319557189943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,4,power_law_1.2,0.47982082366943357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,2,balanced,0.05444266895453135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,4,power_law_1.2,1.0699263572692872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,2,balanced,0.04781333108743032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,4,power_law_1.2,0.022841599583625794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,4,power_law_1.2,0.3209536075592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,2,balanced,0.055919999877611794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,2,power_law_1.01,0.026131200790405273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,4,power_law_1.2,0.7021696090698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,2,balanced,0.04842666784922282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,4,power_law_1.2,0.023155200481414794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,2,balanced,0.056757330894470215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,4,power_law_1.2,0.3901439905166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,2,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,4,power_law_1.2,0.02475520074367523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,4,power_law_1.2,2.239014434814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,2,power_law_1.01,0.027110400795936584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,4,power_law_1.2,0.9514816284179688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,2,balanced,0.057914664347966514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,2,balanced,0.04837333162625631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,4,power_law_1.2,0.02565760016441345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,4,power_law_1.2,0.624729585647583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,2,power_law_1.01,0.03669120073318481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,2,balanced,0.056517332792282104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,2,balanced,0.051882664362589516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,4,power_law_1.2,0.024851199984550477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,4,power_law_1.2,1.456057643890381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,2,balanced,0.057760000228881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,2,balanced,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,2,power_law_1.01,0.03976320028305054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,4,power_law_1.2,0.8618176460266114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,4,power_law_1.2,0.02942720055580139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,2,balanced,0.05936000247796377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,2,power_law_1.01,0.04874880015850067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,2,balanced,0.0561706672112147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,4,power_law_1.2,0.02963840067386627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,4,power_law_1.2,1.02542724609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,2,power_law_1.01,0.050444799661636355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,2,balanced,0.05964266757170359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,4,power_law_1.2,0.029504001140594482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,2,balanced,0.06673066814740498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,4,power_law_1.2,3.2251327514648436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,2,power_law_1.01,0.05422719717025757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,2,balanced,0.06529599924882253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,4,power_law_1.2,0.04650880098342895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,4,power_law_1.2,1.670854377746582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,2,balanced,0.06648000081380208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,4,power_law_1.2,0.047270399332046506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,2,balanced,0.06574933230876923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,2,power_law_1.01,0.053830397129058835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,4,power_law_1.2,0.04885759949684143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,2,balanced,0.07044800122578938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,4,power_law_1.2,3.8681983947753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,2,balanced,0.06815466781457265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,4,power_law_1.2,0.05123199820518494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,2,balanced,0.08568533261617024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,2,power_law_1.01,0.05671679973602295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,4,power_law_1.2,0.05324800014495849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,2,balanced,0.0841919978459676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,4,power_law_1.2,0.059462398290634155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,2,power_law_1.01,0.03260799944400787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,2,balanced,0.09333866834640503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,2,power_law_1.01,0.058508801460266116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,4,power_law_1.2,0.06441599726676941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,2,balanced,0.09380267063776652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,4,power_law_1.2,0.07545599937438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,2,power_law_1.01,0.029420799016952513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,2,power_law_1.01,0.05984640121459961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,2,balanced,0.10541866223017375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,4,power_law_1.2,0.0930239975452423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,2,balanced,0.11800533533096313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,2,power_law_1.01,0.03249920010566711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,4,power_law_1.2,0.11910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,2,power_law_1.01,0.060787200927734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,2,balanced,0.13499733805656433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,2,balanced,0.11709866921106975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,2,power_law_1.01,0.03578880131244659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,2,power_law_1.01,0.06294400095939637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,4,power_law_1.2,0.1468608021736145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,4,power_law_1.2,0.19971840381622313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,2,power_law_1.01,0.037868800759315493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,2,power_law_1.01,0.06813439726829529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,4,power_law_1.2,0.2594815969467163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,2,balanced,0.1818293333053589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,2,balanced,0.15050666530927023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,2,power_law_1.01,0.045491200685501096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,2,power_law_1.01,0.07270399928092956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,4,power_law_1.2,0.3079103946685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,2,power_law_1.01,0.046623998880386354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,2,power_law_1.01,0.08025599718093872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,4,power_law_1.2,0.45395841598510744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,2,balanced,0.2179786761601766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,2,power_law_1.01,0.04730879962444305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,2,power_law_1.01,0.09279999732971192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,4,power_law_1.2,0.9252223968505859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,2,power_law_1.01,0.04790399968624115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,2,power_law_1.01,0.09635199904441834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,2,balanced,0.17539199193318686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,2,power_law_1.01,0.04937599897384644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,2,power_law_1.01,0.11211520433425903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,2,power_law_1.01,0.051609599590301515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,2,power_law_1.01,0.13033599853515626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,2,power_law_1.01,0.05429760217666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,2,balanced,0.31251732508341473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,2,power_law_1.01,0.1712831974029541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,2,power_law_1.01,0.05687680244445801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,2,power_law_1.01,0.20767359733581542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,2,power_law_1.01,0.061791998147964475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,2,power_law_1.01,0.29092481136322024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,2,balanced,0.2529226740201314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,2,power_law_1.01,0.06621440052986145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,2,power_law_1.01,0.3732736110687256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,2,power_law_1.01,0.07137280106544494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,2,power_law_1.01,0.07411839962005615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,2,power_law_1.01,0.509503984451294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,2,power_law_1.01,0.09009280204772949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,2,power_law_1.01,0.6910848140716552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,2,balanced,0.4113973379135132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,2,power_law_1.01,0.10282880067825317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,2,power_law_1.01,0.7900928020477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,2,power_law_1.01,0.1282431960105896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,2,power_law_1.01,1.239904022216797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,2,balanced,0.3285973270734151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,2,power_law_1.01,0.1502400040626526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,2,power_law_1.01,0.18825600147247315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,2,power_law_1.01,2.3836032867431642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,2,power_law_1.01,0.23571839332580566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,2,balanced,0.6106346845626831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,2,power_law_1.01,0.35900800228118895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,2,power_law_1.01,0.48480639457702634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,2,balanced,0.4657919804255168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,2,power_law_1.01,0.7065152168273926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,2,power_law_1.01,0.924403190612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,2,power_law_1.01,1.0743680000305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,2,power_law_1.01,1.878598403930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,2,power_law_1.01,3.3837310791015627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,2,balanced,0.8156106472015381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,2,balanced,0.6043893496195475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,2,balanced,1.0227946440378826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,2,balanced,0.7483200232187907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,2,balanced,1.6211199760437012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,2,balanced,1.1608479817708333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,8,balanced,0.059114664793014526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,8,balanced,0.062037333846092224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,8,balanced,0.058335999647776283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,8,balanced,0.06192533175150553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,8,balanced,0.06286933521429698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,8,balanced,0.061936000982920326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,8,balanced,0.06277333199977875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,8,balanced,0.06438399851322174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,8,balanced,0.06669333577156067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,8,balanced,0.06603733201821645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,8,balanced,0.037632000943024956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,8,balanced,0.06462400158246358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,8,balanced,0.06564799944559734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,8,balanced,0.03805333375930786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,8,balanced,0.037418665985266365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,8,balanced,0.0645546664794286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,8,balanced,0.03965866565704346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,8,balanced,0.06640000144640605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,8,balanced,0.03807466725508372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,8,balanced,0.07175999879837036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,8,balanced,0.039749334255854286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,8,balanced,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,8,balanced,0.04159466673930486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,8,balanced,0.07675733168919881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,8,balanced,0.040661332507928215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,8,balanced,0.08675733208656311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,8,balanced,0.040234667559464775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,8,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,8,balanced,0.08758933345476787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,8,balanced,0.042965332667032875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,8,balanced,0.11545600493748982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,8,balanced,0.04465066889921824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,8,balanced,0.043978666265805565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,8,balanced,0.11834667126337688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,8,balanced,0.04789333542188009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,8,balanced,0.15199466546376547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,8,balanced,0.04037333279848099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,8,balanced,0.05620799958705902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,8,balanced,0.05783999959627787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,8,balanced,0.17493865887324014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,8,balanced,0.06001066664854685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,8,balanced,0.040618665516376495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,8,balanced,0.07227733234564464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,8,balanced,0.040421334405740104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,8,balanced,0.23057067394256592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,8,balanced,0.07869333525498708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,2,power_law_1.2,0.03306879997253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,8,balanced,0.041637333730856575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,2,power_law_1.2,0.026790401339530943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,8,balanced,0.04171733558177948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,8,balanced,0.0983893374602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,8,balanced,0.0418453315893809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,8,balanced,0.281658669312795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,2,power_law_1.2,0.029580798745155335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,8,balanced,0.10571733117103577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,8,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,2,power_law_1.2,0.03167999982833862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,8,balanced,0.04358933369318644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,2,power_law_1.2,0.02597759962081909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,8,balanced,0.04353600243727366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,8,balanced,0.13423466682434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,2,power_law_1.2,0.03484799861907959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,8,balanced,0.04378133515516917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,8,balanced,0.39445332686106366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,2,balanced,3.0761388142903647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,2,power_law_1.2,0.037350401282310486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,2,power_law_1.2,0.02895359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,8,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,8,balanced,0.15922133127848306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,2,power_law_1.2,0.044633600115776065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,8,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,2,power_law_1.2,0.035750401020050046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,2,power_law_1.2,0.047007998824119566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,8,balanced,0.04375466704368591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,8,balanced,0.20816532770792642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,8,balanced,0.04574400186538696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,2,power_law_1.2,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,2,power_law_1.2,0.04028159976005554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,8,balanced,0.5008639891942342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,8,balanced,0.05073600014050802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,2,power_law_1.2,0.048819199204444885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,2,power_law_1.2,0.05077760219573975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,8,balanced,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,2,power_law_1.2,0.05011839866638183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,8,balanced,0.2576319972674052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,8,balanced,0.051856001218159996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,2,power_law_1.2,0.051769602298736575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,2,power_law_1.2,0.05167999863624573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,2,power_law_1.2,0.05475839972496033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,8,balanced,0.058687999844551086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,2,power_law_1.2,0.0578495979309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,2,power_law_1.2,0.05454080104827881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,8,balanced,0.0639626681804657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,8,balanced,0.6090026696523031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,2,power_law_1.2,0.061312001943588254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,2,power_law_1.2,0.055795198678970336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,8,balanced,0.36133333047231037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,2,balanced,2.1840640703837075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,8,balanced,0.07654400169849396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,2,power_law_1.2,0.06830080151557923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,2,power_law_1.2,0.056883198022842404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,8,balanced,0.023525332411130268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,2,power_law_1.2,0.07299200296401978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,8,balanced,0.08380267024040222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,2,power_law_1.2,0.05867519974708557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,2,power_law_1.2,0.07669119834899903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,8,balanced,0.02350933353106181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,2,power_law_1.2,0.06023039817810059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,2,power_law_1.2,0.08902400135993957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,8,balanced,0.45389334360758465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,8,balanced,0.10317867000897725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,2,power_law_1.2,0.10527360439300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,8,balanced,0.023232000569502514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,2,power_law_1.2,0.06064000129699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,2,power_law_1.2,0.13417600393295287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,8,balanced,0.023226665953795116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,8,balanced,0.9044586817423502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,8,balanced,0.11895466844240825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,2,power_law_1.2,0.1566848039627075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,2,power_law_1.2,0.06421120166778564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,8,balanced,0.02329600105683009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,2,power_law_1.2,0.2034303903579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,8,balanced,0.025957333544890087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,2,power_law_1.2,0.07423359751701356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,8,balanced,0.14755200346310934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,8,balanced,0.5544533332188925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,8,balanced,0.04427200059096018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,2,power_law_1.2,0.23401598930358886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,2,power_law_1.2,0.07399680018424988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,8,balanced,0.04450133442878723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,2,power_law_1.2,0.3905344009399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,2,power_law_1.2,0.0828224003314972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,8,balanced,0.04727466901143392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,8,balanced,0.18023467063903809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,2,power_law_1.2,0.5074687957763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,8,balanced,0.04359466830889384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,2,power_law_1.2,0.09381759762763978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,8,balanced,0.03786666691303253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,2,power_law_1.2,0.09794560074806213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,2,power_law_1.2,0.7496384143829345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,8,balanced,0.0652159998814265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,2,power_law_1.2,0.11886080503463745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,2,power_law_1.2,0.9780415534973145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,8,balanced,0.05765866736570994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,8,balanced,0.24225066105524698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,2,power_law_1.2,0.1312127947807312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,8,balanced,0.8372960090637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,8,balanced,0.04790933430194855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,2,power_law_1.2,1.257196807861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,2,power_law_1.2,0.17136000394821166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,8,balanced,0.0661653329928716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,2,power_law_1.2,0.20752639770507814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,8,balanced,0.05728533367315928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,8,balanced,1.7229599952697754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,2,power_law_1.2,0.32284159660339357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,2,power_law_1.2,2.016409683227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,8,balanced,0.06436799963315327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,8,balanced,0.3060906728108724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,2,power_law_1.2,0.4127295970916748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,8,balanced,0.08444266517957051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,8,balanced,0.10134933392206828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,2,power_law_1.2,0.5341951847076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,2,power_law_1.2,3.8780353546142576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,8,balanced,0.1399626632531484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,2,power_law_1.2,0.7838335990905761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,8,balanced,0.1730239987373352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,8,balanced,0.37452268600463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,2,power_law_1.2,0.9431872367858887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,8,balanced,0.24174932638804117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,8,balanced,1.6157760620117188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,2,power_law_1.2,1.407475185394287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,8,balanced,0.30246400833129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,2,power_law_1.2,2.5023359298706054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,8,balanced,0.56386665503184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,8,balanced,0.4384320179621379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,8,balanced,0.5670719941457113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,8,balanced,0.8315786520640055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,8,balanced,1.101296027501424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,8,balanced,1.0984960397084553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,8,balanced,1.3615999221801758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,8,balanced,2.1532533963521323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,8,balanced,4.265690803527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,8,power_law_1.01,0.05726720094680786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,8,power_law_1.01,0.0591808021068573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,8,power_law_1.01,0.056531202793121335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,8,power_law_1.01,0.05986559987068176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,8,power_law_1.01,0.06072319746017456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,8,power_law_1.01,0.06349440217018128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,8,power_law_1.01,0.06869760155677795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,8,power_law_1.01,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,8,power_law_1.01,0.3174720048904419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,8,power_law_1.01,0.31661438941955566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,8,power_law_1.01,0.06700159907341004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,8,power_law_1.01,0.07725440263748169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,8,power_law_1.01,0.08025599718093872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,8,power_law_1.01,0.08775039911270141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,8,power_law_1.01,0.09662719964981079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,8,power_law_1.01,0.11932159662246704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,8,power_law_1.01,0.1120576024055481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,8,power_law_1.01,0.12303359508514404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,8,power_law_1.01,0.1688704013824463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,8,power_law_1.01,0.2036992073059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,8,power_law_1.01,0.03639039993286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,8,power_law_1.01,0.26917119026184083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,8,power_law_1.01,0.036627200245857236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,8,power_law_1.01,0.3647104024887085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,8,power_law_1.01,0.03615359961986542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,8,power_law_1.01,0.4795584201812744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,8,power_law_1.01,0.0385919988155365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,8,power_law_1.01,0.7151040077209473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,8,power_law_1.01,0.03871360123157501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,8,power_law_1.01,0.8742783546447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,8,power_law_1.01,0.039724799990653994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,8,power_law_1.01,0.04588800072669983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,8,power_law_1.01,1.3317631721496581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,8,power_law_1.01,0.048089599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,8,power_law_1.01,1.6985599517822265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,8,power_law_1.01,0.051648002862930295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,8,power_law_1.01,0.05777279734611511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,8,power_law_1.01,2.177631950378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,8,power_law_1.01,0.056390398740768434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,8,power_law_1.01,3.0975231170654296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,8,power_law_1.01,0.060844802856445314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,8,power_law_1.01,0.06322559714317322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,8,power_law_1.01,0.06483839750289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,8,power_law_1.01,6.28674545288086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,8,power_law_1.01,0.07505279779434204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,8,power_law_1.01,0.07808640003204345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,8,power_law_1.01,0.08942720293998718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,8,power_law_1.01,0.10295039415359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,8,power_law_1.01,0.12206720113754273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,8,power_law_1.01,0.1581760048866272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,8,power_law_1.01,0.17754240036010743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,8,power_law_1.01,0.2649728059768677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,8,power_law_1.01,0.03980799913406372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,8,power_law_1.01,0.2879168033599854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,8,power_law_1.01,0.023071999847888946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,8,power_law_1.01,0.47162880897521975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,8,power_law_1.01,0.040140798687934874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,8,power_law_1.01,0.040166398882865904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,8,power_law_1.01,0.582144021987915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,8,power_law_1.01,0.04085760116577149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,8,power_law_1.01,0.022201600670814513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,8,power_law_1.01,0.8613823890686035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,8,power_law_1.01,0.04154239892959595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,8,power_law_1.01,0.02147199958562851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,8,power_law_1.01,1.0416831970214844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,8,power_law_1.01,0.041868799924850465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,8,power_law_1.01,0.020531199872493744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,8,power_law_1.01,1.330361557006836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,8,power_law_1.01,0.04273920059204102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,8,power_law_1.01,0.021452799439430237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,8,power_law_1.01,2.2356672286987305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,8,power_law_1.01,0.04355199933052063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,8,power_law_1.01,0.025094398856163026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,8,power_law_1.01,0.04444159865379334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,8,power_law_1.01,0.04349440038204193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,8,power_law_1.01,4.405414581298828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,8,power_law_1.01,0.042982399463653564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,8,power_law_1.01,0.04549759924411774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,8,power_law_1.01,0.04562560021877289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,8,power_law_1.01,0.04681600034236908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,8,power_law_1.01,0.042854401469230655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,8,power_law_1.01,0.04755200147628784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,8,power_law_1.01,0.03666560053825378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,8,power_law_1.01,0.052857601642608644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,2,8,power_law_1.2,0.05736320018768311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,2,8,power_law_1.2,0.03678719997406006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,8,power_law_1.01,0.06033279895782471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,8,power_law_1.01,0.051052802801132204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,8,power_law_1.01,0.06074879765510559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,8,power_law_1.01,0.06156799793243408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,2,8,power_law_1.2,0.05941759943962097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,2,8,power_law_1.2,0.03699840009212494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,8,power_law_1.01,0.06388480067253113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,8,power_law_1.01,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,8,power_law_1.01,0.07294080257415772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,2,8,power_law_1.2,0.05575680136680603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,8,power_law_1.01,0.06758400201797485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,2,8,power_law_1.2,0.036211198568344115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,8,power_law_1.01,0.0855679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,8,power_law_1.01,0.054995197057724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,2,8,power_law_1.2,0.06002560257911682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,2,8,power_law_1.2,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.01,0.10295679569244384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,8,power_law_1.01,0.06219519972801209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,2,8,power_law_1.2,0.05995519757270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,2,8,power_law_1.2,0.039059200882911684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.01,0.12672640085220338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,2,8,power_law_1.2,0.03858560025691986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,2,8,power_law_1.2,0.06272640228271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,2,8,power_law_1.2,0.039961600303649904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.01,0.1484671950340271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,8,power_law_1.01,0.08320639729499817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,2,8,power_law_1.2,0.06985599994659424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,2,8,power_law_1.2,0.04878079891204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.01,0.191539204120636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,2,8,power_law_1.2,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.01,0.10035200119018554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,2,8,power_law_1.2,0.07486720085144043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.01,0.21895039081573486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,2,8,power_law_1.2,0.03935999870300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,2,8,power_law_1.2,0.050387197732925416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,2,8,power_law_1.2,0.2215104103088379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.01,0.32766079902648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,2,8,power_law_1.2,0.040217599272727965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.01,0.13784960508346558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,2,8,power_law_1.2,0.05181440114974976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,2,8,power_law_1.2,0.36796159744262696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.01,0.39901440143585204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,2,8,power_law_1.2,0.040300801396369934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.01,0.1710271954536438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,2,8,power_law_1.2,0.0536575973033905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,2,8,power_law_1.2,0.06765440106391907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.01,0.6641088008880616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,2,8,power_law_1.2,0.041484799981117246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,2,8,power_law_1.2,0.05669119954109192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.01,0.2396159887313843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,2,8,power_law_1.2,0.08305919766426087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,2,8,power_law_1.2,0.062412798404693604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,2,8,power_law_1.2,0.041503998637199405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.01,0.8668992042541503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.01,0.30032639503479003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,2,8,power_law_1.2,0.08105599880218506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,2,8,power_law_1.2,0.06348800063133239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,8,power_law_1.01,1.0934656143188477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.01,0.4349247932434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,2,8,power_law_1.2,0.042803201079368594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,2,8,power_law_1.2,0.08982399702072144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,2,8,power_law_1.2,0.06508799791336059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,8,power_law_1.01,1.5484543800354005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,2,8,power_law_1.2,0.04557439982891083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,2,8,power_law_1.2,0.10337280035018921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,2,8,power_law_1.2,0.0775871992111206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,2,8,power_law_1.2,0.11819519996643066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,2,8,power_law_1.2,0.08176640272140503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.01,0.5642367839813233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,8,power_law_1.01,3.071014404296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,2,8,power_law_1.2,0.10471680164337158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,2,8,power_law_1.2,0.046751999855041505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,2,8,power_law_1.2,0.09283199906349182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,2,8,power_law_1.2,0.13890559673309327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.01,0.8289600372314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,2,8,power_law_1.2,0.16647679805755616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,2,8,power_law_1.2,0.10797439813613892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,2,8,power_law_1.2,0.22802560329437255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,2,8,power_law_1.2,0.046758401393890384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,2,8,power_law_1.2,0.12896640300750734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,2,8,power_law_1.2,0.2707200050354004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.01,1.0927743911743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,2,8,power_law_1.2,0.16580480337142944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,2,8,power_law_1.2,0.04976640045642853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,2,8,power_law_1.2,0.37623679637908936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,2,8,power_law_1.2,0.48711681365966797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,2,8,power_law_1.2,0.1938367962837219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,8,power_law_1.01,1.3554431915283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,2,8,power_law_1.2,0.051801598072052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,2,8,power_law_1.2,0.6391808032989502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,2,8,power_law_1.2,0.2575743913650513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,2,8,power_law_1.2,0.05189759731292724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,8,power_law_1.01,2.1472896575927733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,2,8,power_law_1.2,0.9261695861816406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,2,8,power_law_1.2,0.337772798538208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,2,8,power_law_1.2,0.02336000055074692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,2,8,power_law_1.2,0.061862397193908694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,2,8,power_law_1.2,0.49333758354187013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,8,power_law_1.01,4.257651138305664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,2,8,power_law_1.2,0.06606720089912414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,2,8,power_law_1.2,1.307590389251709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,2,8,power_law_1.2,0.022694399952888487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,2,8,power_law_1.2,0.6051968097686767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,2,8,power_law_1.2,0.07014399766921997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,2,8,power_law_1.2,0.08911359906196595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,2,8,power_law_1.2,1.7485055923461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,2,8,power_law_1.2,0.9051456451416016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,2,8,power_law_1.2,0.02192640006542206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.2,0.10517760515213012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,2,8,power_law_1.2,2.1540607452392577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,2,8,power_law_1.2,0.021087999641895293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,2,8,power_law_1.2,1.1570367813110352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.2,0.11663999557495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,2,8,power_law_1.2,3.157708740234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,2,8,power_law_1.2,0.022086399793624877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.2,0.15003520250320435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,2,8,power_law_1.2,0.02484479993581772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,2,8,power_law_1.2,1.4297472000122071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,2,8,power_law_1.2,5.8572032928466795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.2,0.17895679473876952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,2,8,power_law_1.2,0.0438400000333786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,2,8,power_law_1.2,2.125926399230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.2,0.24033279418945314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,2,8,power_law_1.2,0.04353919923305512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.2,0.327513599395752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,2,8,power_law_1.2,4.404697418212891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,2,8,power_law_1.2,0.045798400044441225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.2,0.40428800582885743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,2,8,power_law_1.2,0.043296000361442565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.2,0.6685887813568115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,2,8,power_law_1.2,0.037222400307655334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.2,0.8839167594909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,2,8,power_law_1.2,0.060070401430130003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,2,8,power_law_1.2,1.0553024291992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,2,8,power_law_1.2,0.061414402723312375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,2,8,power_law_1.2,0.04652799963951111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,2,8,power_law_1.2,1.7915712356567384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,2,8,power_law_1.2,0.06622080206871032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,2,8,power_law_1.2,3.350592041015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,2,8,power_law_1.2,0.055193597078323366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,2,8,power_law_1.2,0.06234880089759827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,2,8,power_law_1.2,0.0837823987007141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.2,0.0990015983581543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.2,0.13854080438613892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.2,0.171014404296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.2,0.24004480838775635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.2,0.3004224061965942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.2,0.4356224060058594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.2,0.564467191696167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.2,0.8289983749389649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.2,1.0961600303649903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,2,8,power_law_1.2,1.3579456329345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,2,8,power_law_1.2,2.149420738220215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,4,balanced,0.05091199775536855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,2,8,power_law_1.2,4.259731292724609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,4,balanced,0.037818667789300285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,4,balanced,0.056426664193471275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,4,balanced,0.05965333183606466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,4,balanced,0.06136000156402588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,4,balanced,0.060496002435684204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,4,balanced,0.06177600224812826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,4,balanced,0.060645331939061485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,4,balanced,0.06035733222961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,4,balanced,0.05979733169078827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,4,balanced,0.06205333272616068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,4,balanced,0.061434666315714516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,4,balanced,0.0617386649052302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,4,balanced,0.06553600231806438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,4,balanced,0.033759998778502144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,4,balanced,0.06814933319886525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,4,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,4,balanced,0.0711413323879242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,4,balanced,0.037647999823093414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,4,balanced,0.07801066835721333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,4,balanced,0.03734933336575826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,4,balanced,0.03722666700681051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,4,balanced,0.08492799599965413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,4,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,4,balanced,0.08870933453241985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,4,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,4,balanced,0.038975998759269714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,4,balanced,0.1076746682325999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,4,balanced,0.03835200021664301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,4,balanced,0.11531200011571248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,4,balanced,0.039664000272750854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,4,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,4,balanced,0.14775466918945312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,4,balanced,0.042394667863845825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,4,balanced,0.0417546679576238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,4,balanced,0.17599467436472574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,4,balanced,0.047498668233553566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,4,balanced,0.05601066847642263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,4,balanced,0.0543146679798762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,4,balanced,0.23054399092992148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,4,balanced,0.06196799874305725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,4,balanced,0.07210666437943776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,4,balanced,0.28916800022125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,4,balanced,0.08201066652933757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,4,balanced,0.09648533662160237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,4,balanced,0.1029866635799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,4,balanced,0.4005013306935628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,4,balanced,0.1317813297112783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,4,balanced,0.16260266304016113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,4,balanced,0.49566932519276935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,4,balanced,0.048351998130480446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,4,balanced,0.21498133738835654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,4,balanced,0.04804266492525736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,4,balanced,0.2712159951527913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,4,balanced,0.05633600056171417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,4,balanced,0.05805333455403646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,4,balanced,0.604416012763977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,4,balanced,0.05669866502285004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,4,balanced,0.05737066765626272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,4,balanced,0.3745439847310384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,4,balanced,0.060362666845321655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,4,balanced,0.05729599793752035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,4,balanced,0.05825600028038025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,4,balanced,0.02015999952952067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,4,balanced,0.06041066845258077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,4,balanced,0.021157334248224895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,4,balanced,0.9126666386922201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,4,balanced,0.47009599208831787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,4,balanced,0.052799999713897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,4,balanced,0.019343999524911244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,4,balanced,0.0561653325955073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,4,balanced,0.020981334149837494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,4,balanced,0.056128000219662987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,4,balanced,0.021168000996112823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,4,balanced,0.05286933481693268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,4,balanced,0.021045332153638203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,4,balanced,0.5740053256352743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,4,balanced,0.060271998246510826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,4,balanced,0.027263998985290527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,4,balanced,0.05569600065549215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,4,balanced,0.024698667228221893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,4,balanced,0.02942399928967158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,4,balanced,0.054117331902186074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,4,balanced,0.0276053324341774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,4,balanced,0.07295466462771098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,4,balanced,0.029130667448043823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,4,balanced,0.07331733405590057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,4,balanced,0.031210665901501972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,4,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,4,balanced,0.08469333251317342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,4,balanced,0.8701653480529785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,4,balanced,0.03540800015131632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,4,balanced,0.10116266210873921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,4,balanced,0.04565866788228353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,4,balanced,1.755066712697347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,4,balanced,0.1265600025653839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,4,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,4,balanced,0.0458186666170756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,4,balanced,0.15215466419855753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,4,balanced,0.05858133236567179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,4,balanced,0.06648533542950948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,4,power_law_1.01,0.06343680024147033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,4,power_law_1.01,0.03700479865074158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,4,balanced,0.20513067642847696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,4,balanced,0.08656533559163411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,4,power_law_1.01,0.05118079781532288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,4,power_law_1.01,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,4,balanced,0.1072213351726532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,4,balanced,0.2558559974034627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,4,power_law_1.01,0.03418880105018616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,4,power_law_1.01,0.051283198595047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,4,balanced,0.14661866426467896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,4,power_law_1.01,0.036313599348068236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,4,power_law_1.01,0.05750399827957153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,4,power_law_1.01,0.03669120073318481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,4,power_law_1.01,0.058380800485610965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,4,balanced,0.35740800698598224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,4,balanced,0.1811359922091166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,4,balanced,1.6651466687520344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,4,power_law_1.01,0.03937920033931732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,4,power_law_1.01,0.06319360136985779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,4,power_law_1.01,0.05245440006256104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,4,power_law_1.01,0.044249600172042845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,4,power_law_1.01,0.06718720197677612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,4,power_law_1.01,0.04415999948978424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,4,balanced,0.2561759948730469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,4,power_law_1.01,0.06385279893875122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,4,balanced,0.4623626470565796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,4,power_law_1.01,0.06606079936027527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,4,power_law_1.01,0.044607999920845035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,4,power_law_1.01,0.04775680005550385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,4,balanced,0.3283519943555196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,4,power_law_1.01,0.04885759949684143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,4,power_law_1.01,0.0661952018737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,4,power_law_1.01,0.04979200065135956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,4,power_law_1.01,0.06901119947433472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,4,power_law_1.01,0.05000960230827332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,4,power_law_1.01,0.07065600156784058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,4,power_law_1.01,0.05175039768218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,4,power_law_1.01,0.056883198022842404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,4,power_law_1.01,0.05509120225906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,4,power_law_1.01,0.06800640225410462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,4,balanced,0.5650240182876587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,4,balanced,0.4759253263473511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,4,power_law_1.01,0.05494400262832642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,4,power_law_1.01,0.05920640230178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,4,power_law_1.01,0.07759360074996949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,4,power_law_1.01,0.05626239776611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,4,power_law_1.01,0.06659839749336242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,4,power_law_1.01,0.05667200088500977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,4,power_law_1.01,0.08156800270080566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,4,power_law_1.01,0.07305600047111512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,4,power_law_1.01,0.05496960282325745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,4,power_law_1.01,0.08655359745025634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,4,balanced,0.6236159801483154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,4,power_law_1.01,0.055103999376296994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,4,power_law_1.01,0.09923840165138245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,4,power_law_1.01,0.08133119940757752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,4,balanced,0.8710453510284424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,4,power_law_1.01,0.11475199460983276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,4,power_law_1.01,0.05482239723205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,4,power_law_1.01,0.092467200756073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,4,power_law_1.01,0.1378368020057678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,4,power_law_1.01,0.05541120171546936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,4,power_law_1.01,0.1653439998626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,4,power_law_1.01,0.10551040172576905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,4,power_law_1.01,0.05766400098800659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,4,balanced,0.7697439988454183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,4,power_law_1.01,0.18620799779891967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,4,power_law_1.01,0.059411197900772095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,4,power_law_1.01,0.1312831997871399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,4,power_law_1.01,0.13628159761428832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,4,power_law_1.01,0.06217600107192993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,4,power_law_1.01,0.27352960109710694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,4,power_law_1.01,0.07117440104484558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,4,power_law_1.01,0.19143680334091187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,4,power_law_1.01,0.31496319770812986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,4,power_law_1.01,0.07308800220489502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,4,power_law_1.01,0.25615360736846926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,4,power_law_1.01,0.465715217590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,4,power_law_1.01,0.07997440099716187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,4,power_law_1.01,0.38815999031066895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,4,power_law_1.01,0.02001280039548874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,4,power_law_1.01,0.646073579788208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,4,power_law_1.01,0.09440640211105347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,4,power_law_1.01,0.5079936027526856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.01,0.11084159612655639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,4,power_law_1.01,0.8404800415039062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,4,balanced,1.2152746518452961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,4,power_law_1.01,0.01968639940023422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,4,power_law_1.01,0.7616511821746826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.01,0.132096004486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,4,power_law_1.01,0.019443200528621675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,4,power_law_1.01,1.1849791526794433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,4,power_law_1.01,0.01930239945650101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.01,0.13725440502166747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,4,power_law_1.01,0.909721565246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,4,power_law_1.01,0.019750399887561797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,4,power_law_1.01,1.4698687553405763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,4,balanced,1.693621317545573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.01,0.22211840152740478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,4,power_law_1.01,0.020230400562286376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,4,power_law_1.01,1.088102436065674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,4,power_law_1.01,2.6033023834228515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.01,0.26750080585479735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,4,power_law_1.01,0.025286400318145753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,4,power_law_1.01,0.02316800057888031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.01,0.36389760971069335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,4,power_law_1.01,1.5570176124572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,4,power_law_1.01,0.02837119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,4,power_law_1.01,4.797299194335937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.01,0.5587776184082032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,4,power_law_1.01,0.026982399821281432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,4,power_law_1.01,0.027923199534416198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.01,0.8539199829101562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,4,power_law_1.01,3.14718074798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,4,power_law_1.01,0.030067199468612672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,4,power_law_1.01,0.033395200967788696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.01,0.833407974243164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,4,power_law_1.01,0.03475840091705322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,4,power_law_1.01,0.04506239891052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,4,power_law_1.01,0.04336639940738678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,4,balanced,2.3957759539286294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,4,power_law_1.01,0.044512000679969785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,4,power_law_1.01,1.3028223991394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,4,power_law_1.01,0.05737599730491638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.01,0.0658623993396759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.01,0.084934401512146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,4,power_law_1.01,1.932147216796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.01,0.10771199464797973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.01,0.14623359441757203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.01,0.17894400358200074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,4,power_law_1.01,4.052889633178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.01,0.25396480560302737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.01,0.32654719352722167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.01,0.47272319793701173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.01,0.6217408180236816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,4,power_law_1.01,0.7690815925598145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,4,power_law_1.01,1.2126848220825195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,4,power_law_1.01,2.394918441772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,4,power_law_1.2,0.06159999966621399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,4,power_law_1.2,0.047654399275779726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,4,power_law_1.2,0.05449600219726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,4,power_law_1.2,0.05738880038261414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,4,power_law_1.2,0.058956801891326904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,4,power_law_1.2,0.06110720038414001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,4,power_law_1.2,0.07058560252189636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,4,power_law_1.2,0.06320639848709106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,4,power_law_1.2,0.06488320231437683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,4,power_law_1.2,0.06595199704170226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,4,power_law_1.2,0.0701312005519867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,4,power_law_1.2,0.06996480226516724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,4,power_law_1.2,0.06998400092124939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,4,power_law_1.2,0.07836800217628478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,4,power_law_1.2,0.08254079818725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,4,power_law_1.2,0.08876799941062927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,4,power_law_1.2,0.10338560342788697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,4,power_law_1.2,0.035148799419403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,4,power_law_1.2,0.12616959810256959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,4,power_law_1.2,0.13893760442733766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,4,power_law_1.2,0.034860798716545106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,4,power_law_1.2,0.17834880352020263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,4,power_law_1.2,0.03547520041465759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,4,power_law_1.2,0.19347840547561646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,4,power_law_1.2,0.035743999481201175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,4,power_law_1.2,0.03632639944553375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,4,power_law_1.2,0.2922688007354736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,4,power_law_1.2,0.03834879994392395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,4,power_law_1.2,0.3522559881210327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,4,power_law_1.2,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,4,power_law_1.2,0.04666880071163178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,4,power_law_1.2,0.5465216159820556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,4,power_law_1.2,0.04742400050163269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,4,power_law_1.2,0.7548416137695313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,4,power_law_1.2,0.05045120120048523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,4,power_law_1.2,1.0291711807250976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,4,power_law_1.2,0.049721598625183105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,4,power_law_1.2,0.05610880255699158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,4,power_law_1.2,1.0963264465332032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,4,power_law_1.2,0.05549439787864685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,4,power_law_1.2,1.5046655654907226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,4,power_law_1.2,0.05923200249671936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,4,power_law_1.2,2.6090688705444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,4,power_law_1.2,0.06982399821281433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,4,power_law_1.2,0.07433599829673768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,4,power_law_1.2,4.757036972045898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,4,power_law_1.2,0.08284159898757934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,4,power_law_1.2,0.09512959718704224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,4,power_law_1.2,0.12086399793624877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,4,power_law_1.2,0.1372159957885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,4,power_law_1.2,0.1497663974761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,4,power_law_1.2,0.2224128007888794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,4,power_law_1.2,0.3133631944656372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,4,power_law_1.2,0.44576640129089357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,4,power_law_1.2,0.05139200091361999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,4,power_law_1.2,0.5255807876586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,4,power_law_1.2,0.018943999707698823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,4,power_law_1.2,0.7594175815582276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,4,power_law_1.2,0.05713919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,4,power_law_1.2,0.8172287940979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,4,power_law_1.2,0.019046400487422944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,4,power_law_1.2,0.05299839973449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,4,power_law_1.2,1.1729791641235352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,4,power_law_1.2,0.01881600022315979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,4,power_law_1.2,0.0584384024143219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,4,power_law_1.2,1.819321632385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,4,power_law_1.2,0.01953279972076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,4,power_law_1.2,0.05685120224952698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,4,power_law_1.2,3.4049022674560545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,4,power_law_1.2,0.019923199713230134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,4,power_law_1.2,0.05678719878196716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,4,power_law_1.2,0.020339199900627138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,4,power_law_1.2,0.024979199469089507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,4,power_law_1.2,0.05836799740791321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,4,power_law_1.2,0.023238399624824525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,4,power_law_1.2,0.05541759729385376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,4,power_law_1.2,0.028307199478149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,4,power_law_1.2,0.05598719716072083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,2,balanced,0.031898667414983116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,4,power_law_1.2,0.02656640112400055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,2,balanced,0.030432000756263733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,4,power_law_1.2,0.027481600642204285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,2,balanced,0.04172799984614054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,2,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,4,power_law_1.2,0.05548800230026245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,2,balanced,0.03358400116364161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,4,power_law_1.2,0.03000960052013397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,4,power_law_1.2,0.032307198643684386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,2,balanced,0.019173332800467808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,2,balanced,0.05388266841570536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,2,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,2,balanced,0.044293334086736046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,4,power_law_1.2,0.03375999927520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,4,power_law_1.2,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,2,balanced,0.019109333554903667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,2,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,2,balanced,0.058005332946777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,4,power_law_1.2,0.04439040124416351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,2,balanced,0.05392533540725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,2,balanced,0.01903466631968816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,4,power_law_1.2,0.04264959990978241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,2,balanced,0.05816000203291575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,2,balanced,0.05422399938106537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,2,balanced,0.035599999129772186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,2,balanced,0.019727999965349834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,4,power_law_1.2,0.044736000895500186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,4,power_law_1.2,0.061318397521972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,2,balanced,0.021040000021457672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,4,power_law_1.2,0.05730559825897217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,2,balanced,0.05421866476535797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,2,balanced,0.021157334248224895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,2,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,2,balanced,0.05764799813429514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,4,power_law_1.2,0.06221439838409424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.2,0.06525440216064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,2,balanced,0.021525333325068157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,2,balanced,0.05602133274078369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,2,balanced,0.03800000001986822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,2,balanced,0.021141332884629566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,2,balanced,0.0599839985370636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,2,balanced,0.025333332518736523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,4,power_law_1.2,0.05985919833183288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,2,balanced,0.058890665570894875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,2,balanced,0.03908800085385641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.2,0.08558080196380616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,2,balanced,0.029616000751654308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,2,balanced,0.060346667965253196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,2,balanced,0.039647998909155525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,2,balanced,0.05392533540725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.2,0.10698879957199096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,2,balanced,0.02532800038655599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,4,power_law_1.2,0.07109760046005249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,2,balanced,0.03985599925120672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,2,balanced,0.05924266576766968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,2,balanced,0.02718399961789449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,2,balanced,0.05407466491063436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,2,balanced,0.029322666426499683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,2,balanced,0.04178133110205332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,4,power_law_1.2,0.07556480169296265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.2,0.1449280023574829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,2,balanced,0.06001066664854685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,2,balanced,0.029648000995318096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,2,balanced,0.05482133229573568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,2,balanced,0.043893332282702126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,2,balanced,0.03166399896144867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,2,balanced,0.06150400141874949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,4,power_law_1.2,0.07953919768333435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,2,balanced,0.05406933526198069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.2,0.17823359966278077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,2,balanced,0.04358933369318644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,2,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,2,balanced,0.06189866860707601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,2,balanced,0.0469813346862793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,2,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,2,balanced,0.059989333152770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,4,power_law_1.2,0.09270399808883667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.2,0.2543103933334351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,2,balanced,0.04782933493455251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,2,balanced,0.05898133416970571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,2,balanced,0.062122667829195656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,2,balanced,0.05216533442338308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,2,balanced,0.049327999353408813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,2,balanced,0.060133333007494606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.2,0.32588160037994385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.2,0.10956799983978271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,2,balanced,0.06414400041103363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,2,balanced,0.06434666613737743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,2,balanced,0.05422399938106537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,2,balanced,0.062352001667022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,2,balanced,0.07665066421031952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.2,0.47367038726806643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.2,0.14967039823532105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,2,balanced,0.07381866872310638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,2,balanced,0.07867200175921123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,2,balanced,0.0580213318268458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,2,balanced,0.10021866361300151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,2,balanced,0.08169066905975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,2,balanced,0.07667199770609538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.2,0.6207808017730713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.2,0.16935039758682252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,2,balanced,0.05932799975077311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,2,balanced,0.12296000123023987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,2,balanced,0.09853333234786987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,2,balanced,0.07467199862003326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.2,0.2382335901260376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,2,balanced,0.05796800057093302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,2,balanced,0.11428800225257874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,4,power_law_1.2,0.7705471992492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,2,balanced,0.08713600039482117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,2,balanced,0.16725333531697592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.2,0.30779519081115725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,2,balanced,0.0761706680059433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,4,power_law_1.2,1.2121343612670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,2,balanced,0.14894400040308634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,2,balanced,0.08753599723180135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.2,0.4545279979705811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,2,balanced,0.2105813423792521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,2,balanced,0.07842666904131572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.2,0.6562880039215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,2,balanced,0.1069653332233429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,4,power_law_1.2,2.3948032379150392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,2,balanced,0.09201600154240926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,2,balanced,0.1851146618525187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.2,0.7664447784423828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,2,balanced,0.11823466420173645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,2,balanced,0.2969706654548645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.2,0.9565119743347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,2,balanced,0.10957333445549011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,2,balanced,0.24708799521128336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,2,balanced,0.15187733372052512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,4,power_law_1.2,1.332313632965088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,2,balanced,0.140474667151769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,4,power_law_1.2,2.228160095214844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,2,balanced,0.19142399231592813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,2,balanced,0.3847999970118205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,2,balanced,0.3128906687100728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,4,power_law_1.2,4.106425476074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,2,balanced,0.1731520096460978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,2,balanced,0.2493600050608317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,2,balanced,0.44276265303293866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,2,balanced,0.23245867093404135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,2,balanced,0.473194678624471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,2,balanced,0.3006613254547119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,2,balanced,0.29416000843048096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,2,balanced,0.5674826701482137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,2,balanced,0.41285868485768634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,2,balanced,0.7375946839650472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,2,balanced,0.4166613419850667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,2,balanced,0.5234666665395101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,2,balanced,0.6897066434224447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,2,balanced,0.5405013163884481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,2,balanced,0.6366986831029257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,2,balanced,0.6613866488138834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,2,balanced,1.0606506665547688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,2,balanced,1.4453013737996419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,2,balanced,0.96833602587382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,2,balanced,1.028106689453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,2,balanced,2.045584042867025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,2,balanced,1.854042689005534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,2,balanced,2.019498666127523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,2,power_law_1.01,0.037811198830604555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,2,power_law_1.01,0.037510401010513304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,2,power_law_1.01,0.04538240134716034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,2,power_law_1.01,0.056383997201919556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,2,power_law_1.01,0.05619840025901794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,2,power_law_1.01,0.05699840188026428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,2,power_law_1.01,0.058284801244735715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,2,power_law_1.01,0.05927680134773254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,2,power_law_1.01,0.060819202661514284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,2,power_law_1.01,0.06272000074386597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,2,power_law_1.01,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,2,power_law_1.01,0.0664192020893097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,2,power_law_1.01,0.06691200137138367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,2,power_law_1.01,0.07063040137290955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,2,power_law_1.01,0.07457919716835022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,2,power_law_1.01,0.08276479840278625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,2,power_law_1.01,0.08373759984970093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,2,power_law_1.01,0.031148800253868104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,2,power_law_1.01,0.09848960041999817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,2,power_law_1.01,0.11435519456863404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,2,power_law_1.01,0.03157120048999786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,2,power_law_1.01,0.12371200323104858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,2,power_law_1.01,0.03172479867935181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,2,power_law_1.01,0.1622655987739563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,2,power_law_1.01,0.03497599959373474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,2,power_law_1.01,0.20147199630737306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,2,power_law_1.01,0.03573119938373566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,2,power_law_1.01,0.23716480731964112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,2,power_law_1.01,0.040320000052452086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,2,power_law_1.01,0.3323647975921631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,2,power_law_1.01,0.04276480078697205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,2,power_law_1.01,0.04226559996604919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,2,power_law_1.01,0.4075647830963135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,2,power_law_1.01,0.0425024002790451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,2,power_law_1.01,0.6378367900848388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,2,power_law_1.01,0.0468095988035202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,2,power_law_1.01,0.6815999984741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,2,power_law_1.01,0.04418559968471527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,2,power_law_1.01,0.9802047729492187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,2,power_law_1.01,0.04935680031776428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,2,power_law_1.01,0.05251839756965637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,2,power_law_1.01,1.4860671997070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,2,power_law_1.01,0.05261440277099609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,2,power_law_1.01,3.197324752807617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,2,power_law_1.01,0.06471679806709289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,2,power_law_1.01,0.06856319904327393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,2,power_law_1.01,0.07151359915733338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,2,power_law_1.01,0.08981760144233704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,2,power_law_1.01,0.09358720183372497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,2,power_law_1.01,0.11768959760665894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,2,power_law_1.01,0.13737599849700927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,2,power_law_1.01,0.17745920419692993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,2,power_law_1.01,0.20276479721069335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,2,power_law_1.01,0.31332480907440186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,2,power_law_1.01,0.3965951919555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,2,power_law_1.01,0.6323584079742431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,2,power_law_1.01,0.7766784191131592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,2,power_law_1.01,0.8714879989624024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,2,power_law_1.01,1.6327936172485351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,2,power_law_1.01,2.854924774169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,2,power_law_1.01,0.043161600828170776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,2,power_law_1.01,0.01886720061302185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,2,power_law_1.01,0.04436480104923248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,2,power_law_1.01,0.0491456001996994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,2,power_law_1.01,0.01842560023069382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,2,power_law_1.01,0.053376001119613645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,2,power_law_1.01,0.053887999057769774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,2,power_law_1.01,0.018406400084495546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,2,power_law_1.01,0.05382400155067444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,2,power_law_1.01,0.018688000738620758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,2,power_law_1.01,0.0541375994682312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,2,power_law_1.01,0.01918720006942749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,2,power_law_1.01,0.05594879984855652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,2,power_law_1.01,0.01910399943590164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,2,power_law_1.01,0.054425597190856934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,2,power_law_1.01,0.020339199900627138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,2,power_law_1.01,0.05454080104827881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,2,power_law_1.01,0.020716799795627593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,2,power_law_1.01,0.05416960120201111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,2,power_law_1.01,0.024307200312614442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,2,power_law_1.01,0.05498239994049072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,2,power_law_1.01,0.028838399052619933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,2,power_law_1.01,0.05614079833030701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,2,power_law_1.01,0.024243199825286867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,2,power_law_1.01,0.06103039979934692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,2,power_law_1.01,0.06828160285949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,2,power_law_1.01,0.025811201333999632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,2,power_law_1.01,0.06794880032539367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,2,power_law_1.01,0.027801600098609925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,2,power_law_1.01,0.02871679961681366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,2,power_law_1.01,0.07525119781494141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,2,power_law_1.2,0.03457919955253601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,2,power_law_1.01,0.031795200705528257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,2,power_law_1.01,0.08478080034255982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,2,power_law_1.2,0.039468801021575926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.01,0.09489920139312744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,2,power_law_1.01,0.034380799531936644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.01,0.11104639768600463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,2,power_law_1.2,0.03865599930286408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,2,power_law_1.01,0.0346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.01,0.13863680362701417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,2,power_law_1.2,0.055929601192474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,2,power_law_1.01,0.04622080028057098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.01,0.20015358924865723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,2,power_law_1.2,0.05631359815597534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.01,0.04778240025043488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.01,0.23582720756530762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,2,power_law_1.2,0.057625597715377806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.01,0.33409919738769533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.01,0.06213759779930115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,2,power_law_1.2,0.05902720093727112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.01,0.3722879886627197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.01,0.07578240036964416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,2,power_law_1.2,0.06261119842529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.01,0.6249472141265869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.01,0.09700480103492737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,2,power_law_1.2,0.062080001831054686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.01,0.8617919921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,2,power_law_1.2,0.0649407982826233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.01,0.12079999446868897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,2,power_law_1.01,1.0111680030822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,2,power_law_1.2,0.06505600214004517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.01,0.16488959789276122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,2,power_law_1.2,0.06676480174064636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,2,power_law_1.01,1.5793984413146973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.01,0.20711040496826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,2,power_law_1.2,0.06803200244903565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,2,power_law_1.01,2.487513542175293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.01,0.2954751968383789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,2,power_law_1.2,0.06896640062332153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,2,power_law_1.2,0.031327998638153075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,2,power_law_1.2,0.07663360238075256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.01,0.3839040040969849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,2,power_law_1.2,0.04314880073070526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,2,power_law_1.2,0.08253440260887146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,2,power_law_1.2,0.03218559920787811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,2,power_law_1.01,0.4725056171417236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,2,power_law_1.2,0.08670079708099365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,2,power_law_1.2,0.04289279878139496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,2,power_law_1.2,0.031692799925804135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,2,power_law_1.01,0.7339903831481933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,2,power_law_1.2,0.09254400134086609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,2,power_law_1.2,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,2,power_law_1.2,0.04819200038909912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,2,power_law_1.01,1.4404159545898438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,2,power_law_1.2,0.10728319883346557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,2,power_law_1.2,0.036262398958206175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,2,power_law_1.2,0.019116799533367156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,2,power_law_1.2,0.05532159805297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,2,power_law_1.2,0.13050240278244019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,2,power_law_1.2,0.05379840135574341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,2,power_law_1.2,0.039724799990653994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,2,power_law_1.2,0.15838079452514647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,2,power_law_1.2,0.018105599284172057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,2,power_law_1.2,0.05419520139694214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,2,power_law_1.2,0.04280959963798523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,2,power_law_1.2,0.19982080459594725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,2,power_law_1.2,0.054585599899291994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,2,power_law_1.2,0.018086400628089905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,2,power_law_1.2,0.041119998693466185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,2,power_law_1.2,0.24978559017181395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,2,power_law_1.2,0.05509120225906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,2,power_law_1.2,0.01802240014076233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,2,power_law_1.2,0.04350079894065857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,2,power_law_1.2,0.3739711999893188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,2,power_law_1.2,0.05503360033035278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,2,power_law_1.2,0.046079999208450316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,2,power_law_1.2,0.018016000092029572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,2,power_law_1.2,0.05469440221786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,2,power_law_1.2,0.4322239875793457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,2,power_law_1.2,0.0465472012758255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,2,power_law_1.2,0.01850239932537079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,2,power_law_1.2,0.05046399831771851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,2,power_law_1.2,0.05345919728279114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,2,power_law_1.2,0.6016704082489014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,2,power_law_1.2,0.019251200556755065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,2,power_law_1.2,0.055212801694869994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,2,power_law_1.2,0.020319999754428865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,2,power_law_1.2,0.0574783980846405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,2,power_law_1.2,0.7148799896240234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,2,power_law_1.2,0.05509759783744812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,2,power_law_1.2,0.024160000681877136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,2,power_law_1.2,0.05824000239372253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,2,power_law_1.2,0.9706624031066895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,2,power_law_1.2,0.028191998600959778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,2,power_law_1.2,0.06462079882621766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,2,power_law_1.2,0.057196801900863646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,2,power_law_1.2,0.06988800168037415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,2,power_law_1.2,0.024486400187015533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,2,power_law_1.2,0.0649407982826233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,2,power_law_1.2,1.5255807876586913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,2,power_law_1.2,0.07524480223655701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,2,power_law_1.2,0.06767359972000123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,2,power_law_1.2,0.02481279969215393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,2,power_law_1.2,0.08814079761505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,2,power_law_1.2,0.07429119944572449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,2,power_law_1.2,0.027513599395751952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,2,power_law_1.2,3.1390144348144533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,2,power_law_1.2,0.08544639945030212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,2,power_law_1.2,0.10268800258636475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,2,power_law_1.2,0.02903040051460266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,1,balanced,0.031770666440327965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.2,0.10032639503479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,2,power_law_1.2,0.12629760503768922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,2,power_law_1.2,0.03054719865322113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.2,0.12517759799957276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,2,power_law_1.2,0.1351680040359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,1,balanced,0.03605333218971888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,2,power_law_1.2,0.19608319997787477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.2,0.14273279905319214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,2,power_law_1.2,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,1,balanced,0.055386667450269066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,2,power_law_1.2,0.24511361122131348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.2,0.19601279497146606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,1,balanced,0.058042665322621666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,2,power_law_1.2,0.0347135990858078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,1,balanced,0.058186665177345276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,2,power_law_1.2,0.3433664083480835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.2,0.24681599140167237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,2,power_law_1.2,0.04562560021877289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,1,balanced,0.05831466615200043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,2,power_law_1.2,0.4440959930419922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,1,balanced,0.060319999853769936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.2,0.04750719964504242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.2,0.34946560859680176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,1,balanced,0.060533334811528526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.2,0.062489598989486694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,2,power_law_1.2,0.5652800083160401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.2,0.3865407943725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,1,balanced,0.060005332032839455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,1,balanced,0.06002666552861532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.2,0.07530879974365234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,2,power_law_1.2,0.8205183982849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.2,0.5803840160369873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,1,balanced,0.06193066636721293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,1,balanced,0.061936000982920326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.2,0.09765120148658753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.2,0.7817215919494629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,2,power_law_1.2,0.9511103630065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,1,balanced,0.06326933205127716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,1,balanced,0.06817066669464111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.2,0.12007679939270019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,2,power_law_1.2,1.0216959953308105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,2,power_law_1.2,1.3631360054016113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,1,balanced,0.07431999842325847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,1,balanced,0.07691733539104462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,2,power_law_1.2,1.3977791786193847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.2,0.1649216055870056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,1,balanced,0.08046400050322215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,2,power_law_1.2,2.991916847229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,1,balanced,0.09618666768074036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.2,0.2066431999206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,2,power_law_1.2,3.3170814514160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,1,balanced,0.09744532903035481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.2,0.2939840078353882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,1,balanced,0.12143466869990031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.2,0.3815295934677124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,1,balanced,0.14728533228238425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,2,power_law_1.2,0.47034239768981934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,1,balanced,0.1886133352915446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,2,power_law_1.2,0.7357247829437256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,1,balanced,0.22658665974934897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,2,power_law_1.2,1.442240047454834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,1,balanced,0.2944906751314799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,1,balanced,0.03139200061559677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,1,balanced,0.03331733246644338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,1,balanced,0.38174935181935626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,1,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,1,balanced,0.03842133283615112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,1,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,1,balanced,0.5330346822738647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,1,balanced,0.03782399992148081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,1,balanced,0.04014399896065394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,1,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,1,balanced,0.039520000418027244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,1,balanced,0.6832533677419027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,1,balanced,0.04142933338880539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,1,balanced,0.0421066681543986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,1,balanced,0.04402133325735728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,1,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,1,balanced,0.05175999800364176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,1,balanced,0.8335946400960287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,1,balanced,0.061573331554730736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,1,balanced,0.06398933132489522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,1,balanced,0.07310933371384938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,1,balanced,0.09090133508046468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,1,balanced,0.09460266431172688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,1,balanced,0.12141866485277812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,1,balanced,0.13612266381581625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,1,balanced,1.2821813424428303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,1,balanced,0.18140800793965658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,1,balanced,0.22220800320307413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,1,balanced,0.31215999523798627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,1,balanced,0.40296534697214764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,1,balanced,0.58078400293986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,1,balanced,2.4976906776428223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,1,balanced,0.7526453336079916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,1,balanced,0.9257813294728597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,1,balanced,0.04182399809360504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,1,balanced,0.04789333542188009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,1,balanced,0.05575466652711233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,1,balanced,0.057855998476346336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,1,balanced,0.05692266424496969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,1,balanced,1.4454986254374187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,1,balanced,0.05428266525268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,1,balanced,0.060133333007494606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,1,balanced,0.05387733379999796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,1,balanced,0.05629866818586985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,1,balanced,0.05658133327960968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,1,balanced,0.06038933495680491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,1,balanced,0.06036800146102905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,1,balanced,0.056645333766937256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,1,balanced,0.057946667075157166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,1,balanced,0.01933866615096728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,1,balanced,0.06425066788991292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,1,balanced,0.06203199923038483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,1,balanced,0.01907733331123988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,1,balanced,0.06614933411280315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,1,balanced,0.01930133377512296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,1,balanced,0.08830400307973225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,1,balanced,0.019354666272799175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,1,balanced,2.823493321736654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,1,balanced,0.09481066465377808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,1,balanced,0.01934933289885521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,1,balanced,0.11361599961916606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,1,balanced,0.01929066702723503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,1,power_law_1.01,0.03312000036239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,1,balanced,0.021295999487241108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,1,balanced,0.13756266236305237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,1,balanced,0.021322667598724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,1,power_law_1.01,0.033036801218986514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,1,balanced,0.18150933583577475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,1,power_law_1.01,0.03982079923152924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,1,balanced,0.021520001192887623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,1,power_law_1.01,0.05566080212593079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,1,balanced,0.021386665602525074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,1,balanced,0.22902933756510416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,1,power_law_1.01,0.05663359761238098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,1,balanced,0.023200000325838726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,1,balanced,0.02550933261712392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,1,power_law_1.01,0.05697280168533325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,1,power_law_1.01,0.05746560096740723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,1,balanced,0.02762666592995326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,1,balanced,0.31908265749613446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,1,power_law_1.01,0.057868802547454835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,1,balanced,0.025770666698614757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,1,power_law_1.01,0.05777919888496399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,1,balanced,0.02770666778087616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,1,balanced,0.4113119840621948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,1,power_law_1.01,0.059462398290634155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,1,balanced,0.029824001093705494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,1,power_law_1.01,0.060915201902389526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,1,balanced,0.031658666829268135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,1,power_law_1.01,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,1,balanced,0.03788266579310099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,1,balanced,0.586085319519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,1,power_law_1.01,0.06663680076599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,1,balanced,0.041477332512537636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,1,balanced,0.056287998954455055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,1,power_law_1.01,0.06725119948387145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,1,power_law_1.01,0.07390080094337463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,1,balanced,0.769536018371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,1,balanced,0.06041066845258077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,1,power_law_1.01,0.08128640055656433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,1,balanced,0.0817386656999588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,1,power_law_1.01,0.08115199804306031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,1,power_law_1.01,0.09599999785423279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,1,balanced,0.09743466973304749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,1,power_law_1.01,0.10073599815368653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,1,power_law_1.01,0.12261760234832764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,1,balanced,0.9502560297648112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,1,balanced,0.127893328666687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,1,power_law_1.01,0.14434560537338256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,1,power_law_1.01,0.18613120317459106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,1,power_law_1.01,0.22654080390930176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,1,balanced,0.16119999686876932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,1,power_law_1.01,0.3041215896606445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,1,power_law_1.01,0.37770240306854247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,1,power_law_1.01,0.5313727855682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,1,balanced,0.22522133588790894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,1,power_law_1.01,0.6807295799255371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,1,power_law_1.01,0.8334464073181153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,1,balanced,1.499008019765218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,1,power_law_1.01,1.2824511528015137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,1,balanced,0.28813334306081134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,1,power_law_1.01,2.4999488830566405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,1,power_law_1.01,0.017395199835300447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,1,power_law_1.01,0.029151999950408937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,1,power_law_1.01,0.040601599216461184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,1,power_law_1.01,0.01764480024576187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,1,power_law_1.01,0.029126399755477907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,1,power_law_1.01,0.0440447986125946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,1,balanced,0.35309867064158124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,1,power_law_1.01,0.017529599368572235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,1,power_law_1.01,0.03128319978713989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,1,power_law_1.01,0.045484799146652224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,1,power_law_1.01,0.036671999096870425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,1,power_law_1.01,0.017798399925231932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,1,power_law_1.01,0.036160001158714296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,1,power_law_1.01,0.018092800676822663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,1,power_law_1.01,0.05617920160293579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,1,power_law_1.01,0.0371071994304657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,1,power_law_1.01,0.01820800006389618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,1,power_law_1.01,0.04110719859600067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,1,power_law_1.01,0.0568448007106781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,1,power_law_1.01,0.01852799952030182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,1,power_law_1.01,0.040031999349594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,1,balanced,2.9490559895833335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,1,power_law_1.01,0.019424000382423402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,1,power_law_1.01,0.056403201818466184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,1,power_law_1.01,0.04165120124816894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,1,balanced,0.545136014620463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,1,power_law_1.01,0.02035840004682541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,1,power_law_1.01,0.04259200096130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,1,power_law_1.01,0.05591040253639221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,1,power_law_1.01,0.02014079988002777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,1,power_law_1.01,0.05667200088500977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,1,power_law_1.01,0.0451200008392334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,1,power_law_1.01,0.021779200434684752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,1,power_law_1.01,0.055718398094177245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,1,power_law_1.01,0.023104000091552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,1,power_law_1.01,0.047731199860572816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,1,power_law_1.01,0.027276799082756042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,1,power_law_1.01,0.0573311984539032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,1,power_law_1.01,0.050911998748779295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,1,power_law_1.01,0.02446720004081726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,1,power_law_1.01,0.056480002403259275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,1,power_law_1.01,0.0548799991607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,1,power_law_1.01,0.02592639923095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,1,power_law_1.01,0.06531839966773986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,1,power_law_1.01,0.056627202033996585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,1,power_law_1.01,0.028883200883865357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,1,power_law_1.01,0.030623999238014222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,1,power_law_1.01,0.06776319742202759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,1,power_law_1.01,0.060159999132156375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,1,power_law_1.01,0.037049600481987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,1,power_law_1.01,0.07430400252342224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,1,power_law_1.01,0.06380800008773804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,1,balanced,1.058666706085205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.01,0.040524798631668094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,1,power_law_1.01,0.0900223970413208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,1,power_law_1.01,0.07062399983406067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.01,0.053881597518920896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,1,power_law_1.01,0.09545599818229675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,1,power_law_1.01,0.07245439887046815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.01,0.05983999967575073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,1,power_law_1.01,0.12206720113754273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.01,0.07964800000190735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,1,power_law_1.01,0.13451520204544068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,1,power_law_1.01,0.07584000229835511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.01,0.09504640102386475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,1,power_law_1.01,0.17944320440292358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,1,power_law_1.01,0.09015679955482483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.01,0.12521599531173705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,1,power_law_1.01,0.22493441104888917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,1,power_law_1.2,0.0331712007522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.01,0.15799039602279663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,1,power_law_1.01,0.31794559955596924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.01,0.09982720017433167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,1,power_law_1.01,0.4060351848602295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.01,0.2204416036605835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.01,0.12242560386657715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,1,power_law_1.2,0.0350271999835968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,1,power_law_1.01,0.5808767795562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.01,0.2846400022506714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,1,power_law_1.2,0.038649600744247434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.01,0.14205440282821655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,1,power_law_1.01,0.7546175956726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,1,power_law_1.2,0.055846399068832396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.01,0.3491519927978516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.01,0.1866368055343628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,1,power_law_1.2,0.05671039819717407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,1,power_law_1.01,0.9268223762512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,1,power_law_1.2,0.05723519921302796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,1,power_law_1.01,0.5419328212738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.01,0.23500800132751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,1,power_law_1.2,0.057068800926208495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,1,power_law_1.01,1.4440128326416015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,1,power_law_1.01,1.0552767753601073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,1,power_law_1.2,0.058380800485610965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,1,power_law_1.2,0.05914880037307739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.01,0.32584960460662843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,1,power_law_1.01,2.8168767929077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,1,power_law_1.2,0.05852800011634827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.01,0.4171328067779541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,1,power_law_1.2,0.0599295973777771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,1,power_law_1.2,0.06212480068206787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.01,0.5948031902313232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,1,power_law_1.2,0.06843519806861878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,1,power_law_1.2,0.06725119948387145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.01,0.775987195968628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,1,power_law_1.2,0.08024960160255432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,1,power_law_1.2,0.08278399705886841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.01,0.955571174621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,1,power_law_1.2,0.08391680121421814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,1,power_law_1.2,0.09505280256271362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,1,power_law_1.01,1.4960127830505372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,1,power_law_1.2,0.09945600032806397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,1,power_law_1.2,0.12396160364151002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,1,power_law_1.2,0.14345599412918092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,1,power_law_1.01,2.9578624725341798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,1,power_law_1.2,0.1864575982093811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,1,power_law_1.2,0.22704000473022462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,1,power_law_1.2,0.3094016075134277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,1,power_law_1.2,0.3790719985961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,1,power_law_1.2,0.5330111980438232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,1,power_law_1.2,0.6805439949035644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,1,power_law_1.2,0.8299839973449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,1,power_law_1.2,1.2833727836608886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,1,power_law_1.2,2.500089645385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,1,power_law_1.2,0.029529601335525513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,1,power_law_1.2,0.03079040050506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,1,power_law_1.2,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,1,power_law_1.2,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,1,power_law_1.2,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,1,power_law_1.2,0.03841919898986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,1,power_law_1.2,0.04219520092010498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,1,power_law_1.2,0.04021120071411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,1,power_law_1.2,0.04286080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,1,power_law_1.2,0.0438975989818573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,1,power_law_1.2,0.0458624005317688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,1,power_law_1.2,0.05010560154914856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,1,power_law_1.2,0.051481598615646364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,1,power_law_1.2,0.05497599840164184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,1,power_law_1.2,0.06700800061225891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,1,power_law_1.2,0.04102399945259094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,1,power_law_1.2,0.06766080260276794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,1,power_law_1.2,0.04319359958171844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,1,power_law_1.2,0.07420799732208253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,1,power_law_1.2,0.04561919867992401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,1,power_law_1.2,0.09067519903182983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,1,power_law_1.2,0.05627520084381103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,1,power_law_1.2,0.09516159892082214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,1,power_law_1.2,0.05806080102920532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,1,power_law_1.2,0.1227455973625183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,1,power_law_1.2,0.056032001972198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,1,power_law_1.2,0.05605760216712952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,1,power_law_1.2,0.13659520149230958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,1,power_law_1.2,0.055718398094177245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,1,power_law_1.2,0.05496320128440857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,1,power_law_1.2,0.18321919441223145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,1,power_law_1.2,0.05702400207519531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,1,power_law_1.2,0.22403199672698976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,1,power_law_1.2,0.055264002084732054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,1,power_law_1.2,0.05715199708938599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,1,power_law_1.2,0.3160831928253174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,1,power_law_1.2,0.05966079831123352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,1,power_law_1.2,0.06214399933815003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,1,power_law_1.2,0.4061888217926025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,1,power_law_1.2,0.06988800168037415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,1,power_law_1.2,0.5818240165710449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,1,power_law_1.2,0.06824319958686828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,1,power_law_1.2,0.07524480223655701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,1,power_law_1.2,0.7551551818847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,1,power_law_1.2,0.0894976019859314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,1,power_law_1.2,0.9266752243041992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.2,0.09807999730110169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.2,0.12015360593795776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,1,power_law_1.2,1.4424575805664062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,1,power_law_1.2,0.016684800386428833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.2,0.14063359498977662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.2,0.1870527982711792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,1,power_law_1.2,2.8199743270874023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,1,power_law_1.2,0.017228800058364867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.2,0.23193600177764892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.2,0.3246272087097168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,1,power_law_1.2,0.016940799355506898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.2,0.4135424137115479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,1,power_law_1.2,0.017555199563503265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.2,0.5939455986022949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,1,power_law_1.2,0.018406400084495546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.2,0.7738944053649902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,1,power_law_1.2,0.018080000579357148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.2,0.9538496017456055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,1,power_law_1.2,0.018611200153827667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,1,power_law_1.2,1.4997823715209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,1,power_law_1.2,0.01966720074415207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,1,power_law_1.2,2.9409088134765624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,8,balanced,0.08061333497365315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,1,power_law_1.2,0.020351999998092653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,1,power_law_1.2,0.021004800498485566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,8,balanced,0.08259200056393941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,1,power_law_1.2,0.02126079946756363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,8,balanced,0.08319999774297078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,1,power_law_1.2,0.022886399924755097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,1,power_law_1.2,0.026924800872802735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,8,balanced,0.08081600069999695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,1,power_law_1.2,0.024985599517822265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,8,balanced,0.08277866741021474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,1,power_law_1.2,0.026022401452064515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,1,power_law_1.2,0.029023998975753786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,8,balanced,0.0807360013326009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,1,power_law_1.2,0.030611199140548707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,8,balanced,0.08095466593901317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,1,power_law_1.2,0.037062400579452516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,8,balanced,0.0830506682395935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.2,0.040729600191116336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,8,balanced,0.08203733464082082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.2,0.05358719825744629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.2,0.05924479961395264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,8,balanced,0.08141333361466725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.2,0.07927039861679078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,8,balanced,0.084714670976003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.2,0.0956928014755249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,8,balanced,0.08453866839408875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.2,0.12524160146713256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,8,balanced,0.0848479966322581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.2,0.1590208053588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,8,balanced,0.09001066287358601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.2,0.22197120189666747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,8,balanced,0.09354666868845622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,8,balanced,0.09539199868837993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.2,0.2857599973678589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,8,balanced,0.09703999757766724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.2,0.34980480670928954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,8,balanced,0.12065066893895467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,1,power_law_1.2,0.5419519901275635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,8,balanced,0.11655466755231221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,1,power_law_1.2,1.0571840286254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,8,balanced,0.13398399949073792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,8,balanced,0.16260799765586853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,8,balanced,0.05596800148487091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,8,balanced,0.07622933387756348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,8,balanced,0.07499200105667114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,8,balanced,0.05853333572546641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,8,balanced,0.19874666134516397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,8,balanced,0.07482133309046428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,8,balanced,0.05613866448402405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,8,balanced,0.08027733365694682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,8,balanced,0.05978133281071981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,8,balanced,0.24698134263356528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,8,balanced,0.05813866853713989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,8,balanced,0.07647466659545898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,8,balanced,0.059893334905306496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,8,balanced,0.07548800110816956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,8,balanced,0.3309066692988078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,8,balanced,0.06091199815273285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,8,balanced,0.07859733204046886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,8,balanced,0.060922667384147644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,8,balanced,0.07532800237337749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,8,balanced,0.027679999669392902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,8,balanced,0.06052800019582113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,8,balanced,0.07869866490364075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,8,balanced,0.061893333991368614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,8,balanced,0.41042133172353107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,8,balanced,0.07840533554553986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,8,balanced,0.02752000093460083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,8,balanced,0.06325866778691609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,8,balanced,0.025242666403452556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,8,balanced,0.07649066547552745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,8,power_law_1.01,0.0804032027721405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,8,balanced,0.02585600068171819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,8,balanced,0.06426133215427399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,8,balanced,0.07462933162848155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,8,balanced,0.025392000873883564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,8,balanced,0.07074133555094402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,8,balanced,0.06378133098284404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,8,power_law_1.01,0.08095359802246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,8,balanced,0.03088533381621043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,8,balanced,0.5774666468302408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,8,balanced,0.0728053351243337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,8,balanced,0.045423999428749084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,8,balanced,0.07044800122578938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,8,power_law_1.01,0.08179839849472045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,8,balanced,0.07106666763623555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,8,balanced,0.037717332442601524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,8,balanced,0.07677866518497467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,8,balanced,0.05212800204753876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,8,balanced,0.07077333331108093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,8,power_law_1.01,0.07927039861679078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,8,balanced,0.04548799991607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,8,balanced,0.0755680004755656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,8,balanced,0.06990933418273926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,8,power_law_1.01,0.07973120212554932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,8,balanced,0.04630400240421295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,8,balanced,0.7632213433583578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,8,balanced,0.08063466846942902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,8,balanced,0.08463467160860698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,8,balanced,0.09514133135477702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,8,power_law_1.01,0.08004480004310607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,8,balanced,0.08010133107503255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,8,power_law_1.01,0.08615040183067321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,8,balanced,0.09060800075531006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,8,balanced,0.09358400106430054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,8,balanced,0.06237333516279856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,8,power_law_1.01,0.08872960209846496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,8,balanced,0.09916266798973083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,8,balanced,0.10749866565068562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,8,balanced,0.09015466769536336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,8,power_law_1.01,0.092985600233078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,8,balanced,0.1236853301525116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,8,balanced,0.07514666517575581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,8,balanced,0.12716266512870789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,8,power_law_1.01,0.09478399753570557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,8,balanced,0.9501919746398926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,8,balanced,0.09659733374913533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,8,balanced,0.14087466398874918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,8,power_law_1.01,0.09400960206985473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,8,balanced,0.16569067041079202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,8,balanced,0.12026666601498921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,8,power_law_1.01,0.09845759868621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,8,balanced,0.14492799838383993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,8,balanced,0.1748159925142924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,8,power_law_1.01,0.10286719799041748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,8,balanced,0.20523732900619507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,8,balanced,0.20334400733311972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,8,power_law_1.01,0.11250560283660889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,8,balanced,0.21456533670425415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,8,balanced,0.2592746615409851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,8,power_law_1.01,0.11930880546569825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,8,balanced,0.27297067642211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,8,power_law_1.01,0.12186880111694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,8,balanced,0.3601653178532918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,8,balanced,0.2892746726671855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,8,power_law_1.01,0.1522752046585083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,8,power_law_1.01,0.20158720016479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,8,balanced,1.4653654098510742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,8,balanced,0.4699999888737996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,8,balanced,0.35023999214172363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,8,power_law_1.01,0.24015359878540038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,8,balanced,0.36105600992838544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,8,power_law_1.01,0.3350656032562256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,8,power_law_1.01,0.3746880054473877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,8,balanced,0.6809173425038656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,8,power_law_1.01,0.5287936210632325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,8,balanced,0.4963839848836263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,8,power_law_1.01,0.7458879947662354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,8,balanced,0.512880007425944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,8,balanced,0.8876372973124186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,8,power_law_1.01,0.9937472343444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,8,power_law_1.01,1.3608320236206055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,8,balanced,0.6423466602961222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,8,power_law_1.01,2.1092096328735352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,8,balanced,1.3109652996063232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,8,power_law_1.01,2.7390079498291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,8,balanced,0.6715520222981771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,8,power_law_1.01,3.3691646575927736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,8,power_law_1.01,5.311398315429687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,8,balanced,2.8456265131632485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,8,balanced,0.7955733140309652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,8,balanced,1.7303840319315593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,8,power_law_1.01,10.5627197265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,8,balanced,0.8240799903869629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,8,balanced,2.14516274134318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,8,balanced,1.2480746905008953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,8,power_law_1.01,0.055852800607681274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,8,power_law_1.01,0.05628160238265991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,8,balanced,1.257424036661784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,8,power_law_1.01,0.0541055977344513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,8,power_law_1.01,0.05742080211639404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,8,power_law_1.01,0.05865600109100342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,8,power_law_1.01,0.06032000184059143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,8,power_law_1.01,0.06193280220031738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,8,balanced,3.4022560119628906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,8,power_law_1.01,0.06600319743156433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,8,power_law_1.01,0.06865919828414917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,8,power_law_1.01,0.06973440051078797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,8,power_law_1.01,0.06789759993553161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,8,power_law_1.01,0.07174400091171265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,8,power_law_1.01,0.07341439723968506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,8,power_law_1.01,0.0828544020652771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,8,balanced,2.4506026903788247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,8,power_law_1.01,0.09453439712524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,8,power_law_1.01,0.0967423975467682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,8,balanced,2.433039983113607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,8,power_law_1.01,0.1157696008682251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,8,power_law_1.01,0.08658559918403626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,8,power_law_1.01,0.1411967992782593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,8,power_law_1.01,0.16699520349502564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,8,power_law_1.01,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,8,power_law_1.01,0.2232448101043701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,8,power_law_1.01,0.07964800000190735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,8,power_law_1.01,0.2483839988708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,8,power_law_1.01,0.08170880079269409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,8,power_law_1.01,0.38581759929656984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,8,power_law_1.01,0.08133760094642639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,8,balanced,6.7446028391520185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,8,power_law_1.01,0.4501952171325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,8,power_law_1.01,0.07323520183563233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,8,power_law_1.01,0.07763199806213379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,8,power_law_1.01,0.6576896190643311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,8,power_law_1.01,0.07276160120964051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,8,power_law_1.01,0.9259776115417481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,8,power_law_1.01,0.07155839800834655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,8,power_law_1.01,1.3914431571960448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,8,power_law_1.01,0.07078400254249573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,8,power_law_1.01,0.05960320234298706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,8,power_law_1.01,1.7137727737426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,8,power_law_1.01,0.07486079931259156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,8,power_law_1.01,0.09034879803657532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,8,power_law_1.01,2.4535039901733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,8,power_law_1.01,0.08310400247573853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,8,power_law_1.01,3.3840641021728515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,8,power_law_1.01,0.09349120259284974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,8,power_law_1.01,6.641305541992187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,8,power_law_1.01,0.09904000163078308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,8,power_law_1.01,0.11164799928665162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,8,power_law_1.01,0.15066239833831788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.01,0.1675711989402771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.01,0.22387840747833251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.01,0.29051520824432375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,8,power_law_1.01,0.02584959864616394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.01,0.37953920364379884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.01,0.5047423839569092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.01,0.7268415927886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,8,power_law_1.01,0.025644800066947936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.01,0.9357952117919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.01,1.268166446685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,8,power_law_1.01,0.024115200340747833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.01,1.6218496322631837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,8,power_law_1.01,0.024243199825286867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,8,power_law_1.01,0.024748800694942473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,8,power_law_1.01,2.2731391906738283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,8,power_law_1.01,0.02892799973487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,8,power_law_1.01,0.044012799859046936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,8,power_law_1.01,3.990758514404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,8,power_law_1.01,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,8,power_law_1.01,7.315647888183594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,8,power_law_1.01,0.05015680193901062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,8,power_law_1.01,0.043680000305175784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,8,power_law_1.01,0.04607360064983368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,8,power_law_1.01,0.08232960104942322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,8,power_law_1.01,0.07960320115089417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,8,power_law_1.01,0.05958399772644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,8,power_law_1.01,0.08871039748191833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,8,power_law_1.01,0.07305600047111512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,8,power_law_1.01,0.0950976014137268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,8,power_law_1.01,0.11870080232620239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.01,0.14283519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.01,0.2032831907272339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.01,0.25868160724639894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.01,0.3584320068359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.01,0.4689023971557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.01,0.6793791770935058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.01,0.8880064010620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.01,1.3081727981567384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.01,1.7271808624267577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,8,power_law_1.01,2.1465728759765623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,8,power_law_1.01,3.413343811035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,8,power_law_1.01,6.7537788391113285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,2,8,power_law_1.2,0.08030719757080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,2,8,power_law_1.2,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,2,8,power_law_1.2,0.08133760094642639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,2,8,power_law_1.2,0.08132479786872863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,2,8,power_law_1.2,0.05663359761238098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,2,8,power_law_1.2,0.07964800000190735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,2,8,power_law_1.2,0.055743998289108275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,2,8,power_law_1.2,0.05866879820823669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,2,8,power_law_1.2,0.07926399707794189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,2,8,power_law_1.2,0.05909759998321533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,2,8,power_law_1.2,0.08109440207481385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,2,8,power_law_1.2,0.06058239936828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,2,8,power_law_1.2,0.06458240151405334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,2,8,power_law_1.2,0.08067839741706848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,2,8,power_law_1.2,0.06612480282783509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,2,8,power_law_1.2,0.08911359906196595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,2,8,power_law_1.2,0.0706816017627716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,2,8,power_law_1.2,0.09570559859275818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,2,8,power_law_1.2,0.0714303970336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,2,8,power_law_1.2,0.09376639723777772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,2,8,power_law_1.2,0.06617599725723267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,2,8,power_law_1.2,0.09438719749450683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,2,8,power_law_1.2,0.0757311999797821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,2,8,power_law_1.2,0.10049279928207397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,2,8,power_law_1.2,0.0756991982460022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,2,8,power_law_1.2,0.0827455997467041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,2,8,power_law_1.2,0.10147839784622192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,2,8,power_law_1.2,0.09745280146598816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,2,8,power_law_1.2,0.1097216010093689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,2,8,power_law_1.2,0.09875839948654175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,2,8,power_law_1.2,0.12558079957962037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,2,8,power_law_1.2,0.11691520214080811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,2,8,power_law_1.2,0.12394880056381226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,2,8,power_law_1.2,0.14336639642715454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,2,8,power_law_1.2,0.15526399612426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,2,8,power_law_1.2,0.17431039810180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,4,balanced,0.05384000142415365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,2,8,power_law_1.2,0.20213119983673095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,2,8,power_law_1.2,0.2292479991912842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,2,8,power_law_1.2,0.08520960211753845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,4,balanced,0.03790933390458425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,2,8,power_law_1.2,0.02627840042114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,2,8,power_law_1.2,0.25537919998168945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,4,balanced,0.05574933191140493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,2,8,power_law_1.2,0.27127039432525635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,2,8,power_law_1.2,0.3458751916885376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,2,8,power_law_1.2,0.365120005607605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,2,8,power_law_1.2,0.07832959890365601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,4,balanced,0.08017066617806752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,2,8,power_law_1.2,0.0255295991897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,2,8,power_law_1.2,0.38889598846435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,2,8,power_law_1.2,0.49383039474487306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,4,balanced,0.03738133360942205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,2,8,power_law_1.2,0.07950080037117005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,4,balanced,0.07724800209204356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,2,8,power_law_1.2,0.024748800694942473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,2,8,power_law_1.2,0.553926420211792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,2,8,power_law_1.2,0.7450111865997314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,4,balanced,0.05380799869696299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,4,balanced,0.07905599971612294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,2,8,power_law_1.2,0.07976319789886474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,2,8,power_law_1.2,0.02439039945602417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,2,8,power_law_1.2,0.7536320209503173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,4,balanced,0.05620799958705902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,2,8,power_law_1.2,0.8575936317443847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,4,balanced,0.05620799958705902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,4,balanced,0.07834666470686595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,2,8,power_law_1.2,0.08010879755020142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,2,8,power_law_1.2,0.024672000110149382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,2,8,power_law_1.2,1.3196928024291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,2,8,power_law_1.2,1.0766592025756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,4,balanced,0.05781333148479462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,4,balanced,0.08084799846013387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,2,8,power_law_1.2,0.07447680234909057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,2,8,power_law_1.2,0.029875200986862183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,4,balanced,0.05996266504128774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,2,8,power_law_1.2,1.4738368034362792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,2,8,power_law_1.2,1.922559928894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,4,balanced,0.058261334896087646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,4,balanced,0.07952000200748444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,2,8,power_law_1.2,0.07352319955825806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,2,8,power_law_1.2,0.0433023989200592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,4,balanced,0.05840000013510386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,2,8,power_law_1.2,2.067168045043945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,2,8,power_law_1.2,2.3934911727905273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,4,balanced,0.07964799801508586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,4,balanced,0.05994133154551188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,2,8,power_law_1.2,0.07221119999885559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,2,8,power_law_1.2,0.03676159977912903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,2,8,power_law_1.2,2.6025728225708007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,4,balanced,0.0621066689491272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,4,balanced,0.07918400069077809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,2,8,power_law_1.2,3.389811325073242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,2,8,power_law_1.2,0.04925439953804016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,2,8,power_law_1.2,0.06954240202903747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,4,balanced,0.06378666559855144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,2,8,power_law_1.2,3.5508480072021484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,4,balanced,0.062133332093556724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,4,balanced,0.08349333206812541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,2,8,power_law_1.2,0.04312959909439087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,2,8,power_law_1.2,0.06528000235557556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,4,balanced,0.06649066507816315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,2,8,power_law_1.2,6.656121826171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,2,8,power_law_1.2,0.045433598756790164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,4,balanced,0.0820000022649765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,4,balanced,0.07611200213432312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,2,8,power_law_1.2,4.87196159362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,2,8,power_law_1.2,0.06280959844589233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,4,balanced,0.07453866799672444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,2,8,power_law_1.2,0.08520960211753845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,4,balanced,0.08378666639328003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,2,8,power_law_1.2,0.08312960267066956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,4,balanced,0.07655466596285503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,2,8,power_law_1.2,0.07846400141716003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,4,balanced,0.08449066678682964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,2,8,power_law_1.2,10.60904312133789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,2,8,power_law_1.2,0.09074559807777405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,4,balanced,0.09503466884295146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,4,balanced,0.05461333195368449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,2,8,power_law_1.2,0.06053119897842407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,4,balanced,0.09082133571306865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,4,balanced,0.09854400157928467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,2,8,power_law_1.2,0.08091520071029663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,4,balanced,0.055455997586250305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,2,8,power_law_1.2,0.09152640104293823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,4,balanced,0.07640000184377034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,4,balanced,0.09359467029571533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,4,balanced,0.11932800213495891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,2,8,power_law_1.2,0.09723520278930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,4,balanced,0.07373866438865662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,2,8,power_law_1.2,0.07299839854240417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,4,balanced,0.07035733262697856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,2,8,power_law_1.2,0.09744639992713929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,4,balanced,0.09876267115275066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,4,balanced,0.07276266813278198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,4,balanced,0.13958932956059775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,2,8,power_law_1.2,0.09401599764823913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,4,balanced,0.0765173335870107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,2,8,power_law_1.2,0.10933120250701904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,4,balanced,0.11215466260910034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,4,balanced,0.07859200239181519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,2,8,power_law_1.2,0.11738879680633545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,4,balanced,0.1774239937464396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,4,balanced,0.0751200020313263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,2,8,power_law_1.2,0.13346560001373292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.2,0.1429695963859558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,4,balanced,0.1200373371442159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,4,balanced,0.08284266789754231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.2,0.17749119997024537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,4,balanced,0.07257066667079926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,4,balanced,0.21703465779622397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.2,0.2005631923675537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,4,balanced,0.16665599743525186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,4,balanced,0.07392533123493195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.2,0.2292799949645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.2,0.2565567970275879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,4,balanced,0.07081066568692525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.2,0.2603327989578247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,4,balanced,0.16539200146993002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,4,balanced,0.06820799907048543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,4,balanced,0.2985759973526001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.2,0.3589632034301758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,4,balanced,0.07347199817498525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.2,0.3611776113510132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,4,balanced,0.0708426684141159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.2,0.46851840019226076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,4,balanced,0.20907199382781982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,4,balanced,0.07006399830182393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.2,0.5009215831756592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.2,0.6772736072540283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,4,balanced,0.0941493312517802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,4,balanced,0.36904001235961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.2,0.7392127990722657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,4,balanced,0.09233066439628601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.2,0.8857536315917969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,4,balanced,0.25009065866470337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.2,1.023481559753418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.2,1.3118975639343262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,4,balanced,0.10920000076293945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,4,balanced,0.3335946798324585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,4,balanced,0.13170133034388223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.2,1.7265600204467773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.2,1.489568042755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,4,balanced,0.5241173505783081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,4,balanced,0.1693013310432434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,2,8,power_law_1.2,2.147161674499512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.2,1.821299171447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,4,balanced,0.40987201531728107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,4,balanced,0.2070186734199524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,2,8,power_law_1.2,2.490713691711426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,2,8,power_law_1.2,3.403392028808594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,4,balanced,0.021456000705560047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,2,8,power_law_1.2,3.9818561553955076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,4,balanced,0.2835893432299296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,4,balanced,0.021744000415007275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,2,8,power_law_1.2,6.779705810546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,4,balanced,0.673583984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,4,balanced,0.022389332453409832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,4,balanced,0.6077119906743368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,2,8,power_law_1.2,7.812397003173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,4,balanced,0.021242665747801464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,4,balanced,0.3582133452097575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,4,balanced,0.02128533273935318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,4,balanced,0.02348800003528595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,4,balanced,0.029946667452653248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,4,balanced,0.02717333287000656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,4,balanced,0.8291467030843099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,4,balanced,0.03555733213822047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,4,balanced,0.5136586825052897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,4,balanced,0.7697227001190186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,4,balanced,0.03153600047032038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,4,balanced,0.03562133262554804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,4,balanced,0.03984000037113825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,4,balanced,0.04014399896065394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,4,balanced,0.04153066625197729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,4,balanced,0.6674026648203532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,4,balanced,0.06443200012048085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,4,balanced,0.05333333214124044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,4,balanced,0.9471893310546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,4,balanced,0.05990933378537496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,4,balanced,1.2710879643758137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,4,balanced,0.07461866736412048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,4,balanced,0.09227200349171956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,4,balanced,0.8178880214691162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,4,balanced,0.12781332929929098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,4,balanced,0.15279466907183328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,4,balanced,0.21225599447886148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,4,balanced,1.4544639587402344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,4,balanced,0.27075199286142987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,4,balanced,1.28056534131368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,4,balanced,0.38651732603708905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,4,balanced,2.4661760330200195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,4,balanced,0.5021653175354004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,4,balanced,0.7331039905548096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,4,balanced,2.837631861368815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,4,balanced,0.959824005762736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,4,balanced,2.5360159873962402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,4,balanced,1.1910186608632405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,4,balanced,1.8839732805887859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,4,balanced,3.725536028544108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,4,power_law_1.01,0.04776960015296936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,4,power_law_1.01,0.06496639847755432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,4,power_law_1.01,0.07035520076751708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,4,power_law_1.01,0.07334399819374085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,4,power_law_1.01,0.07706239819526672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,4,power_law_1.01,0.05084800124168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,4,power_law_1.01,0.07838079929351807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,4,power_law_1.01,0.07860479950904846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,4,power_law_1.01,0.049132800102233885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,4,power_law_1.01,0.08169599771499633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,4,power_law_1.01,0.05594239830970764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,4,power_law_1.01,0.08413439989089966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,4,power_law_1.01,0.05585920214653015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,4,power_law_1.01,0.08769279718399048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,4,power_law_1.01,0.059084802865982056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,4,power_law_1.01,0.09404799938201905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,4,power_law_1.01,0.060819202661514284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,4,power_law_1.01,0.05961599946022034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,4,power_law_1.01,0.08604800105094909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,4,power_law_1.01,0.06282240152359009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,4,power_law_1.01,0.06302719712257385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,4,power_law_1.01,0.10079360008239746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,4,power_law_1.01,0.06650879979133606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,4,power_law_1.01,0.09447680115699768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,4,power_law_1.01,0.06773120164871216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,4,power_law_1.01,0.10840320587158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,4,power_law_1.01,0.07027199864387512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,4,power_law_1.01,0.0711296021938324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,4,power_law_1.01,0.12347520589828491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,4,power_law_1.01,0.07343360185623168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,4,power_law_1.01,0.11940480470657348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,4,power_law_1.01,0.08421120047569275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,4,power_law_1.01,0.0704255998134613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,4,power_law_1.01,0.08542720079421998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,4,power_law_1.01,0.1367743968963623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,4,power_law_1.01,0.07104640007019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,4,power_law_1.01,0.09388800263404846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,4,power_law_1.01,0.16757760047912598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,4,power_law_1.01,0.11971839666366577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,4,power_law_1.01,0.07483519911766053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,4,power_law_1.01,0.19843200445175171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,4,power_law_1.01,0.13299839496612548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,4,power_law_1.01,0.07331200242042542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,4,power_law_1.01,0.25084800720214845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,4,power_law_1.01,0.16885759830474853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,4,power_law_1.01,0.07588480114936828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,4,power_law_1.01,0.231276798248291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,4,power_law_1.01,0.31556479930877684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,4,power_law_1.01,0.0222463995218277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,4,power_law_1.01,0.3462464094161987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,4,power_law_1.01,0.07272319793701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,4,power_law_1.01,0.436678409576416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,4,power_law_1.01,0.3762815952301025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,4,power_law_1.01,0.07192320227622986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,4,power_law_1.01,0.46698880195617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,4,power_law_1.01,0.02129279971122742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,4,power_law_1.01,0.5351871967315673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,4,power_law_1.01,0.0716543972492218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,4,power_law_1.01,0.5885183811187744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,4,power_law_1.01,0.5331647872924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,4,power_law_1.01,0.07046399712562561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,4,power_law_1.01,0.020979200303554536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,4,power_law_1.01,1.036409568786621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,4,power_law_1.01,1.0324159622192384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,4,power_law_1.01,0.06870399713516236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,4,power_law_1.01,0.020787200331687926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,4,power_law_1.01,1.3301440238952638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,4,power_law_1.01,1.4553728103637695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,4,power_law_1.01,0.07723519802093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,4,power_law_1.01,0.021779200434684752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,4,power_law_1.01,1.3862848281860352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,4,power_law_1.01,2.0925056457519533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,4,power_law_1.01,0.07836800217628478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,4,power_law_1.01,0.022463999688625336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,4,power_law_1.01,2.439558410644531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,4,power_law_1.01,0.029446399211883544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,4,power_law_1.01,0.08060160279273987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,4,power_law_1.01,2.4942911148071287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,4,power_law_1.01,4.315008163452148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,4,power_law_1.01,0.08748800158500672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,4,power_law_1.01,0.027187201380729675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,4,power_law_1.01,0.09185280203819275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,4,power_law_1.01,4.385644912719727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,4,power_law_1.01,0.033292800188064575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,4,power_law_1.01,0.10609279870986939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,4,power_law_1.01,0.030297601222991945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,4,power_law_1.01,0.12836480140686035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,4,power_law_1.01,0.03457280099391937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,4,power_law_1.01,7.384352111816407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.01,0.14380799531936644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,4,power_law_1.01,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.01,0.16535040140151977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,4,power_law_1.01,0.039673599600791934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.01,0.20751359462738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,4,power_law_1.01,0.04012799859046936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.01,0.29656960964202883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,4,power_law_1.01,0.059552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.01,0.3762880086898804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,4,power_law_1.01,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.01,0.6708992004394532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,4,power_law_1.01,0.05886719822883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.01,0.8211392402648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,4,power_law_1.01,0.07303680181503296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.01,1.0922559738159179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,4,power_law_1.2,0.0779263973236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.01,0.09132159948348999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,4,power_law_1.2,0.06367999911308289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.01,0.12572799921035765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.01,1.3853695869445801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,4,power_law_1.2,0.07756159901618957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.01,0.15104000568389891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,4,power_law_1.2,0.07702400088310242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,4,power_law_1.01,1.5327808380126953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.01,0.21000959873199462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,4,power_law_1.2,0.0778880000114441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.01,0.26730239391326904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,4,power_law_1.2,0.07923840284347534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,4,power_law_1.01,2.5006784439086913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,4,power_law_1.2,0.0818943977355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,4,power_law_1.2,0.04950399994850159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.01,0.38102400302886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,4,power_law_1.2,0.06750720143318176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,4,power_law_1.01,7.063641357421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,4,power_law_1.2,0.05186560153961182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,4,power_law_1.2,0.08281599879264831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,4,power_law_1.2,0.07328640222549439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.01,0.4986559867858887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,4,power_law_1.2,0.048179200291633605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,4,power_law_1.2,0.08888319730758668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,4,power_law_1.2,0.0536191999912262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.01,0.7297152042388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,4,power_law_1.2,0.08815360069274902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,4,power_law_1.2,0.07129600048065185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,4,power_law_1.2,0.05483520030975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,4,power_law_1.2,0.08574720025062561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.01,0.957369613647461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,4,power_law_1.2,0.05672320127487183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,4,power_law_1.2,0.09956480264663696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,4,power_law_1.2,0.07688959836959838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,4,power_law_1.2,0.06094080209732056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,4,power_law_1.2,0.10024960041046142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,4,power_law_1.01,1.18984956741333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,4,power_law_1.2,0.05973759889602661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,4,power_law_1.2,0.1076416015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,4,power_law_1.2,0.0747648000717163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,4,power_law_1.2,0.06280959844589233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,4,power_law_1.2,0.11424000263214111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,4,power_law_1.2,0.07439360022544861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,4,power_law_1.2,0.06457599997520447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,4,power_law_1.01,1.880633544921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,4,power_law_1.2,0.12712960243225097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,4,power_law_1.2,0.07327359914779663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,4,power_law_1.2,0.06435840129852295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,4,power_law_1.2,0.14128639698028564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,4,power_law_1.2,0.07420160174369812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,4,power_law_1.2,0.0686079978942871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,4,power_law_1.2,0.17989120483398438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,4,power_law_1.01,3.7204479217529296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,4,power_law_1.2,0.07114880084991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,4,power_law_1.2,0.2017728090286255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,4,power_law_1.2,0.07365120053291321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,4,power_law_1.2,0.07034239768981934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,4,power_law_1.2,0.07525119781494141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,4,power_law_1.2,0.256115198135376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,4,power_law_1.2,0.0804095983505249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,4,power_law_1.2,0.07070720195770264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,4,power_law_1.2,0.2798271894454956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,4,power_law_1.2,0.02111999988555908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,4,power_law_1.2,0.08512639999389648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,4,power_law_1.2,0.41251840591430666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,4,power_law_1.2,0.07182719707489013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,4,power_law_1.2,0.10120960474014282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,4,power_law_1.2,0.5569151878356934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,4,power_law_1.2,0.08099200129508972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,4,power_law_1.2,0.12547839879989625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,4,power_law_1.2,0.7416319847106934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,4,power_law_1.2,0.020883199572563172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,4,power_law_1.2,0.1369279980659485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,4,power_law_1.2,0.07646080255508422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,4,power_law_1.2,1.0689536094665528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,4,power_law_1.2,0.021196800470352172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,4,power_law_1.2,0.18743679523468018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,4,power_law_1.2,0.09905920028686524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,4,power_law_1.2,1.9997119903564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,4,power_law_1.2,0.2549823999404907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,4,power_law_1.2,0.091430401802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,4,power_law_1.2,2.127142333984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,4,power_law_1.2,0.020627200603485107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,4,power_law_1.2,0.32085120677948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,4,power_law_1.2,0.1031872034072876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,4,power_law_1.2,0.3702143907546997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,4,power_law_1.2,2.9401599884033205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,4,power_law_1.2,0.021158400177955627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,4,power_law_1.2,0.11975040435791015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,4,power_law_1.2,0.5903744220733642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,4,power_law_1.2,0.02255360037088394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,4,power_law_1.2,4.340697479248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.2,0.1493888020515442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,4,power_law_1.2,0.6522687911987305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,4,power_law_1.2,0.02917119860649109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.2,0.1929535984992981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,4,power_law_1.2,1.0477888107299804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,4,power_law_1.2,8.809407806396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,4,power_law_1.2,0.026111999154090883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.2,0.2449728012084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,4,power_law_1.2,1.5009984016418456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,4,power_law_1.2,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.2,0.3734976053237915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,4,power_law_1.2,1.6128896713256835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,4,power_law_1.2,0.03204480111598969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.2,0.416428804397583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,4,power_law_1.2,0.034169599413871765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,4,power_law_1.2,2.8302783966064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.2,0.5744832038879395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,4,power_law_1.2,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.2,0.8086400032043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,4,power_law_1.2,5.071270370483399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,4,power_law_1.2,0.040608000755310056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.2,1.2595392227172852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,4,power_law_1.2,0.039827200770378116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,4,power_law_1.2,0.062118399143219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.2,1.87127685546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,4,power_law_1.2,0.05220479965209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,4,power_law_1.2,2.0683967590332033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,4,power_law_1.2,0.0585919976234436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,4,power_law_1.2,2.9645952224731444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,4,power_law_1.2,0.07203199863433837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,4,power_law_1.2,7.903065490722656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.2,0.09063680171966552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.2,0.12407040596008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.2,0.15092480182647705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.2,0.21000959873199462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.2,0.2682751893997192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.2,0.38467841148376464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.2,0.49689598083496095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.2,0.730617618560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.2,0.9607551574707032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,4,power_law_1.2,1.1926848411560058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,4,power_law_1.2,1.8821632385253906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,4,power_law_1.2,3.722367858886719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,2,balanced,0.03356266766786575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,2,balanced,0.03703466554482778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,2,balanced,0.03548266738653183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,2,balanced,0.052842666705449425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,2,balanced,0.05402133365472158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,2,balanced,0.057114665706952415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,2,balanced,0.07799466451009114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,2,balanced,0.057674666245778404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,2,balanced,0.07714666426181793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,2,balanced,0.05819199979305267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,2,balanced,0.07808533310890198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,2,balanced,0.05983466903368632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,2,balanced,0.07646933197975159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,2,balanced,0.05913599828879038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,2,balanced,0.0791786660750707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,2,balanced,0.05989866455396017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,2,balanced,0.07898133496443431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,2,balanced,0.060864001512527466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,2,balanced,0.07753600180149078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,2,balanced,0.06223999957243601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,2,balanced,0.07791999975840251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,2,balanced,0.06437333424886067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,2,balanced,0.08080000181992848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,2,balanced,0.06266666452089946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,2,balanced,0.08075733482837677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,2,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,2,balanced,0.0664213349421819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,2,balanced,0.08124266564846039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,2,balanced,0.07773333291212718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,2,balanced,0.08661866188049316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,2,balanced,0.05599466462930044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,2,balanced,0.08935999870300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,2,balanced,0.07828266421953838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,2,balanced,0.06881066660086314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,2,balanced,0.09283199906349182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,2,balanced,0.08378666639328003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,2,balanced,0.0664213349421819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,2,balanced,0.06650666892528534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,2,balanced,0.09297600388526917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,2,balanced,0.10088533163070679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,2,balanced,0.07223999996980031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,2,balanced,0.1033066709836324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,2,balanced,0.1083679993947347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,2,balanced,0.07030400137106578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,2,balanced,0.12502933541933695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,2,balanced,0.11345600088437398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,2,balanced,0.076773335536321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,2,balanced,0.15241066614786783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,2,balanced,0.13988266388575235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,2,balanced,0.06739200154940288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,2,balanced,0.021295999487241108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,2,balanced,0.06821866830190022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,2,balanced,0.16928533713022867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,2,balanced,0.02109866589307785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,2,balanced,0.1998400092124939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,2,balanced,0.06617600222428639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,2,balanced,0.021013334393501282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,2,balanced,0.07049066821734111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,2,balanced,0.021007999777793884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,2,balanced,0.24665067593256632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,2,balanced,0.0668693333864212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,2,balanced,0.021168000996112823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,2,balanced,0.2203893264134725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,2,balanced,0.06642666459083557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,2,balanced,0.021242665747801464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,2,balanced,0.023056000471115112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,2,balanced,0.07088533540566762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,2,balanced,0.33609068393707275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,2,balanced,0.025349333882331848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,2,balanced,0.2649493416150411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,2,balanced,0.07218666871388753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,2,balanced,0.03149333347876867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,2,balanced,0.0705973356962204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,2,balanced,0.033520000676314034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,2,balanced,0.10270933310190837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,2,balanced,0.02920000006755193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,2,balanced,0.41972800095876056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,2,balanced,0.3568426767985026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,2,balanced,0.03161066770553589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,2,balanced,0.10004799564679463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,2,balanced,0.03332266708215078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,2,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,2,balanced,0.11627733707427979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,2,balanced,0.04018666595220566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,2,balanced,0.4288320144017537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,2,balanced,0.6041920185089111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,2,balanced,0.14006400108337402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,2,balanced,0.04318933188915253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,2,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,2,balanced,0.18515199422836304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,2,balanced,0.0562666654586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,2,balanced,0.6206239859263102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,2,balanced,0.22816532850265503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,2,balanced,0.06168533364931742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,2,balanced,0.08299200236797333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,2,balanced,0.776917298634847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,2,balanced,0.3155413269996643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,2,balanced,0.10432533423105876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,2,balanced,0.13915733496348062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,2,balanced,0.7892639636993408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,2,balanced,0.40396801630655926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,2,balanced,0.16800000270207724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,2,balanced,0.9598506291707357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,2,balanced,0.5790293216705322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,2,balanced,0.23874133825302124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,2,balanced,0.9746560255686442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,2,balanced,0.30268265803654987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,2,power_law_1.01,0.03449600040912628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,2,power_law_1.01,0.0463808000087738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,2,balanced,0.7580426534016927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,2,balanced,0.4400213162104289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,2,power_law_1.01,0.0395904004573822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,2,power_law_1.01,0.055251199007034305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,2,power_law_1.01,0.03866879940032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,2,power_law_1.01,0.05921919941902161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,2,power_law_1.01,0.051686400175094606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,2,power_law_1.01,0.07306879758834839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,2,balanced,1.4795145988464355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,2,power_law_1.01,0.055878400802612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,2,balanced,1.5035999615987141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,2,power_law_1.01,0.07528319954872131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,2,balanced,0.9341973463694254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,2,balanced,0.576255997021993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,2,power_law_1.01,0.058387202024459836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,2,power_law_1.01,0.07601919770240784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,2,power_law_1.01,0.05824000239372253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,2,power_law_1.01,0.07608320116996765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,2,power_law_1.01,0.06209920048713684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,2,power_law_1.01,0.07747200131416321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,2,power_law_1.01,0.06208639740943909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,2,power_law_1.01,0.0812928020954132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,2,power_law_1.01,0.06275839805603027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,2,power_law_1.01,0.08236160278320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,2,balanced,0.7100266615549723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,2,power_law_1.01,0.06037759780883789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,2,power_law_1.01,0.06396160125732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,2,power_law_1.01,0.08479999899864196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,2,power_law_1.01,0.06595199704170226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,2,power_law_1.01,0.08416000008583069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,2,power_law_1.01,0.05864319801330566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,2,power_law_1.01,0.06769279837608337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,2,balanced,1.4597919782002766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,2,power_law_1.01,0.08937600255012512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,2,power_law_1.01,0.01929599940776825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,2,power_law_1.01,0.07306240200996399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,2,power_law_1.01,0.09761919975280761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,2,power_law_1.01,0.05980799794197082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,2,power_law_1.01,0.0815999984741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,2,power_law_1.01,0.10663679838180543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,2,power_law_1.01,0.01950719952583313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,2,power_law_1.01,0.0747648000717163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,2,power_law_1.01,0.08377599716186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,2,power_law_1.01,0.01959040015935898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,2,power_law_1.01,0.10737919807434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,2,power_law_1.01,0.06944640278816223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,2,power_law_1.01,0.019315199553966524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,2,power_law_1.01,0.09790719747543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,2,balanced,1.1141599814097087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,2,power_law_1.01,0.13034240007400513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,2,power_law_1.01,0.020019200444221497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,2,power_law_1.01,0.10775680541992187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,2,power_law_1.01,0.07215359807014465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,2,power_law_1.01,0.1360383987426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,2,balanced,2.8922506968180337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,2,power_law_1.01,0.020729599893093108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,2,balanced,2.9630613327026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,2,power_law_1.01,0.12845439910888673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,2,power_law_1.01,0.06877440214157104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,2,power_law_1.01,0.02160640060901642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,2,power_law_1.01,0.15067520141601562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,2,power_law_1.01,0.02389120012521744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,2,power_law_1.01,0.15168000459671022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,2,power_law_1.01,0.0686016023159027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,2,power_law_1.01,0.18138240575790404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,2,power_law_1.01,0.03078399896621704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,2,power_law_1.01,0.17745920419692993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,2,power_law_1.01,0.24423038959503174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,2,power_law_1.01,0.07013760209083557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,2,power_law_1.01,0.0316864013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,2,power_law_1.01,0.25948801040649416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,2,power_law_1.01,0.3044287919998169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,2,power_law_1.01,0.02720640003681183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,2,power_law_1.01,0.06891520023345947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,2,power_law_1.01,0.33753600120544436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,2,power_law_1.01,0.031353598833084105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,2,power_law_1.01,0.36110720634460447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,2,power_law_1.01,0.06829439997673034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,2,power_law_1.01,0.032332798838615416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,2,power_law_1.01,0.4272448062896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,2,power_law_1.01,0.544755220413208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,2,power_law_1.01,0.03409920036792755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,2,power_law_1.01,0.07363839745521546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,2,power_law_1.01,0.5379136085510254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,2,balanced,2.8978026707967124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,2,power_law_1.01,0.0387584000825882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,2,power_law_1.01,0.5789760112762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,2,power_law_1.01,0.04090240001678467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,2,power_law_1.01,0.8636032104492187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,2,power_law_1.01,0.0756928026676178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,2,power_law_1.01,0.852620792388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,2,power_law_1.01,0.04216319918632507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,2,power_law_1.01,1.1115455627441406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,2,power_law_1.01,0.05456640124320984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,2,power_law_1.01,0.0776639997959137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,2,power_law_1.01,1.0707903861999513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.01,0.059385597705841064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,2,balanced,2.192394733428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,2,power_law_1.01,1.3238080024719239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,2,power_law_1.01,1.4428288459777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,2,power_law_1.01,0.08489599823951721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.01,0.08112000226974488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.01,0.10193920135498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,2,power_law_1.01,0.09095039963722229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,2,power_law_1.01,2.1511552810668944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,2,power_law_1.01,2.065881538391113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.01,0.13575680255889894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,2,power_law_1.01,0.0996288001537323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.01,0.16425600051879882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,2,power_law_1.01,4.800614547729492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,2,power_law_1.01,4.074278259277344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,2,power_law_1.01,0.11443840265274048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.01,0.23617920875549317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.01,0.29989120960235593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.01,0.1285823941230774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.01,0.437011194229126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.01,0.1557760000228882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.01,0.5734655857086182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,2,power_law_1.01,0.7090240001678467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.01,0.19978879690170287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,2,power_law_1.01,1.1125568389892577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.01,0.24717440605163574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.01,0.33742079734802244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,2,power_law_1.01,2.1886463165283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.01,0.45189762115478516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.01,0.5860544204711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.01,0.8531071662902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.01,1.0831551551818848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,2,power_law_1.01,1.394918441772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,2,power_law_1.01,2.242265510559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,2,power_law_1.01,3.614080047607422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,2,power_law_1.2,0.047660800814628604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,2,power_law_1.2,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,2,power_law_1.2,0.05556480288505554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,2,power_law_1.2,0.05084159970283508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,2,power_law_1.2,0.0645632028579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,2,power_law_1.2,0.038950398564338684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,2,power_law_1.2,0.07322880029678344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,2,power_law_1.2,0.07582719922065735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,2,power_law_1.2,0.054745602607727054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,2,power_law_1.2,0.03886080086231232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,2,power_law_1.2,0.062009602785110474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,2,power_law_1.2,0.07637119889259339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,2,power_law_1.2,0.05538560152053833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,2,power_law_1.2,0.07381119728088378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,2,power_law_1.2,0.07715839743614197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,2,power_law_1.2,0.05549439787864685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,2,power_law_1.2,0.07080960273742676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,2,power_law_1.2,0.08117120265960694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,2,power_law_1.2,0.058432000875473025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,2,power_law_1.2,0.0699455976486206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,2,power_law_1.2,0.08366079926490784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,2,power_law_1.2,0.060262399911880496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,2,power_law_1.2,0.07009279727935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,2,power_law_1.2,0.06884480118751526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,2,power_law_1.2,0.08455680012702942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,2,power_law_1.2,0.062003201246261595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,2,power_law_1.2,0.0673088014125824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,2,power_law_1.2,0.0834879994392395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,2,power_law_1.2,0.061977601051330565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,2,power_law_1.2,0.0681984007358551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,2,power_law_1.2,0.08561279773712158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,2,power_law_1.2,0.06281599998474122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,2,power_law_1.2,0.0685696005821228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,2,power_law_1.2,0.019865599274635316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,2,power_law_1.2,0.08666239976882935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,2,power_law_1.2,0.06607999801635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,2,power_law_1.2,0.07630079984664917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,2,power_law_1.2,0.07500799894332885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,2,power_law_1.2,0.09820799827575684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,2,power_law_1.2,0.06668800115585327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,2,power_law_1.2,0.01964160054922104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,2,power_law_1.2,0.0757311999797821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,2,power_law_1.2,0.10366079807281495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,2,power_law_1.2,0.06801279783248901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,1,balanced,0.03379199902216593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,2,power_law_1.2,0.01958400011062622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,1,balanced,0.032511999209721885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,2,power_law_1.2,0.08831359744071961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,1,balanced,0.04827199876308441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,2,power_law_1.2,0.1030079960823059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,2,power_law_1.2,0.0751039981842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,2,power_law_1.2,0.019551999866962433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,2,power_law_1.2,0.0926144003868103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,2,power_law_1.2,0.12833280563354493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,2,power_law_1.2,0.08151040077209473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,2,power_law_1.2,0.020134399831295013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,1,balanced,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,1,balanced,0.03579200059175491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,2,power_law_1.2,0.09717119932174682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,1,balanced,0.053930665055910744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,2,power_law_1.2,0.08487039804458618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,2,power_law_1.2,0.020633600652217865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,2,power_law_1.2,0.14006400108337402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,2,power_law_1.2,0.1173632025718689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,1,balanced,0.07450133562088013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,1,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,2,power_law_1.2,0.021823999285697938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,2,power_law_1.2,0.10056320428848267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,2,power_law_1.2,0.147980797290802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,1,balanced,0.06917333106199901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.2,0.12961920499801635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,2,power_law_1.2,0.02428800016641617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,2,power_law_1.2,0.11088000535964966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,1,balanced,0.057904000083605446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,1,balanced,0.07595733304818471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.2,0.15749759674072267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,2,power_law_1.2,0.19662079811096192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,2,power_law_1.2,0.030502399802207945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,1,balanced,0.06831466654936473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,2,power_law_1.2,0.12237440347671509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.2,0.19089920520782472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,1,balanced,0.05621333420276642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,2,power_law_1.2,0.03141759932041168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,2,power_law_1.2,0.22878720760345458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,1,balanced,0.07721066474914551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.2,0.270579195022583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,2,power_law_1.2,0.027910399436950683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,1,balanced,0.06555200119813283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,2,power_law_1.2,0.16135040521621705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,1,balanced,0.05806399881839752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,2,power_law_1.2,0.3120575904846191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,2,power_law_1.2,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.2,0.2971199989318848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,1,balanced,0.07684800028800964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,2,power_law_1.2,0.17505279779434205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,1,balanced,0.07285866638024648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,1,balanced,0.060032000144322716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,2,power_law_1.2,0.03224320113658905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.2,0.4968255996704102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,2,power_law_1.2,0.3507904052734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,2,power_law_1.2,0.2626879930496216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,1,balanced,0.07830933233102162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,2,power_law_1.2,0.034220799803733826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.2,0.5869696140289307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,1,balanced,0.06025599936644236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,1,balanced,0.06898133456707001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,2,power_law_1.2,0.4960576057434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,2,power_law_1.2,0.2969919919967651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,2,power_law_1.2,0.03869439959526062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.2,0.8270144462585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,1,balanced,0.07863466441631317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,1,balanced,0.06003733476003011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,2,power_law_1.2,0.041228801012039185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,1,balanced,0.07250666618347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,2,power_law_1.2,0.39628798961639405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.2,1.0878335952758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,2,power_law_1.2,0.7086080074310303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,1,balanced,0.06161599854628245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,2,power_law_1.2,0.04248960018157959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,1,balanced,0.021189334491888683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,1,balanced,0.07820799946784973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,1,balanced,0.0692853331565857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,2,power_law_1.2,0.5408319950103759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,2,power_law_1.2,1.4864447593688965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,2,power_law_1.2,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,1,balanced,0.062133332093556724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,2,power_law_1.2,0.9745280265808105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,1,balanced,0.07854933540026347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.2,0.05986559987068176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,1,balanced,0.021189334491888683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,2,power_law_1.2,0.8374272346496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,1,balanced,0.06753066678841908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,2,power_law_1.2,2.2875328063964844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,1,balanced,0.06548266609509786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.2,0.08176000118255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,2,power_law_1.2,1.3210368156433105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,1,balanced,0.08105599880218506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,1,balanced,0.06518933176994324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,1,balanced,0.06874133149782817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,1,balanced,0.02124800036350886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,2,power_law_1.2,1.0894207954406738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.2,0.10227199792861938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,2,power_law_1.2,5.038387298583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,1,balanced,0.07023466626803081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.2,0.1362496018409729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,1,balanced,0.0809440016746521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,2,power_law_1.2,1.476204776763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,1,balanced,0.021194666624069214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,1,balanced,0.06782400111357371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,1,balanced,0.08105066418647766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,2,power_law_1.2,1.5368255615234374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.2,0.16510080099105834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,1,balanced,0.08506666620572408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,1,balanced,0.08214400211970012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,1,balanced,0.020960000654061634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,1,balanced,0.07234133283297221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,2,power_law_1.2,2.4999807357788084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.2,0.23634560108184816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,1,balanced,0.09010133147239685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.2,0.2998016119003296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,2,power_law_1.2,2.0447616577148438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,1,balanced,0.08806399504343669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,1,balanced,0.02107733239730199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,1,balanced,0.07028799752394359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,1,balanced,0.10571199655532837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.2,0.437113618850708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,2,power_law_1.2,4.623430252075195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,1,balanced,0.09337600072224934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.2,0.5742080211639404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,1,balanced,0.11571199695269267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,1,balanced,0.021327999730904896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,1,balanced,0.07643199960390727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,2,power_law_1.2,4.225279998779297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,2,power_law_1.2,0.7071807861328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,1,balanced,0.09666666388511658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,1,balanced,0.15040000279744467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,1,balanced,0.023397333920001984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,1,balanced,0.07597333192825317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,2,power_law_1.2,1.110086441040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,1,balanced,0.10341333349545796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,1,balanced,0.02346666653951009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,1,balanced,0.17702933152516684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,2,power_law_1.2,2.188345527648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,1,balanced,0.08060800035794576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,1,balanced,0.11552000045776367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,1,balanced,0.024874667326609295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,1,balanced,0.1060693363348643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,1,balanced,0.2405866583188375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,1,balanced,0.12988799810409546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,1,balanced,0.026159999271233875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,1,balanced,0.11309867103894551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,1,balanced,0.29875733455022174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,1,balanced,0.16538666685422262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,1,balanced,0.03164266546567281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,1,balanced,0.13693867127100626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,1,balanced,0.03169066707293192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,1,balanced,0.20401599009831747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,1,balanced,0.4238080183664958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,1,balanced,0.029653333127498627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,1,balanced,0.16576533516248068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,1,balanced,0.26362667481104535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,1,balanced,0.03379733363787333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,1,balanced,0.5351359844207764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,1,balanced,0.22426666816075644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,1,balanced,0.30430932839711505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,1,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,1,balanced,0.037808001041412354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,1,balanced,0.7822879950205485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,1,balanced,0.28228267033894855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,1,balanced,0.4228266477584839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,1,balanced,0.04560000201066335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,1,balanced,0.3954613208770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,1,balanced,0.5173066854476929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,1,balanced,0.047930667797724404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,1,balanced,1.014677365620931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,1,balanced,0.5088320175806681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,1,balanced,0.06842133402824402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,1,balanced,0.07787199815114339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,1,balanced,0.7521119912465414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,1,balanced,0.10180800159772237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,1,balanced,0.7355199654897054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,1,balanced,1.2602559725443523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,1,balanced,0.12777599692344666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,1,balanced,0.970250686009725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,1,balanced,0.1742080052693685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,1,balanced,0.9662933349609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,1,balanced,1.2081013520558674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,1,balanced,0.2128373384475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,1,balanced,1.9536906878153484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,1,balanced,1.1945013205210369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,1,balanced,0.30921600262324017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,1,balanced,1.874890645345052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,1,balanced,0.3962453206380208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,1,balanced,1.8855360349019368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,1,balanced,0.49136531352996826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,1,balanced,3.852938652038574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,1,balanced,3.721695899963379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,1,balanced,0.7629813353220621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,1,balanced,3.7466773986816406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,1,balanced,1.4878026644388835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,1,power_law_1.01,0.03638400137424469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,1,power_law_1.01,0.04289279878139496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,1,power_law_1.01,0.05398399829864502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,1,power_law_1.01,0.06888960003852844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,1,power_law_1.01,0.07519360184669495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,1,power_law_1.01,0.07649279832839966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,1,power_law_1.01,0.07676799893379212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,1,power_law_1.01,0.07933440208435058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,1,power_law_1.01,0.0822272002696991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,1,power_law_1.01,0.08158079981803894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,1,power_law_1.01,0.0815936028957367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,1,power_law_1.01,0.08631680011749268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,1,power_law_1.01,0.08671360015869141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,1,power_law_1.01,0.032979199290275575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,1,power_law_1.01,0.09106559753417968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,1,power_law_1.01,0.050444799661636355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,1,power_law_1.01,0.10103679895401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,1,power_law_1.01,0.01935359984636307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,1,power_law_1.01,0.10700800418853759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,1,power_law_1.01,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,1,power_law_1.01,0.053907197713851926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,1,power_law_1.01,0.01905920058488846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,1,power_law_1.01,0.12452479600906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,1,power_law_1.01,0.03708159923553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,1,power_law_1.01,0.05869439840316772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,1,power_law_1.01,0.018611200153827667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,1,power_law_1.01,0.1408192038536072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,1,power_law_1.01,0.07120640277862549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,1,power_law_1.01,0.054713600873947145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,1,power_law_1.01,0.019206400215625762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,1,power_law_1.01,0.14515199661254882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,1,power_law_1.01,0.01974399983882904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,1,power_law_1.01,0.06872959733009339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,1,power_law_1.01,0.056390398740768434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,1,power_law_1.01,0.020179200172424316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,1,power_law_1.01,0.05594239830970764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,1,power_law_1.01,0.18113919496536254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,1,power_law_1.01,0.06961280107498169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,1,power_law_1.01,0.020633600652217865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,1,power_law_1.01,0.061484801769256595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,1,power_law_1.01,0.20799999237060546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,1,power_law_1.01,0.022521600127220154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,1,power_law_1.01,0.07207040190696716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,1,power_law_1.01,0.059552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,1,power_law_1.01,0.022489599883556366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,1,power_law_1.01,0.26730880737304685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,1,power_law_1.01,0.06969599723815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,1,power_law_1.01,0.06302719712257385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,1,power_law_1.01,0.02316800057888031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,1,power_law_1.01,0.06856319904327393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,1,power_law_1.01,0.3087296009063721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,1,power_law_1.01,0.06275839805603027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,1,power_law_1.01,0.02507520020008087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,1,power_law_1.01,0.06846719980239868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,1,power_law_1.01,0.06306560039520263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,1,power_law_1.01,0.42851839065551756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,1,power_law_1.01,0.029728001356124877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,1,power_law_1.01,0.06649600267410279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,1,power_law_1.01,0.07125759720802308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,1,power_law_1.01,0.5462656021118164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,1,power_law_1.01,0.031097599864006044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,1,power_law_1.01,0.06775680184364319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,1,power_law_1.01,0.07294080257415772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,1,power_law_1.01,0.7651264190673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,1,power_law_1.01,0.028064000606536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,1,power_law_1.01,0.07453439831733703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,1,power_law_1.01,0.07419520020484924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,1,power_law_1.01,0.03201279938220978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,1,power_law_1.01,0.9991744041442872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,1,power_law_1.01,0.084307199716568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,1,power_law_1.01,0.03436160087585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,1,power_law_1.01,0.07904000282287597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,1,power_law_1.01,1.2124544143676759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,1,power_law_1.01,0.03669120073318481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,1,power_law_1.01,0.08661119937896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,1,power_law_1.01,0.08859519958496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,1,power_law_1.01,0.04426240026950836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,1,power_law_1.01,0.0982208013534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,1,power_law_1.01,1.905971145629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,1,power_law_1.01,0.08821759819984436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.01,0.04652799963951111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,1,power_law_1.01,0.11660799980163575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,1,power_law_1.01,0.09320319890975952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.01,0.066348797082901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,1,power_law_1.01,0.12665599584579468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,1,power_law_1.01,3.763391876220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,1,power_law_1.01,0.11004159450531006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.01,0.0764415979385376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,1,power_law_1.01,0.15527039766311646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.01,0.12220159769058228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.01,0.09964159727096558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,1,power_law_1.01,0.18165760040283202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.01,0.14965120553970337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.01,0.1250048041343689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,1,power_law_1.01,0.24647679328918456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.01,0.17667839527130128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.01,0.17114880084991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,1,power_law_1.01,0.30487039089202883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.01,0.23820159435272217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.01,0.21008000373840333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,1,power_law_1.01,0.4270336151123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.01,0.3064448118209839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.01,0.29806079864501955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,1,power_law_1.01,0.5502783775329589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.01,0.3935744047164917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.01,0.40616321563720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,1,power_law_1.01,0.7822656154632568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.01,0.4887296199798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.01,0.5214655876159668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,1,power_law_1.01,1.0248319625854492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,1,power_law_1.01,0.759603214263916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.01,0.7457280158996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,1,power_law_1.01,1.2575424194335938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,1,power_law_1.01,1.4840640068054198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.01,0.9796799659729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,1,power_law_1.01,1.970355224609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.01,1.2125951766967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,1,power_law_1.01,1.9002368927001954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,1,power_law_1.01,3.84832649230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,1,power_law_1.01,3.756089782714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,1,power_law_1.2,0.03598720133304596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,1,power_law_1.2,0.04575360119342804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,1,power_law_1.2,0.049881601333618165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,1,power_law_1.2,0.07310720086097718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,1,power_law_1.2,0.05077120065689087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,1,power_law_1.2,0.03294079899787903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,1,power_law_1.2,0.07425919771194459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,1,power_law_1.2,0.053388798236846925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,1,power_law_1.2,0.033497598767280576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,1,power_law_1.2,0.07455999851226806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,1,power_law_1.2,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,1,power_law_1.2,0.03625600039958954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,1,power_law_1.2,0.07639679908752442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,1,power_law_1.2,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,1,power_law_1.2,0.07697920203208923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,1,power_law_1.2,0.072326397895813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,1,power_law_1.2,0.08067200183868409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,1,power_law_1.2,0.05569919943809509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,1,power_law_1.2,0.07180160284042358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,1,power_law_1.2,0.08091520071029663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,1,power_law_1.2,0.056729602813720706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,1,power_law_1.2,0.07196800112724304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,1,power_law_1.2,0.08181120157241821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,1,power_law_1.2,0.0605184018611908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,1,power_law_1.2,0.07069439888000488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,1,power_law_1.2,0.05966079831123352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,1,power_law_1.2,0.08271359801292419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,1,power_law_1.2,0.06892160177230836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,1,power_law_1.2,0.06229119896888733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,1,power_law_1.2,0.019603200256824493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,1,power_law_1.2,0.08403840065002441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,1,power_law_1.2,0.06995840072631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,1,power_law_1.2,0.06307200193405152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,1,power_law_1.2,0.0682752013206482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,1,power_law_1.2,0.09121919870376587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,1,power_law_1.2,0.06239359974861145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,1,power_law_1.2,0.01905920058488846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,1,power_law_1.2,0.07191039919853211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,1,power_law_1.2,0.10277760028839111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,1,power_law_1.2,0.06712960004806519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,1,power_law_1.2,0.018937599658966065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,1,power_law_1.2,0.07336320281028748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,1,power_law_1.2,0.10492160320281982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,1,power_law_1.2,0.01988479942083359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,1,power_law_1.2,0.07696639895439147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,1,power_law_1.2,0.06931840181350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,1,power_law_1.2,0.12464640140533448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,1,power_law_1.2,0.07809919714927674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,1,power_law_1.2,0.01992959976196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,1,power_law_1.2,0.1383936047554016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,1,power_law_1.2,0.07215359807014465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,1,power_law_1.2,0.08412160277366638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,1,power_law_1.2,0.08195840120315552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,1,power_law_1.2,0.08710399866104127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,1,power_law_1.2,0.019993600249290467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,1,power_law_1.2,0.08606719970703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,1,power_law_1.2,0.14346879720687866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,1,power_law_1.2,0.09268479943275451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,1,power_law_1.2,0.09754239916801452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,1,power_law_1.2,0.02054399996995926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,1,power_law_1.2,0.10998400449752807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,1,power_law_1.2,0.11631360054016113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,1,power_law_1.2,0.1769215941429138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.2,0.12298239469528198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,1,power_law_1.2,0.022707200050354003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,1,power_law_1.2,0.12295039892196655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.2,0.15406080484390258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,1,power_law_1.2,0.15604480504989623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,1,power_law_1.2,0.20312960147857667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,1,power_law_1.2,0.022252799570560457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.2,0.17982720136642455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,1,power_law_1.2,0.18072320222854615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,1,power_law_1.2,0.2703423976898193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.2,0.23895680904388428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,1,power_law_1.2,0.023455999791622162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,1,power_law_1.2,0.24888319969177247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,1,power_law_1.2,0.29974400997161865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.2,0.2943552017211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,1,power_law_1.2,0.30858240127563474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,1,power_law_1.2,0.02489600032567978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,1,power_law_1.2,0.428326416015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,1,power_law_1.2,0.4295231819152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,1,power_law_1.2,0.030323201417922975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.2,0.40380158424377444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,1,power_law_1.2,0.030937600135803222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,1,power_law_1.2,0.5487360000610352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,1,power_law_1.2,0.5475264072418213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.2,0.5224127769470215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,1,power_law_1.2,0.028326401114463808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,1,power_law_1.2,0.7819839954376221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,1,power_law_1.2,0.7567232131958008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,1,power_law_1.2,0.03200640082359314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.2,0.7475071907043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,1,power_law_1.2,1.0246399879455566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,1,power_law_1.2,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,1,power_law_1.2,0.9997311592102051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,1,power_law_1.2,0.036723199486732486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.2,0.9738559722900391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,1,power_law_1.2,1.2593983650207519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,1,power_law_1.2,1.209830379486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,1,power_law_1.2,0.044019201397895814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.2,1.197542381286621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,1,power_law_1.2,1.9686527252197266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.2,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,1,power_law_1.2,1.9014272689819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.2,0.06611199975013733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,1,power_law_1.2,1.8963903427124023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,1,power_law_1.2,3.844940948486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.2,0.07538560032844543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,1,power_law_1.2,3.7553470611572264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.2,0.09946240186691284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,1,power_law_1.2,3.7531455993652343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.2,0.12491519451141357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.2,0.17187199592590333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.2,0.2099776029586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.2,0.30693759918212893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.2,0.3940608024597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.2,0.489683198928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,1,power_law_1.2,0.7611839771270752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,1,power_law_1.2,1.4841856002807616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,16,balanced,0.05188799897829691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,16,balanced,0.051925331354141235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,16,balanced,0.06888533135255177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,16,balanced,0.04577066500981649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,16,balanced,0.09750399986902873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,16,balanced,0.14919466773668924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,16,balanced,0.04677866895993551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,16,balanced,0.24690133333206177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,16,balanced,0.05049066742261251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,16,balanced,0.24697067340215048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,16,balanced,0.07262933254241943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,16,balanced,0.24817599852879843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,16,balanced,0.10076266527175903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,16,balanced,0.25065600872039795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,16,balanced,0.1493013302485148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,16,balanced,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,16,balanced,0.15165332953135172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,16,balanced,0.24977066119511923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,16,balanced,0.15266133348147073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,16,balanced,0.2518133322397868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,16,balanced,0.051738664507865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,16,balanced,0.15296000242233276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,16,balanced,0.2559093236923218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,16,balanced,0.05188799897829691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,16,balanced,0.1537920037905375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,16,balanced,0.25568532943725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,16,balanced,0.06365333497524261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,16,balanced,0.05020800232887268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,16,balanced,0.1558080017566681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,16,balanced,0.25332266092300415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,16,balanced,0.08692800005276997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,16,balanced,0.15702933073043823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,16,balanced,0.2606826623280843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,16,balanced,0.0425600012143453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,16,balanced,0.11534399787584941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,16,balanced,0.15836800138155618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,16,balanced,0.26233067115147907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,16,balanced,0.04423466821511587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,16,balanced,0.11779200037320454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,16,balanced,0.1625226636727651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,16,balanced,0.04786666731039683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,16,balanced,0.11804266770680745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,16,balanced,0.2657066583633423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,16,balanced,0.16673066218694052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,16,balanced,0.06012799839178721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,16,balanced,0.11898666620254517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,16,balanced,0.08313600222269694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,16,balanced,0.11885333061218262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,16,balanced,0.16935465733210245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,16,balanced,0.2711840073267619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,16,balanced,0.09457066655158997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,16,balanced,0.12142933408419292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,16,balanced,0.17486933867136636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,16,balanced,0.09202667077382405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,16,balanced,0.12204800049463908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,16,balanced,0.28011200825373334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,16,balanced,0.13344533244768778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,16,balanced,0.12126400073369344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,16,balanced,0.18652266263961792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,16,balanced,0.1281599998474121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,16,balanced,0.12556800246238708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,16,balanced,0.2940373420715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,16,balanced,0.1961066722869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,16,balanced,0.11430933078130086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,16,balanced,0.12922133008639017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,16,balanced,0.18635199467341104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,16,balanced,0.1300106644630432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,16,balanced,0.21614933013916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,16,balanced,0.1361066699028015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,16,balanced,0.31428800026575726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,16,balanced,0.1879840095837911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,16,balanced,0.14401066303253174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,16,balanced,0.1920213301976522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,16,balanced,0.24214933315912882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,16,balanced,0.18818666537602743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,16,balanced,0.3450346787770589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,16,balanced,0.1513866682847341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,16,balanced,0.18635733922322592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,16,balanced,0.1743626594543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,16,balanced,0.2804373304049174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,16,balanced,0.16566933194796243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,16,balanced,0.19646400213241577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,16,balanced,0.3657013177871704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,16,balanced,0.1845759948094686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,16,balanced,0.2189120054244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,16,balanced,0.320357342561086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,16,balanced,0.2542293270428975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,16,balanced,0.21224000056584677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,16,balanced,0.45207464694976807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,16,balanced,0.29928000768025714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,16,balanced,0.24595733483632407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,16,balanced,0.41185065110524494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,16,balanced,0.3710666497548421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,16,balanced,0.318725327650706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,16,balanced,0.5012213389078776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,16,balanced,0.45261867841084796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,16,balanced,0.4859946568806966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,16,balanced,0.6071413358052572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,16,balanced,0.3797173500061035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,16,balanced,0.66102401415507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,16,balanced,0.768618663152059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,16,balanced,0.6783146858215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,16,balanced,0.5207839806874593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,16,balanced,1.1808853149414062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,16,balanced,0.7543573379516602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,16,balanced,0.832810640335083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,16,balanced,0.6571199893951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,16,power_law_1.01,0.1017024040222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,16,balanced,1.4974239667256672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,16,power_law_1.01,0.08951039910316468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,16,power_law_1.01,0.1408959984779358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,16,power_law_1.01,0.1509760022163391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,16,power_law_1.01,0.07562879920005798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,16,power_law_1.01,0.12346240282058715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,16,power_law_1.01,0.24074239730834962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,16,balanced,1.0027039845784504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,16,power_law_1.01,0.06875519752502442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,16,power_law_1.01,0.08206719756126404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,16,power_law_1.01,0.10589439868927002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,16,power_law_1.01,0.07048320174217224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,16,power_law_1.01,0.09127680063247681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,16,power_law_1.01,0.11731840372085571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,16,power_law_1.01,0.08197759985923767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,16,power_law_1.01,0.11227519512176513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,16,power_law_1.01,0.09006720185279846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,16,power_law_1.01,0.14235520362854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,16,power_law_1.01,0.13219840526580812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,16,power_law_1.01,0.10016640424728393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,16,power_law_1.01,0.16652159690856932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,16,power_law_1.01,0.13215999603271483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,16,power_law_1.01,0.10777599811553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,16,balanced,1.0607679684956868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,16,power_law_1.01,0.2060096025466919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,16,power_law_1.01,0.05070719718933105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,16,power_law_1.01,0.10501760244369507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,16,power_law_1.01,0.13692159652709962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,16,balanced,0.7970080375671387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,16,power_law_1.01,0.21760001182556152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,16,power_law_1.01,0.10860799551010132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,16,power_law_1.01,0.14227839708328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,16,power_law_1.01,0.21127679347991943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,16,power_law_1.01,0.07601280212402343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,16,power_law_1.01,0.11418240070343018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,16,power_law_1.01,0.1377408027648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,16,power_law_1.01,0.22494080066680908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,16,power_law_1.01,0.11103359460830689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,16,power_law_1.01,0.06999040246009827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,16,power_law_1.01,0.14304640293121337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,16,power_law_1.01,0.11381759643554687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,16,power_law_1.01,0.22351999282836915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,16,power_law_1.01,0.04632959961891174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,16,power_law_1.01,0.14635519981384276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,16,power_law_1.01,0.11638400554656983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,16,power_law_1.01,0.2215424060821533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,16,power_law_1.01,0.05616639852523804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,16,power_law_1.01,0.12229759693145752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,16,balanced,1.9082506497701008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,16,power_law_1.01,0.15117440223693848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,16,power_law_1.01,0.2357759952545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,16,power_law_1.01,0.1275007963180542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,16,power_law_1.01,0.060812801122665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,16,power_law_1.01,0.15620479583740235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,16,power_law_1.01,0.13228800296783447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,16,power_law_1.01,0.23081600666046143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,16,power_law_1.01,0.07699840068817139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,16,power_law_1.01,0.1554751992225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,16,power_law_1.01,0.1506495952606201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,16,power_law_1.01,0.24318079948425292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,16,power_law_1.01,0.07581440210342408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.01,0.16100480556488037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,16,power_law_1.01,0.1689792037010193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,16,power_law_1.01,0.25137920379638673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.01,0.1892032027244568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,16,power_law_1.01,0.10586880445480347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,16,power_law_1.01,0.20628480911254882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.01,0.19990400075912476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,16,power_law_1.01,0.2738944053649902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,16,power_law_1.01,0.10763520002365112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,16,power_law_1.01,0.20952320098876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.01,0.26184959411621095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,16,power_law_1.01,0.30732800960540774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,16,power_law_1.01,0.10101120471954346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,16,power_law_1.01,0.23991041183471679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.01,0.3035264015197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,16,power_law_1.01,0.31521921157836913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,16,power_law_1.01,0.1509503960609436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.01,0.42969598770141604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,16,power_law_1.01,0.27394559383392336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,16,power_law_1.01,0.15049599409103392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,16,power_law_1.01,0.36611840724945066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.01,0.5153279781341553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,16,power_law_1.01,0.3399807929992676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.01,0.7028736114501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,16,power_law_1.01,0.15834239721298218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,16,balanced,1.363711992899577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,16,power_law_1.01,0.3658240079879761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,16,power_law_1.01,0.41344637870788575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.01,0.9451519966125488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,16,power_law_1.01,0.1490880012512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,16,power_law_1.01,0.4402944087982178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,16,power_law_1.01,0.5492735862731933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,16,power_law_1.01,1.0473024368286132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,16,power_law_1.01,0.15213439464569092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,16,power_law_1.01,0.48732800483703614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,16,power_law_1.01,0.661516809463501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,16,power_law_1.01,0.1549504041671753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,16,power_law_1.01,1.8351680755615234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,16,power_law_1.01,0.6202047824859619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,16,power_law_1.01,0.9350527763366699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,16,power_law_1.01,0.1498047947883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,16,power_law_1.01,0.8216768264770508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,16,balanced,1.5486027399698894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,16,power_law_1.01,3.4973888397216797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,16,balanced,1.2062933444976807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,16,power_law_1.01,1.1632960319519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.01,0.1492416024208069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,16,power_law_1.01,1.0373760223388673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.01,0.17727359533309936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,16,power_law_1.01,1.4622464179992676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.01,0.18787839412689208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,16,power_law_1.01,1.2957568168640137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,16,power_law_1.01,2.114361572265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.01,0.2274240016937256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,16,power_law_1.01,1.6818431854248046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.01,0.282585597038269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,16,balanced,2.946437199910482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,16,power_law_1.01,3.9989185333251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,16,power_law_1.01,2.27825927734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.01,0.37575678825378417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.01,0.4694975852966309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,16,power_law_1.01,4.48919677734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.01,0.6716479778289794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.01,0.8755776405334472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,16,power_law_1.01,1.0000255584716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,16,power_law_1.01,1.7720512390136718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,16,power_law_1.01,3.3605377197265627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,16,balanced,2.6569066047668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,16,balanced,2.307450612386068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,16,balanced,3.025578816731771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,16,balanced,5.846794764200847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,16,power_law_1.2,0.14398720264434814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,16,power_law_1.2,0.23296639919281006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,16,power_law_1.2,0.08588160276412964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,16,power_law_1.2,0.10758399963378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,16,power_law_1.2,0.14405759572982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,16,power_law_1.2,0.15860480070114136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,16,power_law_1.2,0.18499840497970582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,16,power_law_1.2,0.20178558826446533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,16,power_law_1.2,0.22359039783477783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,16,power_law_1.2,0.22398080825805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,16,power_law_1.2,0.21662719249725343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,16,power_law_1.2,0.24439680576324463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,16,power_law_1.2,0.2271872043609619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,16,power_law_1.2,0.25333759784698484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,16,power_law_1.2,0.26191999912261965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,16,power_law_1.2,0.26056320667266847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,16,power_law_1.2,0.28695681095123293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,16,power_law_1.2,0.3256959915161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,16,power_law_1.2,0.3284991979598999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,16,power_law_1.2,0.10076800584793091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,16,power_law_1.2,0.3794816017150879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,16,power_law_1.2,0.0899839997291565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,16,power_law_1.2,0.39296000003814696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,16,power_law_1.2,0.13915519714355468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,16,power_law_1.2,0.49370241165161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,16,power_law_1.2,0.11641600131988525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,16,power_law_1.2,0.578988790512085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,16,power_law_1.2,0.06063359975814819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,16,power_law_1.2,0.058976000547409056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,16,power_law_1.2,0.07768959999084472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,16,power_law_1.2,0.7593599796295166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,16,power_law_1.2,0.07401599884033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,16,power_law_1.2,0.08947839736938476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,16,power_law_1.2,0.07998719811439514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,16,power_law_1.2,0.9491392135620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,16,power_law_1.2,0.08581119775772095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,16,power_law_1.2,0.10195839405059814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,16,power_law_1.2,1.4395456314086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,16,power_law_1.2,0.10243200063705445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,16,power_law_1.2,0.13104000091552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,16,power_law_1.2,1.572704029083252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,16,power_law_1.2,0.10692479610443115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,16,power_law_1.2,0.05021439790725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,16,power_law_1.2,0.13231359720230101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,16,power_law_1.2,0.10341119766235352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,16,power_law_1.2,0.12755839824676513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,16,power_law_1.2,2.0599039077758787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,16,power_law_1.2,0.10802559852600098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,16,power_law_1.2,0.13194880485534669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,16,power_law_1.2,0.07483519911766053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,16,power_law_1.2,0.11336319446563721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,16,power_law_1.2,0.1376512050628662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,16,power_law_1.2,0.0418368011713028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,16,power_law_1.2,2.9953472137451174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,16,power_law_1.2,0.11463040113449097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,16,power_law_1.2,0.14685440063476562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,16,power_law_1.2,0.04385280013084412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,16,power_law_1.2,0.11706240177154541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,16,power_law_1.2,0.14350719451904298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,16,power_law_1.2,0.049472001194953916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,16,power_law_1.2,0.11747200489044189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,16,power_law_1.2,0.15611519813537597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,16,power_law_1.2,7.105171203613281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,16,power_law_1.2,0.06039040088653565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,16,power_law_1.2,0.12375040054321289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,16,power_law_1.2,0.1617535948753357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,16,power_law_1.2,0.12910720109939575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,16,power_law_1.2,0.07773439884185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,16,power_law_1.2,0.16895999908447265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,16,power_law_1.2,0.1420415997505188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,16,power_law_1.2,0.07270399928092956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,16,power_law_1.2,0.18773759603500367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,16,power_law_1.2,0.15134079456329347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,16,power_law_1.2,0.08748160004615783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,16,power_law_1.2,0.2111743927001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.2,0.16260479688644408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,16,power_law_1.2,0.09748479723930359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,16,power_law_1.2,0.21577599048614501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.2,0.19973119497299194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,16,power_law_1.2,0.0883840024471283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.2,0.22729599475860596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,16,power_law_1.2,0.27100160121917727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,16,power_law_1.2,0.15415680408477783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.2,0.27450881004333494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,16,power_law_1.2,0.14883840084075928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,16,power_law_1.2,0.2901887893676758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,16,power_law_1.2,0.15827840566635132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.2,0.34186880588531493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,16,power_law_1.2,0.35117440223693847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,16,power_law_1.2,0.15356800556182862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.2,0.4384255886077881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,16,power_law_1.2,0.43765759468078613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,16,power_law_1.2,0.15093120336532592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,8,balanced,0.04379733403523763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.2,0.6067391872406006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,16,power_law_1.2,0.15589760541915892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,16,power_law_1.2,0.6374015808105469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,8,balanced,0.047007997830708824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,16,power_law_1.2,0.15085439682006835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,8,balanced,0.04531733194986979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.2,0.8115519523620606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,16,power_law_1.2,0.7211071968078613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,8,balanced,0.049413333336512245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.2,0.15578240156173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,8,balanced,0.05382933219273885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,8,balanced,0.07216000060240428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,16,power_law_1.2,1.0033535957336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.2,1.241875171661377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,8,balanced,0.10381333033243816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,8,balanced,0.05615466833114624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.2,0.16030720472335816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,8,balanced,0.06850666801134746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,8,balanced,0.15052800377209982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,16,power_law_1.2,1.368563175201416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.2,0.197817599773407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,16,power_law_1.2,1.3333888053894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,8,balanced,0.15176000197728476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,8,balanced,0.09760533769925435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,8,balanced,0.025626666843891144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,8,balanced,0.15422933300336203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,8,balanced,0.0627040018637975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.2,0.25401599407196046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,8,balanced,0.1564959983030955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,16,power_law_1.2,1.9567680358886719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,16,power_law_1.2,2.0761728286743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,8,balanced,0.14871467153231302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.2,0.30650880336761477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,8,balanced,0.026698666314284008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,8,balanced,0.15685866276423135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,8,balanced,0.07045866549015045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,8,balanced,0.2481706738471985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,8,balanced,0.029520000020662945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,8,balanced,0.1581493318080902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.2,0.40523519515991213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,16,power_law_1.2,2.4149375915527345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,8,balanced,0.04450666904449463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,8,balanced,0.2477440039316813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,8,balanced,0.15864533185958862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,16,power_law_1.2,4.1052734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,8,balanced,0.05681600173314413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.2,0.5579648017883301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,8,balanced,0.09109333157539368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,8,balanced,0.24877333641052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,8,balanced,0.16007999579111734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,8,balanced,0.08297599852085114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,8,balanced,0.2502346634864807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,8,balanced,0.16378666957219443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.2,0.7365824222564697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,8,balanced,0.13884799679120383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,8,balanced,0.09075199564297994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,16,power_law_1.2,5.208038330078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,8,balanced,0.2508959968884786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,8,balanced,0.17126399278640747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,8,balanced,0.08970133463541667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,8,balanced,0.19349332650502524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,8,balanced,0.09621333082516988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,8,balanced,0.2508959968884786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.2,0.9307904243469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,8,balanced,0.17357865969340006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,8,balanced,0.09900266925493877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,8,balanced,0.20200000206629434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,8,balanced,0.17871999740600586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,8,balanced,0.09683733185132344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,8,balanced,0.19357866048812866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,16,power_law_1.2,1.269388771057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,8,balanced,0.252128005027771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,8,balanced,0.10098133484522502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,8,balanced,0.1909653345743815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,8,balanced,0.18766933679580688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,8,balanced,0.1065066655476888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,8,balanced,0.2530826727549235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,16,power_law_1.2,2.035622406005859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,8,balanced,0.20047465960184732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,8,balanced,0.11756267150243123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,8,balanced,0.1931040088335673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,8,balanced,0.12825066844622293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,8,balanced,0.25507734219233197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,8,balanced,0.18227734168370566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,8,balanced,0.13983466227849325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,8,balanced,0.2287786602973938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,8,balanced,0.26293333371480304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,16,power_law_1.2,3.985286331176758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,8,balanced,0.18571199973424277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,8,balanced,0.14358400305112204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,8,balanced,0.2613760034243266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,8,balanced,0.18212266763051352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,8,balanced,0.15731733043988547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,8,balanced,0.2592906753222148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,8,balanced,0.2686240077018738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,8,balanced,0.17971734205881754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,8,balanced,0.20950400829315186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,8,balanced,0.18597332636515299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,8,balanced,0.27874133984247845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,8,balanced,0.19991467396418253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,8,balanced,0.30639467636744183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,8,balanced,0.2888053258260091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,8,balanced,0.229802668094635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,8,balanced,0.19089599450429282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,8,balanced,0.2534986734390259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,8,balanced,0.3561333417892456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,8,balanced,0.3089279929796855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,8,balanced,0.20838399728139242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,8,balanced,0.2064746618270874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,8,balanced,0.30404800176620483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,8,balanced,0.3314879933993022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,8,balanced,0.4686880111694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,8,balanced,0.22256000836690268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,8,balanced,0.36291734377543133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,8,balanced,0.35520001252492267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,8,balanced,0.23598933219909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,8,balanced,0.3959306478500366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,8,balanced,0.26797332366307575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,8,balanced,0.5657493273417155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,8,balanced,0.44014398256937665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,8,balanced,0.3074880043665568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,8,balanced,0.48765333493550617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,8,balanced,0.5368426640828451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,8,balanced,0.4607253472010295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,8,balanced,0.7914079825083414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,8,balanced,0.549402674039205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,8,balanced,0.5196426709493002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,8,balanced,0.8424800237019857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,8,balanced,0.9849599997202555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,8,balanced,0.7659040292104086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,8,balanced,0.7311733563741049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,8,balanced,1.0413333574930828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,8,balanced,1.2387680212656658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,8,balanced,0.8611733118693033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,8,balanced,0.9431573549906412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,8,balanced,1.3382773399353027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,8,balanced,1.168789307276408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,8,balanced,1.1523573398590088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,8,balanced,1.863152027130127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,8,balanced,2.0289653142293296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,8,power_law_1.01,0.09912959933280945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,8,power_law_1.01,0.14534399509429932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,8,power_law_1.01,0.10325759649276733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,8,balanced,1.5789225896199544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,8,power_law_1.01,0.09806720018386841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,8,balanced,1.781376043955485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,8,power_law_1.01,0.12923519611358641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,8,power_law_1.01,0.15467519760131837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,8,power_law_1.01,0.19544320106506347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,8,power_law_1.01,0.1946176052093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,8,power_law_1.01,0.2140415906906128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,8,power_law_1.01,0.2130944013595581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,8,power_law_1.01,0.2064448118209839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,8,power_law_1.01,0.2280639886856079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,8,power_law_1.01,0.22279679775238037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,8,power_law_1.01,0.23064959049224854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,8,power_law_1.01,0.23262081146240235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,8,power_law_1.01,0.2524224042892456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,8,power_law_1.01,0.27121920585632325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,8,power_law_1.01,0.2906176090240479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,8,power_law_1.01,0.2912512063980103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,8,power_law_1.01,0.3220031976699829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,8,power_law_1.01,0.3452415943145752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,8,power_law_1.01,0.40812158584594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,8,power_law_1.01,0.48631677627563474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,8,power_law_1.01,0.6344511985778809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,8,power_law_1.01,0.06895359754562377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,8,power_law_1.01,0.7635456085205078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,8,power_law_1.01,1.0067904472351075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,8,power_law_1.01,0.09447680115699768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,8,power_law_1.01,0.09429759979248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,8,power_law_1.01,0.03118720054626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,8,power_law_1.01,0.07758079767227173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,8,power_law_1.01,1.1868736267089843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,8,power_law_1.01,0.06956160068511963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,8,power_law_1.01,0.13125120401382445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,8,power_law_1.01,1.6055168151855468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,8,power_law_1.01,0.04979200065135956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,8,power_law_1.01,0.08583040237426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,8,power_law_1.01,0.08413439989089966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,8,power_law_1.01,0.0345984011888504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,8,power_law_1.01,0.11152640581130982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,8,power_law_1.01,2.263859176635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,8,power_law_1.01,0.038387200236320494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,8,power_law_1.01,0.09729920029640197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,8,power_law_1.01,0.12681599855422973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,8,power_law_1.01,0.12925440073013306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,8,power_law_1.01,0.04590719938278198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,8,power_law_1.01,0.1143231987953186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,8,balanced,3.628608067830404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,8,power_law_1.01,0.13203840255737304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,8,power_law_1.01,4.185401535034179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,8,power_law_1.01,0.12806400060653686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,8,power_law_1.01,0.05829120278358459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,8,power_law_1.01,0.13119360208511352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,8,power_law_1.01,0.16568959951400758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,8,power_law_1.01,0.069651198387146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,8,power_law_1.01,0.1407487988471985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,8,power_law_1.01,0.07156479954719544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,8,power_law_1.01,0.15935360193252562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,8,power_law_1.01,0.14320000410079955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,8,power_law_1.01,0.0748799979686737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,8,power_law_1.01,0.1683519959449768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,8,power_law_1.01,0.14600319862365724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,8,power_law_1.01,0.07779840230941773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,8,power_law_1.01,0.1744320034980774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,8,power_law_1.01,0.14975359439849853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,8,power_law_1.01,0.15850239992141724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,8,power_law_1.01,0.07710080146789551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,8,power_law_1.01,0.1803455948829651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,8,balanced,4.019381205240886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,8,power_law_1.01,0.16168320178985596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,8,power_law_1.01,0.08021119832992554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,8,power_law_1.01,0.17706880569458008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,8,power_law_1.01,0.173689603805542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,8,power_law_1.01,0.08093439936637878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,8,power_law_1.01,0.1788032054901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,8,power_law_1.01,0.19567359685897828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,8,power_law_1.01,0.17719680070877075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,8,power_law_1.01,0.08723840117454529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,8,power_law_1.01,0.2074944019317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,8,power_law_1.01,0.19360640048980712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,8,power_law_1.01,0.0927616000175476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,8,power_law_1.01,0.2527424097061157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,8,power_law_1.01,0.1897663950920105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,8,power_law_1.01,0.10134400129318237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,8,power_law_1.01,0.2939071893692017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,8,power_law_1.01,0.1984768033027649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,8,power_law_1.01,0.09634559750556945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,8,power_law_1.01,0.10888320207595825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,8,power_law_1.01,0.21765120029449464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,8,power_law_1.01,0.36945281028747556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,8,balanced,3.0959946314493814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.01,0.1588863968849182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,8,balanced,3.4928105672200522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.01,0.2261631965637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,8,power_law_1.01,0.40371198654174806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.01,0.16355199813842775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.01,0.2604991912841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,8,power_law_1.01,0.5737088203430176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.01,0.17834880352020263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.01,0.2958208084106445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,8,power_law_1.01,0.7098495960235596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.01,0.3586751937866211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,8,power_law_1.01,0.9987456321716308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.01,0.2021631956100464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.01,0.43301119804382326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,8,power_law_1.01,1.1866368293762206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.01,0.23719680309295654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,8,power_law_1.01,1.512275218963623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.01,0.5663680076599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.01,0.31312639713287355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,8,power_law_1.01,2.396121597290039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.01,0.7128767967224121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.01,0.38398079872131347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,8,power_law_1.01,4.359302520751953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.01,0.9567168235778809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.01,0.5314752101898194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.01,1.2472960472106933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.01,0.7142271995544434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,8,power_law_1.01,0.8785728454589844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,8,power_law_1.01,1.5590784072875976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,8,power_law_1.01,2.399692726135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,8,power_law_1.01,1.4250432014465333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,8,power_law_1.01,4.6099201202392575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,8,power_law_1.01,2.790995216369629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,8,power_law_1.2,0.06770560145378113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,8,power_law_1.2,0.09884160161018371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,8,power_law_1.2,0.0939136028289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,8,power_law_1.2,0.1396607995033264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,8,power_law_1.2,0.055801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,8,power_law_1.2,0.06472319960594178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,8,power_law_1.2,0.0683456003665924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,8,power_law_1.2,0.09472000002861022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,8,power_law_1.2,0.08785279989242553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,8,power_law_1.2,0.1171455979347229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,8,power_law_1.2,0.10076800584793091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,8,power_law_1.2,0.1451200008392334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,8,power_law_1.2,0.12554240226745605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,8,power_law_1.2,0.1923967957496643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,8,power_law_1.2,0.12876800298690796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,8,power_law_1.2,0.19957760572433472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,8,power_law_1.2,0.129312002658844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,8,power_law_1.2,0.21281919479370118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,8,power_law_1.2,0.13325439691543578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,8,power_law_1.2,0.1363584041595459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,8,power_law_1.2,0.2139456033706665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,8,power_law_1.2,0.1395967960357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,8,power_law_1.2,0.21543679237365723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,8,power_law_1.2,0.1486207962036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,8,power_law_1.2,0.21595520973205568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,8,power_law_1.2,0.09868800044059753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,8,power_law_1.2,0.152729594707489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,8,power_law_1.2,0.2225343942642212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,8,power_law_1.2,0.031007999181747438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,8,power_law_1.2,0.1649216055870056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,8,power_law_1.2,0.23393280506134034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,8,power_law_1.2,0.1349503993988037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,8,power_law_1.2,0.16881920099258424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,8,power_law_1.2,0.24618239402770997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,8,power_law_1.2,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,8,power_law_1.2,0.18584320545196534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,8,power_law_1.2,0.24992640018463136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,8,power_law_1.2,0.0970304012298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,8,power_law_1.2,0.044614401459693906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,8,power_law_1.2,0.20952959060668946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,4,balanced,0.04785599807898203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,8,power_law_1.2,0.27153921127319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,8,power_law_1.2,0.11085439920425415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,8,power_law_1.2,0.21891839504241944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,8,power_law_1.2,0.02890239953994751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,8,power_law_1.2,0.27562880516052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,8,power_law_1.2,0.13378560543060303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,8,power_law_1.2,0.3009727954864502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,4,balanced,0.051776001850763954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,4,balanced,0.06853866577148438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,8,power_law_1.2,0.3203520059585571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,8,power_law_1.2,0.16536320447921754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,8,power_law_1.2,0.3232320070266724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,8,power_law_1.2,0.03893119990825653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,4,balanced,0.09912000099817912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,8,power_law_1.2,0.3890048027038574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,8,power_law_1.2,0.3639872074127197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,8,power_law_1.2,0.17350399494171143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,4,balanced,0.15074666341145834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,8,power_law_1.2,0.04609279930591583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,8,power_law_1.2,0.46158719062805176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,4,balanced,0.24569600820541382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,8,power_law_1.2,0.3878079891204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,8,power_law_1.2,0.17152639627456664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,8,power_law_1.2,0.05400959849357605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,4,balanced,0.24702932437260947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,8,power_law_1.2,0.617471981048584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,8,power_law_1.2,0.1721791982650757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,4,balanced,0.2485333283742269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,8,power_law_1.2,0.4576000213623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,8,power_law_1.2,0.06764159798622131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,8,power_law_1.2,0.759168004989624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,8,power_law_1.2,0.18345600366592407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,4,balanced,0.2523360053698222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,8,power_law_1.2,0.07039999961853027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,8,power_law_1.2,0.5140160083770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,4,balanced,0.2494986653327942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,8,power_law_1.2,1.034489631652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,8,power_law_1.2,0.17319680452346803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,8,power_law_1.2,0.07680000066757202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,8,power_law_1.2,0.6563968181610107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,4,balanced,0.2542240023612976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,8,power_law_1.2,1.6256895065307617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,8,power_law_1.2,0.18277759552001954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,4,balanced,0.25413866837819415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,8,power_law_1.2,0.0755840003490448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,8,power_law_1.2,0.8566847801208496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,4,balanced,0.25965867439905804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,8,power_law_1.2,0.18785279989242554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,8,power_law_1.2,0.07994880080223084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,8,power_law_1.2,1.7107711791992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,8,power_law_1.2,1.1696127891540526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,4,balanced,0.2608480056126912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,4,balanced,0.04533866544564565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,8,power_law_1.2,0.19390079975128174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,8,power_law_1.2,0.07931519746780395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,4,balanced,0.26848000288009644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,8,power_law_1.2,2.704243278503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,8,power_law_1.2,0.19765119552612304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,8,power_law_1.2,0.0849407970905304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,8,power_law_1.2,1.478329563140869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,4,balanced,0.046021332343419395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,4,balanced,0.2688266634941101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,8,power_law_1.2,0.20476160049438477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,4,balanced,0.050954664746920265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,8,power_law_1.2,0.08726400136947632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,4,balanced,0.2760960062344869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,4,balanced,0.07416533430417378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,8,power_law_1.2,1.646291160583496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,8,power_law_1.2,0.2303999900817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,8,power_law_1.2,5.029248046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,8,power_law_1.2,0.09253759980201721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,4,balanced,0.10679466525713603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.2,0.24031360149383546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,4,balanced,0.29039466381073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,4,balanced,0.1536799967288971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,8,power_law_1.2,0.10092799663543701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,8,power_law_1.2,2.556831932067871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.2,0.28910720348358154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,4,balanced,0.15797332922617593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,8,power_law_1.2,0.10069119930267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,4,balanced,0.3054453333218892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,4,balanced,0.16086933016777039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.2,0.3173376083374023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,8,power_law_1.2,0.11308799982070923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,4,balanced,0.16127999623616537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.2,0.3910207986831665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.2,0.14790400266647338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,8,power_law_1.2,5.712678527832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,4,balanced,0.16241066654523215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,4,balanced,0.33242666721343994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.2,0.46810240745544435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,4,balanced,0.16690667470296225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.2,0.16531840562820435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,4,balanced,0.023584000766277313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,4,balanced,0.17004799842834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.2,0.6386240005493165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.2,0.18021759986877442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,4,balanced,0.0521066685517629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,4,balanced,0.17072000106175741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,4,balanced,0.36135466893513996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.2,0.8113792419433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.2,0.2138688087463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,4,balanced,0.024864000578721363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,4,balanced,0.17704532543818155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,4,balanced,0.05417599777380625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.2,0.2637120008468628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,4,balanced,0.18381865819295248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.2,1.058176040649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,4,balanced,0.02716800073782603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,4,balanced,0.057589332262674965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,4,balanced,0.18650666872660318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.2,0.34152319431304934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,4,balanced,0.07027199864387512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,4,balanced,0.04258666435877482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.2,1.3210623741149903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,4,balanced,0.40753066539764404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,4,balanced,0.1013813316822052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,4,balanced,0.05657066901524862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.2,0.4468736171722412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,4,balanced,0.19409066438674927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,4,balanced,0.08306133250395457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,4,balanced,0.14055466651916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,8,power_law_1.2,1.7814975738525392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.2,0.6070528030395508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,4,balanced,0.08524800340334575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,4,balanced,0.14446933070818582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,4,balanced,0.21338667472203574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,4,balanced,0.0906986693541209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,4,balanced,0.14569600423177084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.2,0.7673215866088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,8,power_law_1.2,2.5926719665527345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,4,balanced,0.4691306749979655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,4,balanced,0.09108799695968628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,4,balanced,0.14670933286348978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,4,balanced,0.22934933503468832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,8,power_law_1.2,0.9395839691162109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,4,balanced,0.09299733241399129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,4,balanced,0.14947199821472168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,4,balanced,0.0963200032711029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,8,power_law_1.2,5.661113739013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,4,balanced,0.1518826683362325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,8,power_law_1.2,1.467801570892334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,4,balanced,0.261845330397288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,4,balanced,0.10016000270843506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,4,balanced,0.1536853313446045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,4,balanced,0.10272000233332317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,8,power_law_1.2,2.9236352920532225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,4,balanced,0.5957866509755453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,4,balanced,0.10880000392595927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,4,balanced,0.15357333421707153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,4,balanced,0.29335467020670575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,4,balanced,0.13392000397046408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,4,balanced,0.15572800238927206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,4,balanced,0.13474667072296143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,4,balanced,0.16074132919311523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,4,balanced,0.1390720009803772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,4,balanced,0.3686986764272054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,4,balanced,0.16320000092188516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,4,balanced,0.16850133736928305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,4,balanced,0.21580799420674643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,4,balanced,0.18491200606028238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,4,balanced,0.6879626909891764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,4,balanced,0.24677334229151407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,4,balanced,0.43700265884399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,4,balanced,0.1958613395690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,4,balanced,0.2320853273073832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,4,balanced,0.225983997186025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,4,balanced,0.24602667490641275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,4,balanced,0.5913813511530558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,4,balanced,0.2539199988047282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,4,balanced,0.28142933050791424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,4,balanced,0.9501012961069742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,4,balanced,0.30818132559458417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,4,balanced,0.7208373546600342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,4,balanced,0.36130134264628094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,4,balanced,0.31753599643707275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,4,balanced,0.5091573397318522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,4,balanced,0.38655467828114826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,4,balanced,1.1757653554280598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,4,balanced,1.02238933245341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,4,balanced,0.6243413289388021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,4,balanced,0.4503200054168701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,4,balanced,1.2912533283233643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,4,balanced,0.7139200369517008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,4,balanced,1.472117265065511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,4,balanced,0.8953279654184977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,4,balanced,1.162549336751302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,4,balanced,0.8543626467386881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,4,balanced,1.6113866170247395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,4,balanced,2.177776018778483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,4,balanced,1.1261066595713298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,4,balanced,1.4442133903503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,4,balanced,2.467696030934652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,4,balanced,1.6722346941630046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,4,balanced,2.1954347292582193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,4,power_law_1.01,0.069651198387146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,4,power_law_1.01,0.050291198492050174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,4,power_law_1.01,0.09253119826316833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,4,power_law_1.01,0.07573760151863099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,4,power_law_1.01,0.0694208025932312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,4,power_law_1.01,0.08860160112380981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,4,power_law_1.01,0.05878400206565857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,4,power_law_1.01,0.1226304054260254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,4,power_law_1.01,0.0680512011051178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,4,power_law_1.01,0.15185920000076295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,4,power_law_1.01,0.09285119771957398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,4,balanced,4.222960154215495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,4,power_law_1.01,0.19493759870529176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,4,power_law_1.01,0.10728319883346557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,4,power_law_1.01,0.20497920513153076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,4,power_law_1.01,0.12935680150985718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,4,power_law_1.01,0.20730879306793212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,4,power_law_1.01,0.13756799697875977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,4,power_law_1.01,0.05692160129547119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,4,power_law_1.01,0.20755200386047362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,4,power_law_1.01,0.13448959589004517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,4,power_law_1.01,0.2161087989807129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,4,power_law_1.01,0.0652288019657135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,4,power_law_1.01,0.14176000356674195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,4,power_law_1.01,0.21458559036254882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,4,power_law_1.01,0.059910398721694944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,4,power_law_1.01,0.1506495952606201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,4,power_law_1.01,0.22641279697418212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,4,power_law_1.01,0.06985599994659424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,4,power_law_1.01,0.15070079565048217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,4,power_law_1.01,0.22471680641174316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,4,power_law_1.01,0.09082239866256714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,4,power_law_1.01,0.15915520191192628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,4,power_law_1.01,0.2403264045715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,4,power_law_1.01,0.09950079917907714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,4,power_law_1.01,0.16051199436187744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,4,power_law_1.01,0.12054400444030762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,4,power_law_1.01,0.24277119636535643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,4,power_law_1.01,0.16710400581359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,4,power_law_1.01,0.12313599586486816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,4,power_law_1.01,0.25350399017333985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,4,power_law_1.01,0.02250880002975464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,4,balanced,4.836906750996907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,4,power_law_1.01,0.17375359535217286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,4,power_law_1.01,0.12447999715805054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,4,power_law_1.01,0.2969856023788452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,4,power_law_1.01,0.02858240008354187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,4,power_law_1.01,0.12935680150985718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,4,power_law_1.01,0.2978816032409668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,4,power_law_1.01,0.19094400405883788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,4,power_law_1.01,0.02890239953994751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,4,power_law_1.01,0.1304703950881958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,4,power_law_1.01,0.3531519889831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,4,power_law_1.01,0.03521920144557953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,4,power_law_1.01,0.2188096046447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,4,power_law_1.01,0.13751039505004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,4,power_law_1.01,0.04687359929084778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,4,power_law_1.01,0.3868544101715088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,4,power_law_1.01,0.23454079627990723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,4,power_law_1.01,0.13952640295028687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,4,power_law_1.01,0.05625600218772888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,4,power_law_1.01,0.2817023992538452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,4,power_law_1.01,0.4897600173950195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,4,power_law_1.01,0.1474943995475769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,4,power_law_1.01,0.06809599995613098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,4,power_law_1.01,0.3243583917617798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,4,power_law_1.01,0.15084799528121948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,4,power_law_1.01,0.5316544055938721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,4,power_law_1.01,0.06783360242843628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,4,power_law_1.01,0.41397762298583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,4,power_law_1.01,0.1551424026489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,4,power_law_1.01,0.6936831951141358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,4,power_law_1.01,0.07319039702415467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,4,power_law_1.01,0.16505600214004518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,4,power_law_1.01,0.48660478591918943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,4,power_law_1.01,0.07514879703521729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,4,power_law_1.01,0.8203519821166992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,4,power_law_1.01,0.18410880565643312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,4,power_law_1.01,0.6744063854217529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.01,0.20172159671783446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,4,power_law_1.01,0.07839360237121581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,4,power_law_1.01,0.8347519874572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,4,power_law_1.01,1.1097344398498534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.01,0.241158390045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,4,balanced,3.300800005594889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,4,power_law_1.01,1.187507152557373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,4,power_law_1.01,0.08080639839172363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,4,power_law_1.01,1.484499168395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.01,0.27767679691314695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,4,power_law_1.01,0.08270080089569092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,4,balanced,4.24075190226237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,4,power_law_1.01,1.5406847953796388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.01,0.3397056102752686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,4,power_law_1.01,1.7777088165283204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,4,power_law_1.01,0.08450559973716736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.01,0.426470422744751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,4,power_law_1.01,1.804159927368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,4,power_law_1.01,0.09224960207939148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,4,power_law_1.01,2.5583423614501952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.01,0.5847936153411866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,4,power_law_1.01,0.09317759871482849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,4,power_law_1.01,2.9919551849365233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,4,power_law_1.01,0.09599360227584838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.01,0.711897611618042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,4,power_law_1.01,5.2604927062988285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,4,power_law_1.01,0.17509119510650634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.01,0.19470080137252807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,4,power_law_1.01,5.7611328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.01,1.0404607772827148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.01,0.18188799619674684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.01,1.3899776458740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.01,0.1832383990287781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.01,0.22610559463500976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,4,power_law_1.01,1.6390911102294923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.01,0.24722559452056886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,4,power_law_1.01,2.4268543243408205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.01,0.3195264101028442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,4,power_law_1.01,4.773587036132812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.01,0.3879551887512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.01,0.5498816013336182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.01,0.694758415222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,4,power_law_1.01,0.8263808250427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,4,power_law_1.01,1.2744192123413085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,4,power_law_1.01,2.494246482849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,4,power_law_1.2,0.0698303997516632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,4,power_law_1.2,0.09014400243759155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,4,power_law_1.2,0.059513598680496216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,4,power_law_1.2,0.08627840280532836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,4,power_law_1.2,0.11066240072250366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,4,power_law_1.2,0.1377344012260437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,4,power_law_1.2,0.1855679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,4,power_law_1.2,0.1942720055580139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,4,power_law_1.2,0.050496000051498416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,4,power_law_1.2,0.19920639991760253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,4,power_law_1.2,0.2077183961868286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,4,power_law_1.2,0.06941440105438232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,4,power_law_1.2,0.21062400341033935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,4,power_law_1.2,0.2262592077255249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,4,power_law_1.2,0.05175679922103882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,4,power_law_1.2,0.22780799865722656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,4,power_law_1.2,0.23425281047821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,4,power_law_1.2,0.07130240201950074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,4,power_law_1.2,0.24223361015319825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,4,power_law_1.2,0.08588160276412964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,4,power_law_1.2,0.2597440004348755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,4,power_law_1.2,0.10302079916000366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,4,power_law_1.2,0.27438080310821533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,4,power_law_1.2,0.021964800357818604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,4,power_law_1.2,0.12546559572219848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,4,power_law_1.2,0.3087807893753052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,4,power_law_1.2,0.05754240155220032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,4,power_law_1.2,0.1309440016746521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,4,power_law_1.2,0.3084415912628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,4,power_law_1.2,0.027744001150131224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,4,power_law_1.2,0.13701119422912597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,4,power_law_1.2,0.3698751926422119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,4,power_law_1.2,0.06416000127792358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,4,power_law_1.2,0.027193599939346315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,4,power_law_1.2,0.1381376028060913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,4,power_law_1.2,0.4074560165405273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,4,power_law_1.2,0.05919359922409058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,4,power_law_1.2,0.03304319977760315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,4,power_law_1.2,0.14730240106582643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,4,power_law_1.2,0.07363839745521546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,4,power_law_1.2,0.549241590499878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,4,power_law_1.2,0.15299199819564818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,4,power_law_1.2,0.04596480131149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,4,power_law_1.2,0.1598528027534485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,4,power_law_1.2,0.08424959778785705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,4,power_law_1.2,0.576793622970581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,4,power_law_1.2,0.053376001119613645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,4,power_law_1.2,0.16273280382156372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,4,power_law_1.2,0.09605119824409485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,4,power_law_1.2,0.744320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,4,power_law_1.2,0.1686784029006958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,4,power_law_1.2,0.0667519986629486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,4,power_law_1.2,0.12038400173187255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,4,power_law_1.2,0.1802559971809387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,4,power_law_1.2,0.9493439674377442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,4,power_law_1.2,0.06858239769935608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,4,power_law_1.2,0.19475200176239013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,4,power_law_1.2,0.12193919420242309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,4,power_law_1.2,0.07194240093231201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,4,power_law_1.2,0.22663040161132814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,4,power_law_1.2,1.2819135665893555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,4,power_law_1.2,0.12616319656372071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,4,power_law_1.2,0.07654399871826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,4,power_law_1.2,0.24343039989471435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,4,power_law_1.2,0.12673920392990112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,2,balanced,0.04990399877230326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,4,power_law_1.2,0.29167358875274657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,4,power_law_1.2,0.07878400087356567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,4,power_law_1.2,1.5992639541625977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,4,power_law_1.2,0.1313024044036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,4,power_law_1.2,0.3488447904586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,4,power_law_1.2,0.07985280156135559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,2,balanced,0.053946668903032936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,4,power_law_1.2,0.1385983943939209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,2,balanced,0.047781333327293396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,4,power_law_1.2,0.45294718742370604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,4,power_law_1.2,2.1931455612182615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,4,power_law_1.2,0.08165760040283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,2,balanced,0.07049066821734111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,4,power_law_1.2,0.14669439792633057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,4,power_law_1.2,0.5284287929534912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,4,power_law_1.2,0.08611199855804444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,2,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,2,balanced,0.10233066479365031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,4,power_law_1.2,0.7612095832824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,4,power_law_1.2,3.04901123046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,4,power_law_1.2,0.1455296039581299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,4,power_law_1.2,0.09610880017280579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,2,balanced,0.15572800238927206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,4,power_law_1.2,0.8424768447875977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,2,balanced,0.05598933498064677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,4,power_law_1.2,0.15214719772338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,2,balanced,0.2531306743621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,4,power_law_1.2,0.09602559804916382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,2,balanced,0.07246933380762736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,4,power_law_1.2,1.312985610961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,4,power_law_1.2,5.646527862548828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,4,power_law_1.2,0.16344319581985473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,2,balanced,0.2579253315925598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,2,balanced,0.08110933502515157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,4,power_law_1.2,0.09994239807128906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,4,power_law_1.2,1.7009088516235351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,2,balanced,0.2638186613718669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,2,balanced,0.07776000102361043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,2,balanced,0.11644267042477925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,4,power_law_1.2,0.1672063946723938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,4,power_law_1.2,2.0063743591308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,2,balanced,0.26206399997075397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,4,power_law_1.2,0.18661760091781615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,2,balanced,0.09425600369771321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,2,balanced,0.2605920036633809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,2,balanced,0.17198399702707926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,4,power_law_1.2,0.1888576030731201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,4,power_law_1.2,3.063212776184082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.2,0.20804479122161865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,2,balanced,0.2653866608937581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,2,balanced,0.12282666563987732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,2,balanced,0.17639466126759848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.2,0.21041278839111327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,2,balanced,0.021498667697111767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,2,balanced,0.26848532756169635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.2,0.17569279670715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,4,power_law_1.2,6.021766281127929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,2,balanced,0.18358933925628662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,2,balanced,0.17376534144083658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,2,balanced,0.2733440001805623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,2,balanced,0.023520000278949738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.2,0.24826879501342775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.2,0.19703680276870728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,2,balanced,0.2798186739285787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,2,balanced,0.027215999861558277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,2,balanced,0.27614933252334595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,2,balanced,0.17649600903193155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,2,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.2,0.28195838928222655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.2,0.22556800842285157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,2,balanced,0.2892533342043559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,2,balanced,0.06030400097370148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,2,balanced,0.17722133795420328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,2,balanced,0.2750613292058309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,2,balanced,0.08450133601824443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.2,0.35331199169158933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,2,balanced,0.2937120000521342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.2,0.25575039386749265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,2,balanced,0.08896000186602275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,2,balanced,0.17970667282740274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,2,balanced,0.09145067135492961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.2,0.43839359283447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,2,balanced,0.303765336672465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,2,balanced,0.2695573369661967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,2,balanced,0.18360000848770142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,2,balanced,0.09516266981760661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.2,0.34219520092010497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,2,balanced,0.09784533580144246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.2,0.5609024047851563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,2,balanced,0.18389334281285605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,2,balanced,0.3288000027338664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,2,balanced,0.10058666268984477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.2,0.41818881034851074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,2,balanced,0.2722880045572917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,2,balanced,0.19129600127538046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,2,balanced,0.1033066709836324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.2,0.7563583850860596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,2,balanced,0.3471306562423706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,2,balanced,0.10542399684588115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,2,balanced,0.2718613346417745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,2,balanced,0.19989866018295288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.2,0.5645952224731445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.2,1.077126407623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,2,balanced,0.10965866843859355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,2,balanced,0.27323732773462933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.2,1.4435327529907227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,2,balanced,0.11085866888364156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,2,balanced,0.20496533314387003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,2,balanced,0.2808213432629903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.2,0.7256447792053222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,2,balanced,0.3885973294576009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,2,balanced,0.2762826681137085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,2,balanced,0.11555733283360799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,2,balanced,0.22511466344197592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,4,power_law_1.2,1.6953664779663087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,4,power_law_1.2,0.9019840240478516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,2,balanced,0.2817973295847575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,2,balanced,0.1274773379166921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,2,balanced,0.2545813322067261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,2,balanced,0.14967466394106546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,2,balanced,0.28590933481852215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,4,power_law_1.2,1.3835968017578124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,2,balanced,0.42844800154368085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,4,power_law_1.2,2.9380863189697264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,2,balanced,0.2943519949913025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,2,balanced,0.15998400251070657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,2,balanced,0.27296000719070435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,2,balanced,0.29690666993459064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,4,power_law_1.2,2.601798439025879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,2,balanced,0.20441599686940512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,4,power_law_1.2,5.387295913696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,2,balanced,0.5416106780370077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,2,balanced,0.32233599821726483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,2,balanced,0.32068800926208496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,2,balanced,0.2015413244565328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,2,balanced,0.33507200082143146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,2,balanced,0.3823946714401245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,2,balanced,0.28148800134658813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,2,balanced,0.6276373465855917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,2,balanced,0.37455999851226807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,2,balanced,0.33647998174031574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,2,balanced,0.4800906578699748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,2,balanced,0.409717321395874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,2,balanced,0.8294293085734049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,2,balanced,0.49431467056274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,2,balanced,0.5780266523361206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,2,balanced,0.3924266497294108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,2,balanced,0.5749173164367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,2,balanced,0.9941120147705078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,2,balanced,0.8251893520355225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,2,balanced,0.43806934356689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,2,balanced,0.8590826988220215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,2,balanced,1.0093920230865479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,2,balanced,1.4160265922546387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,2,balanced,0.7131040096282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,2,balanced,1.0266559918721516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,2,balanced,1.4640000661214192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,2,balanced,0.8332800070444742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,2,balanced,1.7431093851725261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,2,balanced,1.4739200274149578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,2,balanced,1.874831994374593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,2,balanced,1.8999892870585124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,2,balanced,1.1112053394317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,2,balanced,2.1650452613830566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,2,balanced,2.3215413093566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,2,balanced,2.3468586603800454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,2,balanced,1.637887954711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,2,balanced,3.257141431172689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,2,balanced,3.570821444193522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,2,balanced,3.648218790690104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,2,power_law_1.01,0.049728000164031984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,2,power_law_1.01,0.04683519899845123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,2,power_law_1.01,0.06783999800682068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,2,power_law_1.01,0.055180799961090085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,2,power_law_1.01,0.059545600414276124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,2,power_law_1.01,0.05379840135574341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,2,power_law_1.01,0.08588160276412964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,2,power_law_1.01,0.071424001455307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,2,power_law_1.01,0.1223296046257019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,2,power_law_1.01,0.09278720021247863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,2,power_law_1.01,0.1456768035888672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,2,power_law_1.01,0.11404160261154175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,2,power_law_1.01,0.19640320539474487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,2,power_law_1.01,0.140447998046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,2,power_law_1.01,0.20465281009674072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,2,power_law_1.01,0.14954240322113038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,2,power_law_1.01,0.20877439975738527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,2,power_law_1.01,0.15185920000076295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,2,power_law_1.01,0.21905920505523682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,2,power_law_1.01,0.15801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,2,power_law_1.01,0.1615615963935852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,2,power_law_1.01,0.2243135929107666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,2,balanced,3.2248853047688804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,2,power_law_1.01,0.1729856014251709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,2,power_law_1.01,0.23442559242248534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,2,balanced,6.273893356323242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,2,power_law_1.01,0.17902719974517822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,2,power_law_1.01,0.2353600025177002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,2,power_law_1.01,0.18906240463256835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,2,power_law_1.01,0.07943040132522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,2,power_law_1.01,0.2537600040435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,2,power_law_1.01,0.020883199572563172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,2,power_law_1.01,0.20595200061798097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,2,power_law_1.01,0.267577600479126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,2,power_law_1.01,0.09433599710464477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,2,power_law_1.01,0.21738240718841553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,2,power_law_1.01,0.022937600314617158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,2,power_law_1.01,0.09800320267677307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,2,power_law_1.01,0.27993600368499755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,2,power_law_1.01,0.22427520751953126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,2,power_law_1.01,0.1116096019744873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,2,power_law_1.01,0.026099199056625368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,2,power_law_1.01,0.2907263994216919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,2,power_law_1.01,0.25656321048736574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,2,power_law_1.01,0.1488960027694702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,2,power_law_1.01,0.03491199910640717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,2,power_law_1.01,0.28447999954223635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,2,power_law_1.01,0.325708794593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,2,power_law_1.01,0.16743040084838867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,2,power_law_1.01,0.04986880123615265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,2,balanced,6.913861592610677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,2,power_law_1.01,0.3538624048233032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,2,power_law_1.01,0.21475839614868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,2,power_law_1.01,0.356332802772522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,2,power_law_1.01,0.06026880145072937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,2,power_law_1.01,0.41296000480651857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,2,power_law_1.01,0.22483201026916505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,2,power_law_1.01,0.4181056022644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,2,power_law_1.01,0.07170559763908387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,2,power_law_1.01,0.22906239032745362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,2,power_law_1.01,0.5236480236053467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,2,power_law_1.01,0.46849918365478516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,2,power_law_1.01,0.24225280284881592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,2,balanced,7.211578369140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,2,power_law_1.01,0.07303040027618408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,2,power_law_1.01,0.6412032127380372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,2,power_law_1.01,0.23941121101379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,2,power_law_1.01,0.07700480222702026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,2,power_law_1.01,0.5848896026611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,2,power_law_1.01,0.8719807624816894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,2,power_law_1.01,0.08174719810485839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,2,power_law_1.01,0.25075840950012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,2,power_law_1.01,1.0802176475524903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,2,power_law_1.01,0.08386560082435608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,2,power_law_1.01,0.250982403755188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,2,power_law_1.01,0.6828479766845703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,2,power_law_1.01,0.0866752028465271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,2,power_law_1.01,0.2565119981765747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,2,power_law_1.01,1.5793024063110352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,2,power_law_1.01,0.08872320055961609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,2,power_law_1.01,0.2642816066741943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,2,power_law_1.01,0.9111616134643554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,2,power_law_1.01,0.08893439769744874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,2,power_law_1.01,0.26942079067230223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,2,power_law_1.01,1.9822271347045899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,2,power_law_1.01,0.2831104040145874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,2,power_law_1.01,0.09441279768943786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,2,power_law_1.01,1.1261887550354004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,2,power_law_1.01,0.30743680000305174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,2,power_law_1.01,0.0956928014755249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,2,power_law_1.01,1.5171839714050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,2,power_law_1.01,2.539116859436035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.01,0.3390336036682129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,2,power_law_1.01,0.10673919916152955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,2,power_law_1.01,1.8670335769653321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.01,0.39091200828552247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,2,power_law_1.01,3.8479167938232424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,2,power_law_1.01,0.1274623990058899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.01,0.45725440979003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,2,power_law_1.01,2.3258623123168944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.01,0.14169600009918212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.01,0.573414421081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,2,power_law_1.01,7.617990112304687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,2,power_law_1.01,3.6458175659179686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.01,0.17618559598922728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.01,0.6696512222290039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.01,0.18812799453735352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.01,0.9127872467041016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,2,power_law_1.01,6.392326354980469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.01,0.23815040588378905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.01,1.1954367637634278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.01,0.28264319896698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.01,1.621913528442383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.01,0.35518081188201905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.01,2.0651840209960937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.01,0.4294464111328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.01,0.5811520099639893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,2,power_law_1.01,2.566726493835449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.01,0.7332863807678223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,2,power_law_1.01,0.8702912330627441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,2,power_law_1.01,4.012876892089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,2,power_law_1.01,1.3207232475280761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,2,power_law_1.01,7.828684997558594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,2,power_law_1.01,2.5842559814453123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,2,power_law_1.2,0.0491456001996994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,2,power_law_1.2,0.0474368005990982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,2,power_law_1.2,0.06695680022239685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,2,power_law_1.2,0.053286397457122804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,2,power_law_1.2,0.06074240207672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,2,power_law_1.2,0.05453439950942993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,2,power_law_1.2,0.0893119990825653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,2,power_law_1.2,0.07092480063438415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,2,power_law_1.2,0.11304960250854493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,2,power_law_1.2,0.08556159734725952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,2,power_law_1.2,0.1340224027633667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,2,power_law_1.2,0.10478719472885131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,2,power_law_1.2,0.18691200017929077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,2,power_law_1.2,0.1411520004272461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,2,power_law_1.2,0.20050559043884278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,2,power_law_1.2,0.08046720027923585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,2,power_law_1.2,0.14687999486923217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,2,power_law_1.2,0.21140480041503906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,2,power_law_1.2,0.020608000457286835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,2,power_law_1.2,0.15072640180587768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,2,power_law_1.2,0.21345279216766358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,2,power_law_1.2,0.0906175971031189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,2,power_law_1.2,0.22177278995513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,2,power_law_1.2,0.1574720025062561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,2,power_law_1.2,0.022988800704479218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,2,power_law_1.2,0.08516479730606079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,2,power_law_1.2,0.23087360858917236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,2,power_law_1.2,0.16363519430160522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,2,power_law_1.2,0.11224960088729859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,2,power_law_1.2,0.026284798979759216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,2,power_law_1.2,0.2381824016571045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,2,power_law_1.2,0.16966400146484376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,2,power_law_1.2,0.1374079942703247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,2,power_law_1.2,0.0316864013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,2,power_law_1.2,0.25742719173431394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,2,power_law_1.2,0.15772160291671752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,2,power_law_1.2,0.1792512059211731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,2,power_law_1.2,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,2,power_law_1.2,0.26866559982299804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,2,power_law_1.2,0.21478400230407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,2,power_law_1.2,0.19782400131225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,2,power_law_1.2,0.2822144031524658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,2,power_law_1.2,0.05633919835090637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,2,power_law_1.2,0.21967360973358155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,2,power_law_1.2,0.2078847885131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,2,power_law_1.2,0.298361611366272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,2,power_law_1.2,0.22522881031036376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,2,power_law_1.2,0.06912639737129211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,2,power_law_1.2,0.22376320362091065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,2,power_law_1.2,0.23478400707244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,2,power_law_1.2,0.3358527898788452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,2,power_law_1.2,0.07404800057411194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,2,power_law_1.2,0.2320512056350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,2,power_law_1.2,0.23987200260162353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,2,power_law_1.2,0.07747200131416321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,2,power_law_1.2,0.3778496026992798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,2,power_law_1.2,0.26577279567718504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,2,power_law_1.2,0.2518784046173096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,2,power_law_1.2,0.080595201253891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,2,power_law_1.2,0.2972480058670044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,2,power_law_1.2,0.4392127990722656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,2,power_law_1.2,0.0825215995311737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,2,power_law_1.2,0.25169920921325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,2,power_law_1.2,0.35583360195159913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,2,power_law_1.2,0.5064511775970459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,2,power_law_1.2,0.08721920251846313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,2,power_law_1.2,0.2576575994491577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,2,power_law_1.2,0.4319295883178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,2,power_law_1.2,0.5963327884674072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,1,balanced,0.05602133274078369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,2,power_law_1.2,0.09203199744224548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,2,power_law_1.2,0.2705984115600586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,2,power_law_1.2,0.6936639785766602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,2,power_law_1.2,0.5513663768768311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,1,balanced,0.06145599981149038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,2,power_law_1.2,0.09055359959602356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,2,power_law_1.2,0.2847104072570801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,2,power_law_1.2,0.9272704124450684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,2,power_law_1.2,0.6410367965698243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,1,balanced,0.07868800063927968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,1,balanced,0.11364799737930298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,2,power_law_1.2,0.2975104093551636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,2,power_law_1.2,0.09562240242958069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,1,balanced,0.05230399966239929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,2,power_law_1.2,0.8638208389282227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,2,power_law_1.2,1.201260757446289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,1,balanced,0.17622933785120645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,2,power_law_1.2,0.31797759532928466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,1,balanced,0.293994665145874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,2,power_law_1.2,0.0995136022567749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,2,power_law_1.2,1.1277312278747558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,2,power_law_1.2,1.6106624603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,1,balanced,0.29923733075459796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,1,balanced,0.05599466462930044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.2,0.34985599517822263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,2,power_law_1.2,0.10911359786987304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,1,balanced,0.3020693262418111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,1,balanced,0.06585066517194112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,2,power_law_1.2,1.6362943649291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,2,power_law_1.2,2.096793556213379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,1,balanced,0.30369067192077637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,2,power_law_1.2,0.13046400547027587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,1,balanced,0.09682666261990865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.2,0.4247039794921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,1,balanced,0.30664533376693726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,1,balanced,0.14637333154678345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,2,power_law_1.2,2.0469375610351563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.2,0.14467840194702147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.2,0.46628479957580565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,2,power_law_1.2,2.467955207824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,1,balanced,0.3099520007769267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,1,balanced,0.2204213341077169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.2,0.17921279668807982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,1,balanced,0.10146666566530864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,1,balanced,0.31666133801142377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.2,0.5798463821411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,1,balanced,0.22523732980092367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,2,power_law_1.2,2.5383039474487306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,1,balanced,0.32037333647410077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.2,0.18699519634246825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,1,balanced,0.22817067305246988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,1,balanced,0.02141333371400833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.2,0.7280896186828614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,2,power_law_1.2,3.7393150329589844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,1,balanced,0.11158399780591328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,1,balanced,0.3253920078277588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,1,balanced,0.2299786607424418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.2,0.24272000789642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.2,0.9703104019165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,1,balanced,0.33609068393707275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,1,balanced,0.23138133684794107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,2,power_law_1.2,4.205721664428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,1,balanced,0.13858133554458618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.2,0.29066240787506104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,1,balanced,0.343173344930013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,1,balanced,0.23845332860946655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.2,1.2058752059936524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,1,balanced,0.023242667317390442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,1,balanced,0.1829973260561625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,2,power_law_1.2,7.037203216552735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.2,0.37355520725250246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,1,balanced,0.24299200375874838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,1,balanced,0.02548266698916753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,1,balanced,0.3577226797739665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.2,1.6709375381469727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,1,balanced,0.2733546694119771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,1,balanced,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,1,balanced,0.24592000246047974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,2,power_law_1.2,7.914816284179688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.2,0.43262081146240233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,1,balanced,0.4394506613413493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,1,balanced,0.4004160165786743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,1,balanced,0.04604800045490265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.2,2.228998374938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,1,balanced,0.2545386751492818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,1,balanced,0.07658133407433827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,1,balanced,0.4357066551844279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.2,0.5760640144348145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,1,balanced,0.26360533634821576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,1,balanced,0.08059733112653096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,1,balanced,0.43594666322072345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,1,balanced,0.4215253194173177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,2,power_law_1.2,2.5735488891601563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,1,balanced,0.08455466230710347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,1,balanced,0.27129065990448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,1,balanced,0.4333440065383911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.2,0.7639039993286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,1,balanced,0.08678399523099263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,1,balanced,0.5121066570281982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,1,balanced,0.42958398660024005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,1,balanced,0.2879786690076192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,1,balanced,0.08892800410588582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,2,power_law_1.2,4.4333953857421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,2,power_law_1.2,0.9060864448547363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,1,balanced,0.43539198239644367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,1,balanced,0.09158933162689209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,1,balanced,0.3324693242708842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,1,balanced,0.44248533248901367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,1,balanced,0.0946560005346934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,1,balanced,0.5707199970881144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,1,balanced,0.09673600395520528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,2,power_law_1.2,1.3577407836914062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,1,balanced,0.44011199474334717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,1,balanced,0.3646719853083293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,2,power_law_1.2,7.5189697265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,1,balanced,0.09823999802271526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,1,balanced,0.4498186508814494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,1,balanced,0.10126399993896484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,1,balanced,0.4554239908854167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,1,balanced,0.7803413073221842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,1,balanced,0.44548265139261883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,2,power_law_1.2,2.855219268798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,1,balanced,0.10589333375295003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,1,balanced,0.46240532398223877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,1,balanced,0.11236266295115153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,1,balanced,0.5148959954579672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,1,balanced,0.12781866391499838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,1,balanced,0.47870934009552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,1,balanced,0.8957173029581705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,1,balanced,0.5173066854476929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,1,balanced,0.14289066195487976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,1,balanced,0.7280373573303223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,1,balanced,0.5482826630274454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,1,balanced,0.18358399470647177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,1,balanced,0.6283733447392782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,1,balanced,0.18645866711934408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,1,balanced,1.2814826965332031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,1,balanced,0.8520853519439697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,1,balanced,0.25725332895914715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,1,balanced,0.6954507033030192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,1,balanced,1.5630292892456055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,1,balanced,0.8495360215504965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,1,balanced,1.2535146872202556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,1,balanced,0.3101653258005778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,1,balanced,1.0069759686787922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,1,balanced,0.4193546772003174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,1,balanced,1.5593172709147136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,1,balanced,2.2650453249613443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,1,balanced,1.5279040336608887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,1,balanced,0.5320693254470825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,1,balanced,2.2845279375712075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,1,balanced,1.8393440246582031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,1,balanced,2.9319680531819663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,1,balanced,0.765221357345581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,1,balanced,2.6677331924438477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,1,balanced,2.978117307027181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,1,balanced,3.5447680155436196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,1,balanced,1.013327995936076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,1,balanced,3.473290761311849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,1,balanced,3.658170700073242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,1,balanced,1.2549280325571697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,1,balanced,4.285530726114909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,1,balanced,5.421845118204753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,1,balanced,5.685658772786458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,1,balanced,1.9569759368896484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,1,balanced,6.717967987060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,1,power_law_1.01,0.04875519871711731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,1,power_law_1.01,0.0506496012210846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,1,power_law_1.01,0.05188480019569397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,1,power_law_1.01,0.059552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,1,power_law_1.01,0.05464320182800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,1,power_law_1.01,0.08695039749145508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,1,power_law_1.01,0.06553599834442139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,1,power_law_1.01,0.11018879413604736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,1,power_law_1.01,0.10117119550704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,1,power_law_1.01,0.13567999601364136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,1,power_law_1.01,0.1318719983100891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,1,balanced,10.437311808268229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,1,power_law_1.01,0.17390719652175904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,1,power_law_1.01,0.16888320446014404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,1,power_law_1.01,0.18471039533615113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,1,power_law_1.01,0.21899518966674805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,1,power_law_1.01,0.19118720293045044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,1,power_law_1.01,0.23656320571899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,1,power_law_1.01,0.1957695960998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,1,power_law_1.01,0.24559359550476073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,1,power_law_1.01,0.09729920029640197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,1,power_law_1.01,0.20292479991912843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,1,power_law_1.01,0.2547647953033447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,1,power_law_1.01,0.22028160095214844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,1,power_law_1.01,0.020812800526618956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,1,power_law_1.01,0.26095359325408934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,1,power_law_1.01,0.10906879901885987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,1,power_law_1.01,0.22933759689331054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,1,power_law_1.01,0.2754431962966919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,1,power_law_1.01,0.12102400064468384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,1,power_law_1.01,0.02240000069141388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,1,power_law_1.01,0.2471679925918579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,1,power_law_1.01,0.28149120807647704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,1,power_law_1.01,0.26973440647125246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,1,power_law_1.01,0.1674496054649353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,1,balanced,11.101744333902994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,1,power_law_1.01,0.29461119174957273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,1,power_law_1.01,0.024940800666809083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,1,power_law_1.01,0.2829184055328369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,1,power_law_1.01,0.2971199989318848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,1,power_law_1.01,0.21045119762420655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,1,power_law_1.01,0.030054399371147157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,1,power_law_1.01,0.27511680126190186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,1,power_law_1.01,0.3127295970916748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,1,power_law_1.01,0.2545023918151855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,1,power_law_1.01,0.33320960998535154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,1,power_law_1.01,0.04731520116329193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,1,power_law_1.01,0.33544321060180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,1,power_law_1.01,0.39365758895874026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,1,power_law_1.01,0.3713023900985718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,1,power_law_1.01,0.3267008066177368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,1,power_law_1.01,0.0688256025314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,1,power_law_1.01,0.43714561462402346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,1,power_law_1.01,0.4710527896881104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,1,power_law_1.01,0.3555327892303467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,1,power_law_1.01,0.08056960105895997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,1,power_law_1.01,0.5258048057556153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,1,power_law_1.01,0.5528255939483643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,1,balanced,3.847317377726237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,1,power_law_1.01,0.34545919895172117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,1,power_law_1.01,0.6009471893310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,1,power_law_1.01,0.08834559917449951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,1,power_law_1.01,0.3541440010070801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,1,power_law_1.01,0.7310976028442383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,1,balanced,13.208768208821615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,1,power_law_1.01,0.3680255889892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,1,power_law_1.01,0.772108793258667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,1,power_law_1.01,0.9115712165832519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,1,power_law_1.01,0.09049599766731262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,1,power_law_1.01,0.38227200508117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,1,power_law_1.01,0.9523648262023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,1,power_law_1.01,1.2542847633361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,1,power_law_1.01,0.0933247983455658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,1,power_law_1.01,0.38224639892578127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,1,power_law_1.01,1.2836095809936523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,1,power_law_1.01,0.09829760193824769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,1,power_law_1.01,0.39836161136627196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,1,power_law_1.01,1.603513526916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,1,power_law_1.01,0.40915842056274415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,1,power_law_1.01,1.5995712280273438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,1,power_law_1.01,0.10065920352935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,1,power_law_1.01,2.2819391250610352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,1,power_law_1.01,0.414796781539917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,1,power_law_1.01,0.10794880390167236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,1,power_law_1.01,2.2327104568481446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,1,power_law_1.01,0.10637439489364624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,1,power_law_1.01,2.975411224365234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,1,power_law_1.01,0.4350143909454346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,1,power_law_1.01,2.8770559310913084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,1,power_law_1.01,0.11270400285720825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,1,power_law_1.01,0.4931072235107422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,1,power_law_1.01,3.521209716796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,1,power_law_1.01,3.6582847595214845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,1,power_law_1.01,0.1158911943435669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.01,0.5417088031768799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,1,power_law_1.01,0.1223680019378662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,1,power_law_1.01,5.434675216674805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,1,power_law_1.01,5.7053375244140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,1,power_law_1.01,0.14364160299301149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.01,0.6452159881591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.01,0.1619328022003174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.01,0.19779200553894044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,1,power_law_1.01,10.367596435546876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,1,power_law_1.01,11.108000183105469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.01,0.7321983814239502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.01,0.22213759422302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.01,0.2762495994567871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.01,0.9456959724426269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.01,0.349126410484314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.01,1.1347264289855956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.01,0.4782144069671631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.01,1.5343744277954101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.01,0.5678400039672852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.01,1.9566463470458983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.01,0.8086848258972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.01,2.754764747619629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.01,1.0398207664489747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.01,3.547808074951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,1,power_law_1.01,1.275692844390869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,1,power_law_1.01,4.337728118896484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,1,power_law_1.01,1.9804416656494142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,1,power_law_1.01,6.732569885253906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,1,power_law_1.01,3.8733440399169923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,1,power_law_1.01,13.113548278808594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,1,power_law_1.2,0.049983999133110045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,1,power_law_1.2,0.054425597190856934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,1,power_law_1.2,0.06488959789276123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,1,power_law_1.2,0.0951744019985199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,1,power_law_1.2,0.113811194896698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,1,power_law_1.2,0.15215359926223754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,1,power_law_1.2,0.21917440891265869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,1,power_law_1.2,0.23057279586791993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,1,power_law_1.2,0.24133119583129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,1,power_law_1.2,0.25191679000854494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,1,power_law_1.2,0.26263039112091063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,1,power_law_1.2,0.2748159885406494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,1,power_law_1.2,0.2810944080352783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,1,power_law_1.2,0.2979327917098999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,1,power_law_1.2,0.30805120468139646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,1,power_law_1.2,0.3197376012802124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,1,power_law_1.2,0.34921600818634035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,1,power_law_1.2,0.4069119930267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,1,power_law_1.2,0.44688000679016116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,1,power_law_1.2,0.5391168117523193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,1,power_law_1.2,0.6119296073913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,1,power_law_1.2,0.7757823944091797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,1,power_law_1.2,0.049055999517440795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,1,power_law_1.2,0.9667840003967285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,1,power_law_1.2,0.09608319997787476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,1,power_law_1.2,0.05317760109901428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,1,power_law_1.2,1.28603515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,1,power_law_1.2,0.06040319800376892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,1,power_law_1.2,0.1071679949760437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,1,power_law_1.2,1.5935872077941895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,1,power_law_1.2,0.08350719809532166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,1,power_law_1.2,0.12021759748458863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,1,power_law_1.2,2.2324544906616213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,1,power_law_1.2,0.1555392026901245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,1,power_law_1.2,0.10393600463867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,1,power_law_1.2,2.8839807510375977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,1,power_law_1.2,0.18974080085754394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,1,power_law_1.2,0.1271999955177307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,1,power_law_1.2,0.23350400924682618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,1,power_law_1.2,3.5306686401367187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,1,power_law_1.2,0.1717120051383972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,1,power_law_1.2,0.021087999641895293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,1,power_law_1.2,0.3141952037811279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,1,power_law_1.2,0.17864320278167725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,1,power_law_1.2,5.442335891723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,1,power_law_1.2,0.3304960012435913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,1,power_law_1.2,0.1883455991744995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,1,power_law_1.2,0.022067199647426605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,1,power_law_1.2,0.3470655918121338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,1,power_law_1.2,0.19201279878616334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,1,power_law_1.2,0.025536000728607178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,1,power_law_1.2,10.365977478027343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,1,power_law_1.2,0.3603904008865356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,1,power_law_1.2,0.2091007947921753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,1,power_law_1.2,0.03191039860248566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,1,power_law_1.2,0.3690560102462769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,1,power_law_1.2,0.22268800735473632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,1,power_law_1.2,0.3767231941223145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,1,power_law_1.2,0.0443583995103836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,1,power_law_1.2,0.2327359914779663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,1,power_law_1.2,0.39105279445648194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,1,power_law_1.2,0.06736639738082886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,1,power_law_1.2,0.2528000116348267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,1,power_law_1.2,0.40021119117736814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,1,power_law_1.2,0.08037760257720947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,1,power_law_1.2,0.27470719814300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,1,power_law_1.2,0.41391358375549314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,1,power_law_1.2,0.08631680011749268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,1,power_law_1.2,0.2906176090240479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,1,power_law_1.2,0.4368256092071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,1,power_law_1.2,0.08922240138053894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,1,power_law_1.2,0.2823040008544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,1,power_law_1.2,0.45995521545410156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,1,power_law_1.2,0.0942463994026184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,1,power_law_1.2,0.3440191984176636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,1,power_law_1.2,0.5201151847839356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,1,power_law_1.2,0.09644799828529357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,1,power_law_1.2,0.38100481033325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.2,0.5644032001495362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,1,power_law_1.2,0.1041983962059021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,1,power_law_1.2,0.47882881164550783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,1,power_law_1.2,0.10548479557037353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,1,power_law_1.2,0.5570367813110352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.2,0.6661759853363037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,1,power_law_1.2,0.10586880445480347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,1,power_law_1.2,0.7363903999328614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.2,0.758406400680542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,1,power_law_1.2,0.11203199625015259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,1,power_law_1.2,0.9272128105163574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.2,0.9606783866882325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,1,power_law_1.2,0.11723519563674926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,1,power_law_1.2,1.2586175918579101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.2,1.1601920127868652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,1,power_law_1.2,0.12417919635772705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,1,power_law_1.2,1.6032896041870117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.2,1.5617919921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,1,power_law_1.2,0.14603519439697266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.2,0.16549760103225708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.2,1.9528959274291993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,16,balanced,0.025290665527184803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,16,balanced,0.029717333614826202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,1,power_law_1.2,2.2745792388916017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.2,0.20153601169586183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.2,2.759769630432129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,1,power_law_1.2,2.969382476806641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,16,balanced,0.027269333600997925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.2,0.22164480686187743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,16,balanced,0.029616000751654308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,16,balanced,0.02962133288383484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,16,balanced,0.025455998877684276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.2,0.27917439937591554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.2,3.5375232696533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,16,balanced,0.033557333052158356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,16,balanced,0.027477333943049114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,1,power_law_1.2,3.6509376525878907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,16,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.2,0.3513024091720581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,16,balanced,0.033546666304270424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,16,balanced,0.035642666121323906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,16,balanced,0.0354720006386439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,16,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.2,0.4776832103729248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,1,power_law_1.2,4.332236862182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,16,balanced,0.03531199942032496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,1,power_law_1.2,5.697094345092774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,16,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.2,0.573747205734253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,16,balanced,0.03588266670703888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,16,balanced,0.03562666724125544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,16,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,1,power_law_1.2,6.7202812194824215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,16,balanced,0.03626666714747747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.2,0.8051839828491211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,16,balanced,0.036677333215872444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,16,balanced,0.03612266729275385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,1,power_law_1.2,11.098521423339843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.2,1.049177646636963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,16,balanced,0.03775999943415324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,16,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,16,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,16,balanced,0.03738133360942205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,1,power_law_1.2,1.2789952278137207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,16,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,1,power_law_1.2,13.093849182128906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,16,balanced,0.037392000357309975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,16,balanced,0.041840001940727234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,16,balanced,0.03761066744724909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,16,balanced,0.04189866781234741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,1,power_law_1.2,1.9808256149291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,16,balanced,0.04177600145339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,16,balanced,0.041877334316571556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,16,balanced,0.039887999494870506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,16,balanced,0.01913600042462349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,16,balanced,0.04420800010363261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,16,balanced,0.03216533362865448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,1,power_law_1.2,3.8740608215332033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,16,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,16,balanced,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,16,balanced,0.01916266605257988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,16,balanced,0.032698666055997215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,16,balanced,0.04373333354791006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,16,balanced,0.02006400004029274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,16,balanced,0.03358400116364161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,16,balanced,0.05227200190226237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,16,balanced,0.035589332381884255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,16,balanced,0.019727999965349834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,16,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,16,balanced,0.04799466828505198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,16,balanced,0.021418665846188862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,16,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,16,balanced,0.056976000467936196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,16,balanced,0.02314666658639908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,16,balanced,0.03789866715669632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,16,balanced,0.050144001841545105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,16,balanced,0.03771200031042099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,16,balanced,0.025578667720158894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,16,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,16,balanced,0.0749013324578603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,16,balanced,0.02513066679239273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,16,balanced,0.039135999977588654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,16,balanced,0.05641599992911021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,16,balanced,0.03562666724125544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,16,balanced,0.03940266619126002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,16,balanced,0.039834665755430855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,16,balanced,0.035775999228159584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,16,balanced,0.03959999978542328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,16,balanced,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,16,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,16,balanced,0.08801066875457764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,16,balanced,0.038719999293486275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,16,balanced,0.07129066685835521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,16,balanced,0.03703466554482778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,16,balanced,0.043893332282702126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,16,balanced,0.035429333647092186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,16,balanced,0.043605332573254905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,16,balanced,0.04371733466784159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,16,balanced,0.04428266485532125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,16,balanced,0.041749333341916404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,16,balanced,0.04571733375390371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,16,balanced,0.07435733576615651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,16,balanced,0.1125333309173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,16,balanced,0.0395359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,16,balanced,0.0498933345079422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,16,balanced,0.04510400195916494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,16,balanced,0.05144000053405762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,16,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,16,balanced,0.08807999889055888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,16,balanced,0.060165335734685264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,16,balanced,0.05895466605822245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,16,balanced,0.141893337170283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,16,balanced,0.06445333361625671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,16,balanced,0.06645333270231883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,16,balanced,0.08261333405971527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,16,balanced,0.07467733323574066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,16,balanced,0.10819733142852783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,16,balanced,0.09869866569836934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,16,balanced,0.18781334161758423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,16,balanced,0.09516266981760661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,16,balanced,0.1340000033378601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,16,balanced,0.13642133275667825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,16,balanced,0.12355200449625652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,16,balanced,0.23060800631841025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,16,balanced,0.1715573271115621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,16,balanced,0.15589333573977152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,16,balanced,0.15777066349983215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,16,balanced,0.24263467391331991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,16,balanced,0.18900267283121744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,16,balanced,0.27967466910680133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,16,balanced,0.1853440006573995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,16,balanced,0.3164373238881429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,16,balanced,0.22592000166575113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,16,power_law_1.01,0.029151999950408937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,16,power_law_1.01,0.02800000011920929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,16,balanced,0.3872426748275757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,16,power_law_1.01,0.02723200023174286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,16,power_law_1.01,0.027667200565338133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,16,balanced,0.4117013216018677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,16,power_law_1.01,0.03162240087985992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,16,power_law_1.01,0.03433600068092346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,16,power_law_1.01,0.03423359990119934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,16,balanced,0.25749866167704266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,16,power_law_1.01,0.03568640053272247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,16,power_law_1.01,0.03709439933300018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,16,power_law_1.01,0.03866879940032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,16,power_law_1.01,0.04094719886779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,16,power_law_1.01,0.04120959937572479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,16,power_law_1.01,0.041126400232315063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,16,power_law_1.01,0.04355199933052063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,16,power_law_1.01,0.04921599924564361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,16,power_law_1.01,0.049081599712371825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,16,power_law_1.01,0.04922879934310913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,16,balanced,0.32073599100112915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,16,power_law_1.01,0.055308800935745236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,16,power_law_1.01,0.06165120005607605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,16,power_law_1.01,0.0697920024394989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,16,power_law_1.01,0.07914239764213563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,16,power_law_1.01,0.09438719749450683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,16,power_law_1.01,0.10781439542770385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,16,power_law_1.01,0.14513280391693115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,16,power_law_1.01,0.17182079553604127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,16,power_law_1.01,0.23515520095825196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,16,power_law_1.01,0.25690879821777346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,16,balanced,0.6054506699244181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,16,power_law_1.01,0.3363647937774658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,16,power_law_1.01,0.511737585067749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,16,power_law_1.01,1.1213631629943848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,16,power_law_1.01,0.03550719916820526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,16,power_law_1.01,0.01852159947156906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,16,power_law_1.01,0.031769600510597226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,16,power_law_1.01,0.04019840061664581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,16,power_law_1.01,0.01774719953536987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,16,power_law_1.01,0.02842240035533905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,16,power_law_1.01,0.018323199450969697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,16,power_law_1.01,0.030297601222991945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,16,power_law_1.01,0.01812479943037033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,16,power_law_1.01,0.033580800890922545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,16,power_law_1.01,0.031142398715019226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,16,power_law_1.01,0.019679999351501463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,16,power_law_1.01,0.03255679905414581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,16,power_law_1.01,0.032595199346542356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,16,power_law_1.01,0.021484799683094025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,16,power_law_1.01,0.03399679958820343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,16,power_law_1.01,0.03320319950580597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,16,power_law_1.01,0.02503040134906769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,16,power_law_1.01,0.034867200255393985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,16,power_law_1.01,0.0341376006603241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,16,power_law_1.01,0.024249599874019624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,16,power_law_1.01,0.03590399920940399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,16,power_law_1.01,0.03449600040912628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,16,power_law_1.01,0.03504000008106232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,16,power_law_1.01,0.03553920090198517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,16,power_law_1.01,0.03500159978866577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,16,power_law_1.01,0.036396801471710205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,16,balanced,0.7922773361206055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,16,power_law_1.01,0.03816959857940674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,16,power_law_1.01,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,16,power_law_1.01,0.03674240112304687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,16,balanced,0.46198399861653644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,16,power_law_1.01,0.038841599225997926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,16,power_law_1.01,0.038867199420928956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,16,power_law_1.01,0.03692800104618073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,16,power_law_1.01,0.040140798687934874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,16,power_law_1.01,0.03790720105171204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,16,power_law_1.01,0.03716480135917664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,16,power_law_1.01,0.03869439959526062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,16,power_law_1.01,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,16,power_law_1.01,0.03573760092258453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,16,power_law_1.01,0.04469760060310364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,16,power_law_1.01,0.03850240111351013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,16,power_law_1.01,0.036735999584198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,16,power_law_1.01,0.04475519955158234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,16,power_law_1.01,0.03888640105724335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,16,power_law_1.01,0.036006399989128114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,16,power_law_1.01,0.048153600096702574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,16,balanced,0.6040106614430746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,16,power_law_1.01,0.05247359871864319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,16,power_law_1.01,0.03992320001125336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,16,power_law_1.01,0.037299200892448425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,16,power_law_1.01,0.05862399935722351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,16,power_law_1.01,0.041388800740242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,16,power_law_1.01,0.0412416011095047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,16,power_law_1.01,0.06789759993553161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,16,power_law_1.01,0.04618239998817444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,16,power_law_1.01,0.04674560129642487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,16,power_law_1.01,0.07861760258674622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,16,power_law_1.01,0.09934080243110657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,16,power_law_1.01,0.04717440009117126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,16,power_law_1.01,0.05000960230827332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,16,power_law_1.01,0.12773120403289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,16,power_law_1.01,0.06401280164718628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,16,power_law_1.01,0.04892799854278564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,16,power_law_1.01,0.07265920042991639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,16,balanced,1.1909386316935222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,16,power_law_1.01,0.158188796043396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,16,power_law_1.01,0.05464959740638733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,16,power_law_1.01,0.10273920297622681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,16,power_law_1.01,0.05601279735565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,16,power_law_1.01,0.1934399962425232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,16,power_law_1.01,0.13107839822769166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,16,power_law_1.01,0.06348800063133239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,16,power_law_1.01,0.3232959985733032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,16,power_law_1.01,0.07571840286254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,16,power_law_1.01,0.17111680507659913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,16,power_law_1.01,0.3639872074127197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,16,power_law_1.01,0.09446399807929992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,16,power_law_1.01,0.21709439754486085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,16,power_law_1.01,0.44681601524353026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,16,power_law_1.01,0.27162880897521974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,16,power_law_1.01,0.11200640201568604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,16,power_law_1.01,0.680083179473877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,16,power_law_1.01,0.43758721351623536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,16,power_law_1.01,0.14227839708328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,16,power_law_1.01,1.3219008445739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,16,power_law_1.01,0.8247679710388184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,16,power_law_1.01,0.18480000495910645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,16,power_law_1.01,0.2660032033920288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,16,power_law_1.01,0.32085120677948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,16,power_law_1.01,0.35681920051574706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,16,power_law_1.01,0.5639808177947998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,16,power_law_1.01,1.340505599975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,16,power_law_1.2,0.02930560111999512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,16,power_law_1.2,0.027353599667549133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,16,power_law_1.2,0.02712959945201874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,16,power_law_1.2,0.03061760067939758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,16,power_law_1.2,0.03115519881248474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,16,power_law_1.2,0.03611519932746887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,16,power_law_1.2,0.03451519906520843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,16,power_law_1.2,0.035776001214981076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,16,power_law_1.2,0.0383103996515274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,16,power_law_1.2,0.03904640078544617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,16,power_law_1.2,0.039366400241851805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,16,power_law_1.2,0.042438399791717527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,16,power_law_1.2,0.04223999977111816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,16,power_law_1.2,0.036294400691986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,16,power_law_1.2,0.045184001326560974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,16,power_law_1.2,0.049318400025367734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,16,power_law_1.2,0.029849600791931153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,16,power_law_1.2,0.05046399831771851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,16,power_law_1.2,0.029023998975753786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,16,power_law_1.2,0.05130239725112915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,16,power_law_1.2,0.03111039996147156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,16,power_law_1.2,0.05663359761238098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,16,power_law_1.2,0.031488001346588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,16,power_law_1.2,0.06576640009880066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,16,power_law_1.2,0.0333759993314743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,16,power_law_1.2,0.0710207998752594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,16,power_law_1.2,0.033580800890922545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,16,power_law_1.2,0.08514559864997864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,16,power_law_1.2,0.0342848002910614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,16,power_law_1.2,0.11585279703140258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,16,power_law_1.2,0.03578880131244659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,16,power_law_1.2,0.03615359961986542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,16,power_law_1.2,0.12749439477920532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,16,power_law_1.2,0.03819519877433777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,16,power_law_1.2,0.1590399980545044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,16,power_law_1.2,0.03969280123710632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,16,power_law_1.2,0.20378880500793456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,16,power_law_1.2,0.04062080085277557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,16,power_law_1.2,0.27550079822540285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,16,power_law_1.2,0.040575999021530154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,16,power_law_1.2,0.03957119882106781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,16,power_law_1.2,0.01825920045375824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,16,power_law_1.2,0.3718208074569702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,16,power_law_1.2,0.044409599900245664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,16,power_law_1.2,0.04828799962997436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,16,power_law_1.2,0.514086389541626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,16,power_law_1.2,0.01693439930677414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,16,power_law_1.2,0.03196159899234772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,16,power_law_1.2,0.04973439872264862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,16,power_law_1.2,0.018303999304771425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,16,power_law_1.2,0.031948798894882204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,16,power_law_1.2,0.8992959976196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,16,power_law_1.2,0.05494400262832642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,16,power_law_1.2,0.018643200397491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,16,power_law_1.2,0.033792001008987424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,16,power_law_1.2,0.06088320016860962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,16,power_law_1.2,0.03591679930686951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,16,power_law_1.2,0.019897599518299103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,16,power_law_1.2,1.627884864807129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,8,balanced,0.024693332612514496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,16,power_law_1.2,0.07173759937286377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,16,power_law_1.2,0.0361407995223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,16,power_law_1.2,0.021555200219154358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,16,power_law_1.2,0.08278399705886841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,16,power_law_1.2,0.03621760010719299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,16,power_law_1.2,0.024953599274158477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,16,power_law_1.2,0.11838079690933227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,8,balanced,0.023610666394233704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,16,power_law_1.2,0.036595198512077334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,16,power_law_1.2,0.024736000597476958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,16,power_law_1.2,0.13727359771728515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,16,power_law_1.2,0.03671680092811584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,8,balanced,0.02422400067249934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,16,power_law_1.2,0.035692799091339114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,16,power_law_1.2,0.20497279167175292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,16,power_law_1.2,0.0379584014415741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,8,balanced,0.02717866748571396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,16,power_law_1.2,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,16,power_law_1.2,0.038675200939178464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,8,balanced,0.028431999186674755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,16,power_law_1.2,0.25956480503082274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,16,power_law_1.2,0.03498879969120026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,16,power_law_1.2,0.038771200180053714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,8,balanced,0.031290667752424874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,16,power_law_1.2,0.0383296012878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,16,power_law_1.2,0.04021120071411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,16,power_law_1.2,0.35749120712280275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,8,balanced,0.029520000020662945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,16,power_law_1.2,0.041843199729919435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,16,power_law_1.2,0.03710080087184906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,8,balanced,0.02935466667016347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,16,power_law_1.2,0.04728319942951202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,16,power_law_1.2,0.511027193069458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,16,power_law_1.2,0.0363072007894516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,8,balanced,0.03160533308982849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,16,power_law_1.2,0.04782080054283142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,16,power_law_1.2,0.03658879995346069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,16,power_law_1.2,0.6175360202789306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,8,balanced,0.03309333324432373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,16,power_law_1.2,0.050425601005554196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,16,power_law_1.2,0.03585920035839081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,8,balanced,0.02918400118748347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,16,power_law_1.2,0.057548797130584715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,16,power_law_1.2,1.0013824462890626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,16,power_law_1.2,0.03586559891700745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,8,balanced,0.029359998802344005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,16,power_law_1.2,0.058963197469711306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,16,power_law_1.2,0.037196800112724304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,16,power_law_1.2,0.06917759776115417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,8,balanced,0.031189332405726116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,16,power_law_1.2,1.845350456237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,16,power_law_1.2,0.08206080198287964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,16,power_law_1.2,0.042931199073791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,8,balanced,0.03376533339420954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,8,balanced,0.029631999631722767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,8,balanced,0.0334346666932106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,16,power_law_1.2,0.09835519790649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,16,power_law_1.2,0.04774399995803833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,8,balanced,0.037615999579429626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,8,balanced,0.029418667157491047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,16,power_law_1.2,0.11744639873504639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,8,balanced,0.034858666360378265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,8,balanced,0.03086400032043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,16,power_law_1.2,0.053260797262191774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,16,power_law_1.2,0.18815360069274903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,8,balanced,0.040021332601706185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,8,balanced,0.037615999579429626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,8,balanced,0.03298133363326391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,16,power_law_1.2,0.24751360416412355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,8,balanced,0.03807999938726425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,8,balanced,0.035487999518712364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,16,power_law_1.2,0.06728320121765137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,16,power_law_1.2,0.30787200927734376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,8,balanced,0.04112533231576284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,8,balanced,0.04106133431196213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,8,balanced,0.036015999813874565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,16,power_law_1.2,0.4850751876831055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,8,balanced,0.04167999823888143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,16,power_law_1.2,0.08172159790992736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,8,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,8,balanced,0.0428959975639979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,16,power_law_1.2,0.5779327869415283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,8,balanced,0.04130133241415024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,8,balanced,0.017477333545684814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,8,balanced,0.03584533433119456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,16,power_law_1.2,0.10972800254821777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,16,power_law_1.2,0.9768768310546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,8,balanced,0.041706666350364685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,8,balanced,0.036848001182079315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,8,balanced,0.04580266773700714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,8,balanced,0.019167999426523846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,8,balanced,0.03761066744724909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,8,balanced,0.04174399872620901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,16,power_law_1.2,0.14035199880599974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,16,power_law_1.2,2.1645376205444338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,8,balanced,0.0377866675456365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,8,balanced,0.05695466697216034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,8,balanced,0.019178666174411774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,8,balanced,0.04187199970086416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,16,power_law_1.2,0.18467199802398682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,8,balanced,0.039493332306543984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,8,balanced,0.019093333433071773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,8,balanced,0.04350399971008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,16,power_law_1.2,0.25095040798187257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,8,balanced,0.0421973317861557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,8,balanced,0.061573331554730736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,8,balanced,0.021173333128293354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,8,balanced,0.04168533285458883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,16,power_law_1.2,0.31496319770812986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,8,balanced,0.023077333966890972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,8,balanced,0.04390400151411692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,8,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,16,power_law_1.2,0.5225024223327637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,8,balanced,0.02622399975856145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,8,balanced,0.06190933287143707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,8,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,8,balanced,0.045968001087506614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,8,balanced,0.024906667570273083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,16,power_law_1.2,1.0291328430175781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,8,balanced,0.04826133449872335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,8,balanced,0.04492799937725067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,8,balanced,0.029301332930723827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,8,balanced,0.05003733436266581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,8,balanced,0.04569066564242045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,8,balanced,0.07940266529719035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,8,balanced,0.02897600084543228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,8,balanced,0.05377600093682607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,8,balanced,0.04796266555786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,8,balanced,0.02733866622050603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,8,balanced,0.054154664278030396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,8,balanced,0.058186665177345276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,8,balanced,0.040218666195869446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,8,balanced,0.090229332447052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,8,balanced,0.03815466662247976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,8,balanced,0.06151466568311056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,8,balanced,0.06266133487224579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,8,balanced,0.03761066744724909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,8,balanced,0.04408533374468485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,8,balanced,0.0702453354994456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,8,balanced,0.06745066742102306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,8,balanced,0.04377066592375437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,8,balanced,0.10910933216412862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,8,balanced,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,8,balanced,0.08660266796747844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,8,balanced,0.08377066254615784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,8,balanced,0.044639999667803444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,8,balanced,0.0498933345079422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,8,balanced,0.10107733805974324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,8,balanced,0.10531733433405559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,8,balanced,0.057861333092053734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,8,balanced,0.13332800070444742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,8,balanced,0.06002133091290792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,8,balanced,0.1387999951839447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,8,balanced,0.1325759987036387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,8,balanced,0.07289066910743713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,8,balanced,0.17535465955734253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,8,balanced,0.08904000123341878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,8,balanced,0.1718826691309611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,8,balanced,0.17867734034856161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,8,balanced,0.12193066875139873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,8,balanced,0.22636799017588297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,8,balanced,0.20934933423995972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,8,balanced,0.23986667394638062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,8,balanced,0.14644267161687216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,8,balanced,0.20334933201471964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,8,balanced,0.2813226580619812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,8,balanced,0.29850133260091144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,8,balanced,0.25513599316279095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,8,balanced,0.26151466369628906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,8,balanced,0.3452213207880656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,8,balanced,0.3736319939295451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,8,balanced,0.3198293248812358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,8,balanced,0.38205333550771076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,8,balanced,0.5233920017878214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,8,balanced,0.5723946491877238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,8,power_law_1.01,0.027603200078010558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,8,power_law_1.01,0.02476799935102463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,8,power_law_1.01,0.022777600586414336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,8,power_law_1.01,0.02622720003128052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,8,power_law_1.01,0.02852480113506317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,8,power_law_1.01,0.03253119885921478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,8,balanced,0.4946613311767578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,8,power_law_1.01,0.03274880051612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,8,power_law_1.01,0.0321152001619339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,8,power_law_1.01,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,8,power_law_1.01,0.035436800122261046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,8,power_law_1.01,0.036601600050926206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,8,power_law_1.01,0.04092159867286682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,8,power_law_1.01,0.041920000314712526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,8,power_law_1.01,0.03473919928073883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,8,power_law_1.01,0.04100480079650879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,8,power_law_1.01,0.042777600884437564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,8,power_law_1.01,0.029574400186538695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,8,power_law_1.01,0.04668160080909729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,8,power_law_1.01,0.0353408008813858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,8,power_law_1.01,0.02810240089893341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,8,power_law_1.01,0.04784640073776245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,8,power_law_1.01,0.03017599880695343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,8,power_law_1.01,0.033529600501060484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,8,power_law_1.01,0.032262399792671204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,8,power_law_1.01,0.03553920090198517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,8,power_law_1.01,0.051596802473068235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,8,power_law_1.01,0.034329599142074584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,8,power_law_1.01,0.03687680065631867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,8,power_law_1.01,0.05797759890556335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,8,power_law_1.01,0.03472639918327332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,8,power_law_1.01,0.03978239893913269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,8,power_law_1.01,0.03500800132751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,8,power_law_1.01,0.06974719762802124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,8,power_law_1.01,0.0398144006729126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,8,power_law_1.01,0.03563520014286041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,8,power_law_1.01,0.039852800965309146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,8,power_law_1.01,0.03581439852714539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,8,power_law_1.01,0.07539839744567871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,8,power_law_1.01,0.04079999923706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,8,power_law_1.01,0.03808000087738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,8,power_law_1.01,0.04106239974498749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,8,power_law_1.01,0.07995520234107971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,8,power_law_1.01,0.038387200236320494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,8,power_law_1.01,0.041766399145126344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,8,power_law_1.01,0.09683200120925903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,8,power_law_1.01,0.03979519903659821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,8,power_law_1.01,0.04199680089950562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,8,power_law_1.01,0.10846079587936401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,8,power_law_1.01,0.04062080085277557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,8,power_law_1.01,0.043756800889968875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,8,power_law_1.01,0.0457152009010315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,8,power_law_1.01,0.13700480461120607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,8,power_law_1.01,0.044223999977111815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,8,power_law_1.01,0.05004799962043762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,8,power_law_1.01,0.045177599787712096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,8,power_law_1.01,0.172326397895813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,8,balanced,0.6967146396636963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,8,power_law_1.01,0.05256320238113403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,8,power_law_1.01,0.04973439872264862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,8,power_law_1.01,0.2568896055221558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,8,power_law_1.01,0.05545600056648255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,8,power_law_1.01,0.05589119791984558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,8,power_law_1.01,0.3088896036148071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,8,power_law_1.01,0.06255360245704651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,8,power_law_1.01,0.0654528021812439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,8,power_law_1.01,0.06415359973907471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,8,power_law_1.01,0.355679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,8,power_law_1.01,0.07412480115890503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,8,power_law_1.01,0.016780799627304076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,8,power_law_1.01,0.0778880000114441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,8,power_law_1.01,0.07950720191001892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,8,power_law_1.01,0.5937856197357178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,8,power_law_1.01,0.0878271996974945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,8,power_law_1.01,0.11240960359573364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,8,power_law_1.01,0.016838400065898894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,8,power_law_1.01,0.11694719791412353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,8,power_law_1.01,0.12896000146865844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,8,power_law_1.01,1.194752025604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,8,power_law_1.01,0.017472000420093538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,8,power_law_1.01,0.1355520009994507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,8,power_law_1.01,0.1965376019477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,8,power_law_1.01,0.01713919937610626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,8,power_law_1.01,0.16725120544433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,8,power_law_1.01,0.22615039348602295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,8,power_law_1.01,0.018751999735832213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,8,power_law_1.01,0.22927360534667968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,8,power_law_1.01,0.35114240646362305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,8,power_law_1.01,0.02019840031862259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,8,power_law_1.01,0.3331007957458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,8,power_law_1.01,0.021382400393486024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,8,power_law_1.01,0.44582400321960447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,8,balanced,1.0137333075205486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,8,power_law_1.01,0.42263040542602537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,8,balanced,1.1153226693471272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,8,power_law_1.01,0.02222079932689667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,8,power_law_1.01,0.5014400005340576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,8,power_law_1.01,0.5251520156860352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,8,power_law_1.01,0.025753599405288697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,8,power_law_1.01,0.8754431724548339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,8,power_law_1.01,0.025407999753952026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,8,power_law_1.01,0.9186047554016114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,8,power_law_1.01,0.025254398584365845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,8,power_law_1.01,1.6970304489135741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,8,power_law_1.01,0.036473599076271054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,8,power_law_1.01,1.8529216766357421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,8,power_law_1.01,0.03678080141544342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,8,power_law_1.01,0.03608959913253784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,8,power_law_1.01,0.039603200554847715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,8,power_law_1.01,0.0400191992521286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,8,power_law_1.01,0.03964160084724426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,8,balanced,0.9663466612497965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,8,power_law_1.01,0.04081920087337494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,8,power_law_1.01,0.04289279878139496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,8,power_law_1.01,0.04852479994297028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,8,power_law_1.01,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,8,power_law_1.01,0.060915201902389526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,8,power_law_1.01,0.0748416006565094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,8,power_law_1.01,0.09863680005073547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,8,power_law_1.01,0.11845120191574096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,8,power_law_1.01,0.16379519701004028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,8,power_law_1.01,0.21404800415039063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,8,power_law_1.01,0.25007359981536864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,8,power_law_1.01,0.37952001094818116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,8,power_law_1.01,0.7418496131896972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,8,power_law_1.2,0.027263998985290527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,8,power_law_1.2,0.02346239984035492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,8,power_law_1.2,0.023468799889087677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,8,power_law_1.2,0.02476799935102463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,8,power_law_1.2,0.031948798894882204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,8,power_law_1.2,0.031251201033592226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,8,power_law_1.2,0.03274880051612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,8,power_law_1.2,0.03473280072212219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,8,power_law_1.2,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,8,power_law_1.2,0.035046398639678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,8,power_law_1.2,0.03898240029811859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,8,power_law_1.2,0.041145598888397215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,8,power_law_1.2,0.035392001271247864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,8,power_law_1.2,0.04359680116176605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,8,power_law_1.2,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,8,power_law_1.2,0.02881920039653778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,8,power_law_1.2,0.04827519953250885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,8,power_law_1.2,0.02890239953994751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,8,power_law_1.2,0.049209600687026976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,8,power_law_1.2,0.030656000971794127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,8,power_law_1.2,0.050944000482559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,8,power_law_1.2,0.03196159899234772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,8,power_law_1.2,0.06139519810676575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,8,power_law_1.2,0.03526400029659271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,8,power_law_1.2,0.07089920043945312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,8,power_law_1.2,0.03537279963493347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,8,power_law_1.2,0.07719680070877075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,8,power_law_1.2,0.03578239977359772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,8,power_law_1.2,0.08257920145988465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,8,power_law_1.2,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,8,power_law_1.2,0.10446079969406127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,8,power_law_1.2,0.03615359961986542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,8,power_law_1.2,0.136735999584198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,8,power_law_1.2,0.03866879940032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,8,power_law_1.2,0.18148479461669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,8,power_law_1.2,0.03894400000572205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,8,power_law_1.2,0.21483519077301025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,8,power_law_1.2,0.03979519903659821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,8,power_law_1.2,0.31061758995056155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,8,power_law_1.2,0.042080000042915344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,8,power_law_1.2,0.045388799905776975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,8,power_law_1.2,0.3633536100387573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,8,power_law_1.2,0.04867840111255646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,8,power_law_1.2,0.49933438301086425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,8,power_law_1.2,0.05356159806251526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,8,power_law_1.2,0.8541888236999512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,8,power_law_1.2,0.059648001194000246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,8,power_law_1.2,0.06842240095138549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,8,power_law_1.2,0.0802944004535675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,8,power_law_1.2,1.994246482849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,8,power_law_1.2,0.0871616005897522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,8,power_law_1.2,0.1338368058204651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,8,power_law_1.2,0.04010240137577057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,4,balanced,0.03148266673088074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,4,balanced,0.02334933231274287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,8,power_law_1.2,0.14325120449066162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,8,power_law_1.2,0.01712000072002411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,4,power_law_1.01,0.0262719988822937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,8,power_law_1.2,0.0323199987411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,8,power_law_1.2,0.23441920280456544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,4,balanced,0.02319466571013133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,4,balanced,0.029466666281223297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,4,power_law_1.01,0.03273600041866302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,4,power_law_1.01,0.024723200500011443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,8,power_law_1.2,0.017056000232696534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,8,power_law_1.2,0.033030399680137636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,8,power_law_1.2,0.2954943895339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,4,balanced,0.025098666548728943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,4,balanced,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,4,power_law_1.01,0.023520000278949738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,4,power_law_1.01,0.028518399596214293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,8,power_law_1.2,0.37370240688323975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,8,power_law_1.2,0.01749120056629181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,8,power_law_1.2,0.03420799970626831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,4,power_law_1.01,0.025254398584365845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,4,balanced,0.025349333882331848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,4,balanced,0.0354720006386439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,4,power_law_1.01,0.029228800535202028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,4,power_law_1.01,0.028300800919532777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,4,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,4,power_law_1.01,0.032576000690460204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,8,power_law_1.2,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,8,power_law_1.2,0.5503168106079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,8,power_law_1.2,0.0173567995429039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,4,balanced,0.03782399992148081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,4,power_law_1.01,0.03322240114212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,4,power_law_1.01,0.0346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,4,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,4,power_law_1.01,0.03353599905967712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,8,power_law_1.2,0.03919999897480011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,4,balanced,0.039733332892258964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,4,power_law_1.01,0.037171199917793274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,8,power_law_1.2,0.018323199450969697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,8,power_law_1.2,0.6606783866882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,4,power_law_1.01,0.03427839875221252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,4,balanced,0.03422933320204417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,4,power_law_1.01,0.037887999415397645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,4,balanced,0.03985599925120672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,8,power_law_1.2,0.039827200770378116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,4,power_law_1.01,0.0356799989938736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,8,power_law_1.2,0.02048639953136444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,4,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,8,power_law_1.2,0.9254655838012695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,4,power_law_1.01,0.03839359879493713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,4,balanced,0.0358240008354187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,4,balanced,0.03978666663169861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,4,power_law_1.01,0.037324801087379456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,8,power_law_1.2,0.039827200770378116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,4,power_law_1.01,0.039468801021575926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,8,power_law_1.2,0.021747200191020964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,4,balanced,0.03551999976237615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,4,power_law_1.01,0.03966720104217529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,4,power_law_1.01,0.039680001139640805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,4,balanced,0.03961600114901861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,4,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,8,power_law_1.2,0.040479999780654904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,8,power_law_1.2,0.02176000028848648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,8,power_law_1.2,1.806617546081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,4,power_law_1.01,0.04249599874019623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,4,power_law_1.01,0.039136001467704774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,4,balanced,0.03548266738653183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,8,power_law_1.2,0.024486400187015533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,4,power_law_1.01,0.04423039853572845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,8,power_law_1.2,0.040729600191116336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,4,power_law_1.01,0.04276480078697205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,4,balanced,0.04046933352947235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,4,balanced,0.03757333258787791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,4,power_law_1.01,0.04743039906024933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,4,power_law_1.01,0.046489599347114566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,8,power_law_1.2,0.025209599733352663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,8,power_law_1.2,0.04136320054531097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,4,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,4,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,4,power_law_1.01,0.04769279956817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,4,power_law_1.01,0.05231999754905701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,8,power_law_1.2,0.02539519965648651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,4,power_law_1.01,0.05146880149841308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,8,power_law_1.2,0.04221439957618713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,4,balanced,0.04452266792456309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,4,balanced,0.04409066836039225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,4,power_law_1.01,0.05490559935569763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,4,power_law_1.01,0.05553920269012451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,8,power_law_1.2,0.03648639917373657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,4,power_law_1.01,0.05567359924316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,4,balanced,0.04691733419895172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,8,power_law_1.2,0.044300800561904906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,4,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,4,power_law_1.01,0.05713919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,4,power_law_1.01,0.06436480283737182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,8,power_law_1.2,0.03592320084571839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,4,power_law_1.01,0.06636160016059875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,8,power_law_1.2,0.04564479887485504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,4,balanced,0.04385599990685781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,4,balanced,0.047983999053637184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,4,power_law_1.01,0.0713536024093628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,4,power_law_1.01,0.07773439884185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,8,power_law_1.2,0.03601920008659363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,4,power_law_1.01,0.08116480112075805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,4,balanced,0.0521919975678126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,8,power_law_1.2,0.0516543984413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,4,power_law_1.01,0.08951039910316468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,4,balanced,0.05179200073083242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,4,power_law_1.01,0.09688320159912109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,4,power_law_1.01,0.10632319450378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,8,power_law_1.2,0.0385919988155365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,4,balanced,0.05193066596984863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,8,power_law_1.2,0.05443199872970581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,4,power_law_1.01,0.1158336043357849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,4,power_law_1.01,0.13697919845581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,8,power_law_1.2,0.04010240137577057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,4,power_law_1.01,0.147705602645874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,4,balanced,0.057328000664711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,4,power_law_1.01,0.1766975998878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,4,balanced,0.051781331499417625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,8,power_law_1.2,0.059059202671051025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,4,power_law_1.01,0.2001728057861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,8,power_law_1.2,0.03958399891853333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,4,power_law_1.01,0.23415040969848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,8,power_law_1.2,0.06550400257110596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,4,power_law_1.01,0.29792640209197996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,4,balanced,0.05635199944178263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,4,power_law_1.01,0.23238399028778076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,8,power_law_1.2,0.039987200498580934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,4,balanced,0.07390399773915608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,8,power_law_1.2,0.0424703985452652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,8,power_law_1.2,0.06823040246963501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,4,power_law_1.01,0.34742400646209715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,4,power_law_1.01,0.4546879768371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,4,balanced,0.06464533507823944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,8,power_law_1.2,0.0492031991481781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,4,power_law_1.01,0.45240960121154783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,4,power_law_1.01,0.6047103881835938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,8,power_law_1.2,0.08065919876098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,8,power_law_1.2,0.05262719988822937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,4,balanced,0.07552533348401387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,4,balanced,0.07706666489442189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,4,power_law_1.01,0.6930624008178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,8,power_law_1.2,0.09019520282745361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,4,power_law_1.01,0.5367231845855713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,8,power_law_1.2,0.06462720036506653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,8,power_law_1.2,0.13141119480133057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,4,power_law_1.01,1.1772031784057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,4,power_law_1.01,0.7744959831237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,8,power_law_1.2,0.0770687997341156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,8,power_law_1.2,0.1456063985824585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,4,balanced,0.08872000376383464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,8,power_law_1.2,0.10164480209350586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,8,power_law_1.2,0.2322240114212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,4,power_law_1.01,2.2413183212280274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,4,power_law_1.01,1.6922367095947266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,4,balanced,0.10078400373458862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,8,power_law_1.2,0.13036799430847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,8,power_law_1.2,0.266483211517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,8,power_law_1.2,0.17517440319061278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,8,power_law_1.2,0.39640960693359373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,4,balanced,0.1232319970925649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,8,power_law_1.2,0.21842560768127442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,8,power_law_1.2,0.550707197189331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,8,power_law_1.2,0.2666048049926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,4,balanced,0.11991999546686809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,8,power_law_1.2,0.7142911911010742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,8,power_law_1.2,0.42088961601257324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,4,balanced,0.14446933070818582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,8,power_law_1.2,1.089913558959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,8,power_law_1.2,0.8336704254150391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,8,power_law_1.2,2.412607955932617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,4,balanced,0.15196800231933594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,4,balanced,0.19688000281651816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,4,balanced,0.1856693426767985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,4,balanced,0.24635199705759683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,4,balanced,0.26053865750630695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,4,balanced,0.35042134920756024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,4,balanced,0.3294293284416199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,4,balanced,0.4572106599807739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,4,balanced,0.4087626536687215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,4,balanced,0.5765386819839478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,4,balanced,0.6159893274307251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,4,balanced,0.9003093242645264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,4,power_law_1.2,0.02627840042114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,4,power_law_1.2,0.024128000438213348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,4,power_law_1.2,0.024326400458812715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,4,power_law_1.2,0.025945600867271424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,4,power_law_1.2,0.027577599883079527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,4,power_law_1.2,0.034169599413871765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,4,power_law_1.2,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,4,power_law_1.2,0.032262399792671204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,4,power_law_1.2,0.0366784006357193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,4,power_law_1.2,0.03765760064125061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,4,power_law_1.2,0.028249600529670717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,4,power_law_1.2,0.038431999087333676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,4,power_law_1.2,0.028825598955154418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,4,power_law_1.2,0.03863680064678192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,4,power_law_1.2,0.031206399202346802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,4,power_law_1.2,0.04007039964199066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,4,power_law_1.2,0.04615679979324341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,4,power_law_1.2,0.033888000249862674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,4,power_law_1.2,0.047628799080848695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,4,power_law_1.2,0.03734399974346161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,4,power_law_1.2,0.05162879824638367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,4,power_law_1.2,0.03759360015392303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,4,power_law_1.2,0.058406400680541995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,4,power_law_1.2,0.03832319974899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,4,power_law_1.2,0.05885440111160278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,4,power_law_1.2,0.038553598523139956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,4,power_law_1.2,0.0643775999546051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,4,power_law_1.2,0.03966720104217529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,4,power_law_1.2,0.07120000123977661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,4,power_law_1.2,0.04209280014038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,4,power_law_1.2,0.08387839794158936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,4,power_law_1.2,0.04470399916172028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,4,balanced,1.1841973463694255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,4,power_law_1.2,0.09843199849128723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,4,power_law_1.2,0.04746240079402923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,4,balanced,1.7823626200358074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,4,power_law_1.2,0.12421120405197143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,4,power_law_1.2,0.04692479968070984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,4,power_law_1.2,0.15875200033187867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,4,power_law_1.2,0.051999998092651364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,8,balanced,0.020954666038354237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,4,power_law_1.2,0.05518720149993896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,4,power_law_1.2,0.18896000385284423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,8,balanced,0.02103466788927714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,4,power_law_1.2,0.2627392053604126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,4,power_law_1.2,0.058508801460266116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,8,balanced,0.0210506667693456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,4,power_law_1.2,0.3988735914230347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,4,power_law_1.2,0.06912639737129211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,8,balanced,0.019167999426523846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,4,power_law_1.2,0.5052800178527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,8,balanced,0.019904000063737232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,4,power_law_1.2,0.07986559867858886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,8,balanced,0.021327999730904896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,4,power_law_1.2,0.6840447902679443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,4,power_law_1.2,0.09701120257377624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,8,balanced,0.02720533311367035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,8,balanced,0.03737599899371465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,4,power_law_1.2,0.11226880550384521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,8,balanced,0.023285334308942158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,4,power_law_1.2,0.9063679695129394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,8,balanced,0.028864001234372456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,4,power_law_1.2,0.15635199546813966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,8,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,8,balanced,0.027488000690937042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,8,balanced,0.037674665451049805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,4,power_law_1.2,0.17726720571517945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,8,balanced,0.03323200096686681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,8,balanced,0.036943999429543815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,4,power_law_1.2,1.7902399063110352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,8,balanced,0.03773866593837738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,8,balanced,0.04207466542720795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,4,power_law_1.2,0.2768831968307495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,8,balanced,0.037717332442601524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,8,balanced,0.04138666639725367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,8,balanced,0.0393653338154157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,4,power_law_1.2,0.3060287952423096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,8,balanced,0.03549333413441976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,8,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,8,balanced,0.046154667933781944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,4,power_law_1.2,0.4837632179260254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,8,balanced,0.037690666814645134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,8,balanced,0.03977066775163015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,8,balanced,0.043509334325790405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,8,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,4,power_law_1.2,0.6996607780456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,8,balanced,0.04557866851488749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,8,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,8,balanced,0.05849599838256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,8,balanced,0.03950933367013931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,4,power_law_1.2,0.8743935585021972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,8,balanced,0.041349334021409355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,8,balanced,0.0671253353357315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,8,balanced,0.047728002071380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,8,balanced,0.08622399965922038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,8,balanced,0.04577599962552389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,4,power_law_1.2,1.4367039680480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,8,balanced,0.046384001771608986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,8,balanced,0.10763733585675557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,8,balanced,0.049642667174339294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,8,balanced,0.052042668064435325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,8,balanced,0.14799466729164124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,4,power_law_1.2,3.1674943923950196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,8,balanced,0.06238399942715963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,8,balanced,0.18066666523615518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,8,balanced,0.0643146683772405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,8,balanced,0.0836853285630544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,8,balanced,0.25623999039332074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,8,balanced,0.09698667128880818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,8,balanced,0.328277329603831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,8,balanced,0.12382933497428894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,8,balanced,0.14350933829943338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,8,balanced,0.4750186602274577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,8,balanced,0.1889280080795288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,8,balanced,0.2369920015335083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,8,balanced,0.6215893427530924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,8,balanced,0.2806826631228129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,8,balanced,0.7700053056081136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,8,balanced,0.4227840105692546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,8,balanced,1.2113333543141682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,8,balanced,0.7797813415527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,8,balanced,2.3934507369995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,8,power_law_1.01,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,8,power_law_1.01,0.04683519899845123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,8,power_law_1.01,0.04565759897232056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,8,power_law_1.01,0.048492801189422605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,8,power_law_1.01,0.047366398572921756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,8,power_law_1.01,0.04821760058403015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,8,power_law_1.01,0.04618239998817444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,8,power_law_1.01,0.04599680006504059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,8,power_law_1.01,0.04552319943904877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,8,power_law_1.01,0.019859200716018675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,8,power_law_1.01,0.044787201285362246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,8,power_law_1.01,0.04289920032024384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,8,power_law_1.01,0.019142399728298187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,8,power_law_1.01,0.04516479969024658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,8,power_law_1.01,0.053600001335144046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,8,power_law_1.01,0.019475199282169342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,8,power_law_1.01,0.049593600630760196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,8,power_law_1.01,0.018675200641155243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,8,power_law_1.01,0.055961602926254274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,8,power_law_1.01,0.019519999623298645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,8,power_law_1.01,0.05742719769477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,8,power_law_1.01,0.02019840031862259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,8,power_law_1.01,0.06282879710197449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,8,power_law_1.01,0.025811201333999632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,8,power_law_1.01,0.07099519968032837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,8,power_law_1.01,0.02319999933242798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.01,0.08080000281333924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,8,power_law_1.01,0.028070399165153505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.01,0.10759680271148682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,8,power_law_1.01,0.026208001375198364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.01,0.126528000831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,8,power_law_1.01,0.032646399736404416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.01,0.160806405544281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,8,power_law_1.01,0.04368639886379242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.01,0.21717119216918945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,8,power_law_1.01,0.041536000370979306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.01,0.29596800804138185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,8,power_law_1.01,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.01,0.3901823997497559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,8,power_law_1.01,0.044844800233840944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.01,0.5443264007568359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,8,power_law_1.01,0.042259201407432556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.01,0.755072021484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,8,power_law_1.01,0.043993601202964784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,8,power_law_1.01,0.0571008026599884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,8,power_law_1.01,0.8773311614990235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.01,0.06565759778022766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,8,power_law_1.01,1.3345151901245118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.01,0.0852288007736206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,8,power_law_1.01,2.8981760025024412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.01,0.10759680271148682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.01,0.14553600549697876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.01,0.1789247989654541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.01,0.25384960174560545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,8,balanced,0.033728001018365227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.01,0.32654719352722167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,8,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.01,0.47343997955322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,8,balanced,0.03395200024048487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.01,0.6217152118682862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,8,balanced,0.035349334279696144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,8,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,8,power_law_1.01,0.7674240112304688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,8,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,8,balanced,0.035301332672437034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,8,power_law_1.01,1.21081600189209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,8,balanced,0.03533333291610082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,8,power_law_1.01,2.3942079544067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,8,balanced,0.03628266602754593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,8,balanced,0.03572266548871994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,8,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,8,balanced,0.03722666700681051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,8,balanced,0.037418665985266365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,8,balanced,0.03802666564782461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,8,balanced,0.04553600152333578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,8,balanced,0.046069333950678505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,8,balanced,0.04375466704368591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,8,balanced,0.05578133463859558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,8,balanced,0.053685332338015236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,8,balanced,0.07048533360163371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,8,balanced,0.04563199977080027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,8,balanced,0.08250133196512859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,8,balanced,0.04417600234349569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,8,balanced,0.10749866565068562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,8,balanced,0.045552000403404236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,8,balanced,0.1246453324953715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,8,balanced,0.04567466676235199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,8,balanced,0.04553066690762838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,8,balanced,0.04584000011285146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,8,balanced,0.16168000300725302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,8,balanced,0.04578666885693868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,8,power_law_1.2,0.03059839904308319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,8,balanced,0.03740799923737844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,8,balanced,0.038021333515644073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,8,power_law_1.2,0.030745598673820495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,8,balanced,0.20123199621836343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,8,balanced,0.0376800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,8,power_law_1.2,0.030419200658798218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,8,balanced,0.0414986660083135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,8,balanced,0.03969600051641464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,8,power_law_1.2,0.031865599751472476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,8,balanced,0.03942399968703588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,8,power_law_1.2,0.03288959860801697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,8,balanced,0.2753653327624003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,8,balanced,0.043418665726979576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,8,power_law_1.2,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,8,balanced,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,8,power_law_1.2,0.03776000142097473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,8,power_law_1.2,0.03716480135917664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,8,balanced,0.051370665431022644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,8,power_law_1.2,0.03916800022125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,8,balanced,0.05197866757710775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,8,power_law_1.2,0.042335999011993405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,8,balanced,0.06185600161552429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,8,power_law_1.2,0.03758080005645752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,8,balanced,0.3399946689605713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,8,balanced,0.06367999811967213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,8,power_law_1.2,0.04688639938831329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,8,balanced,0.08020266890525818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,8,power_law_1.2,0.04668799936771393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,8,balanced,0.08066666622956593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,8,power_law_1.2,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,8,balanced,0.10292266805966695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,8,power_law_1.2,0.055180799961090085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,8,power_law_1.2,0.059487998485565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,8,balanced,0.4059520165125529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,8,balanced,0.11184533437093098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,8,power_law_1.2,0.06537600159645081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,8,power_law_1.2,0.07987840175628662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,8,balanced,0.14897066354751587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,8,power_law_1.2,0.08833280205726624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,8,power_law_1.2,0.11015039682388306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,8,power_law_1.2,0.13839999437332154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,8,balanced,0.17426133155822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,8,power_law_1.2,0.19125759601593018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,8,power_law_1.2,0.20785920619964598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,8,balanced,0.6216479937235514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,8,power_law_1.2,0.31242880821228025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,8,balanced,0.23567465941111246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,8,power_law_1.2,0.40093441009521485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,8,power_law_1.2,0.5768511772155762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,8,power_law_1.2,0.7022208213806153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,8,power_law_1.2,0.8580991744995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,8,balanced,0.2967946728070577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,8,power_law_1.2,1.28918399810791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,8,power_law_1.2,2.6102272033691407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,8,balanced,0.3463040192921956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,8,balanced,1.1764586766560872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,8,balanced,0.5109546581904093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,8,balanced,0.9546933174133301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,4,8,power_law_1.2,0.04650880098342895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,4,8,power_law_1.2,0.046911999583244324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,4,8,power_law_1.2,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,4,8,power_law_1.2,0.048204800486564635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,4,8,power_law_1.2,0.047007998824119566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,4,8,power_law_1.2,0.04795520007610321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,4,8,power_law_1.2,0.046316799521446225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,4,8,power_law_1.2,0.04560000002384186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,4,8,power_law_1.2,0.044998401403427125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,4,8,power_law_1.2,0.04408319890499115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,4,8,power_law_1.2,0.042316800355911253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,4,8,power_law_1.2,0.050393599271774295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,4,8,power_law_1.2,0.049497601389884946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,4,8,power_law_1.2,0.04805119931697845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,4,8,power_law_1.2,0.055289602279663085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,4,8,power_law_1.2,0.060236799716949466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,4,8,power_law_1.2,0.060652798414230345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,4,8,power_law_1.2,0.019980800151824952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,4,8,power_law_1.2,0.07338879704475403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.2,0.08455680012702942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.2,0.10871679782867431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,4,8,power_law_1.2,0.019731199741363524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.2,0.130950403213501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,4,8,power_law_1.2,0.01932159960269928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.2,0.1777791976928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,4,8,power_law_1.2,0.018400000035762788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.2,0.22047359943389894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,4,8,power_law_1.2,0.01943040043115616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.2,0.2893183946609497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,4,8,power_law_1.2,0.019833600521087645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,4,8,power_law_1.2,0.02593280076980591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.2,0.39112319946289065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,4,8,power_law_1.2,0.023193599283695222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.2,0.5672319889068603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,4,8,power_law_1.2,0.028697600960731505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.2,0.7308159828186035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,4,8,power_law_1.2,0.026118400692939758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,4,8,power_law_1.2,0.03271040022373199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,4,8,power_law_1.2,0.9160896301269531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,4,8,power_law_1.2,0.04346239864826203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,4,8,power_law_1.2,1.4028863906860352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,4,8,power_law_1.2,0.041465601325035094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,4,8,power_law_1.2,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,4,8,power_law_1.2,2.7935871124267577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,4,8,power_law_1.2,0.04490880072116852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,4,8,power_law_1.2,0.04215039908885956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,4,8,power_law_1.2,0.04414080083370209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,4,8,power_law_1.2,0.05713919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.2,0.0657151997089386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.2,0.08587520122528076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.2,0.1071936011314392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.2,0.1455296039581299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.2,0.178438401222229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.2,0.25408000946044923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.2,0.3256959915161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.2,0.47325439453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.2,0.6211264133453369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,4,8,power_law_1.2,0.7667263984680176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,4,8,power_law_1.2,1.2127296447753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,4,8,power_law_1.2,2.394700813293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,8,power_law_1.01,0.0440064013004303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,8,power_law_1.01,0.044300800561904906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,4,8,power_law_1.01,0.03391999900341034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,8,power_law_1.01,0.04133760035037994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,4,8,power_law_1.01,0.03325439989566803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,8,power_law_1.01,0.03561600148677826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,4,8,power_law_1.01,0.03249920010566711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,8,power_law_1.01,0.03562879860401154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,4,8,power_law_1.01,0.03317759931087494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,8,power_law_1.01,0.03547520041465759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,4,8,power_law_1.01,0.03306879997253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,8,power_law_1.01,0.04097920060157776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,4,8,power_law_1.01,0.034771201014518735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,8,power_law_1.01,0.0438975989818573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,4,8,power_law_1.01,0.03733119964599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,8,power_law_1.01,0.04737280011177063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,8,power_law_1.01,0.047366398572921756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,4,8,power_law_1.01,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,8,power_law_1.01,0.04766719937324524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,4,8,power_law_1.01,0.03994239866733551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,4,8,power_law_1.01,0.04216319918632507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,8,power_law_1.01,0.054655998945236206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,4,8,power_law_1.01,0.037555199861526486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,8,power_law_1.01,0.060787200927734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,4,8,power_law_1.01,0.04560000002384186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,8,power_law_1.01,0.06035199761390686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,4,8,power_law_1.01,0.04547840058803558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,8,power_law_1.01,0.07235199809074402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,4,8,power_law_1.01,0.047513601183891294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,8,power_law_1.01,0.07275519967079162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,4,8,power_law_1.01,0.05451520085334778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,8,power_law_1.01,0.07545599937438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,4,8,power_law_1.01,0.05571200251579285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,4,8,power_law_1.01,0.06182399988174438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,8,power_law_1.01,0.09683840274810791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,4,8,power_law_1.01,0.0788927972316742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,8,power_law_1.01,0.10959999561309815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,4,8,power_law_1.01,0.0875328004360199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,8,power_law_1.01,0.1312127947807312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,4,8,power_law_1.01,0.11997439861297607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,8,power_law_1.01,0.15969280004501343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,4,8,power_law_1.01,0.13352960348129272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,8,power_law_1.01,0.21683840751647948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,4,8,power_law_1.01,0.17438080310821533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,8,power_law_1.01,0.25408639907836916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,4,8,power_law_1.01,0.2088320016860962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,8,power_law_1.01,0.3617408037185669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,4,8,power_law_1.01,0.30028159618377687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,8,power_law_1.01,0.47208318710327146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,4,8,power_law_1.01,0.3878720045089722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,8,power_law_1.01,0.6837632179260253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,4,8,power_law_1.01,0.5643392086029053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,8,power_law_1.01,0.847430419921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,4,8,power_law_1.01,0.6950655937194824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,4,8,power_law_1.01,0.8299136161804199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,8,power_law_1.01,1.0596927642822265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,4,8,power_law_1.01,1.3261119842529296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,8,power_law_1.01,1.6850496292114259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,4,8,power_law_1.01,2.491967964172363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,8,power_law_1.01,3.0099647521972654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,4,8,power_law_1.2,0.04562560021877289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,4,8,power_law_1.2,0.04524799883365631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,4,8,power_law_1.2,0.04521600008010864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,4,8,power_law_1.2,0.045100799202919005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,4,8,power_law_1.2,0.04533120095729828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,4,8,power_law_1.2,0.05006719827651977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,4,8,power_law_1.2,0.05756800174713135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,4,8,power_law_1.2,0.0430400013923645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,4,8,power_law_1.2,0.0469760000705719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,4,8,power_law_1.2,0.053337597846984865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,4,8,power_law_1.2,0.0477183997631073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,4,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,4,8,power_law_1.2,0.0532800018787384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,4,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,4,8,power_law_1.2,0.05459840297698974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,4,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,4,8,power_law_1.2,0.05911679863929749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,4,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,4,8,power_law_1.2,0.0690559983253479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,4,balanced,0.019130667050679524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,4,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,4,8,power_law_1.2,0.07373440265655518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,4,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,4,balanced,0.01912533367673556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,4,8,power_law_1.2,0.07640960216522216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,4,balanced,0.038736000657081604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,4,balanced,0.01931200052301089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,4,8,power_law_1.2,0.09308159947395325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,4,balanced,0.03742400060097376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,4,balanced,0.021029333273569744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,4,8,power_law_1.2,0.11016319990158081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,4,balanced,0.037530665596326195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,4,balanced,0.021007999777793884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,4,balanced,0.039450667798519135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,4,8,power_law_1.2,0.13550080060958863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,4,balanced,0.02094399929046631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,4,balanced,0.04067733387152354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,4,balanced,0.039808000127474465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,4,8,power_law_1.2,0.15823999643325806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,4,balanced,0.02537599951028824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,4,balanced,0.03942933430274328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,4,8,power_law_1.2,0.1965376019477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,4,balanced,0.02334933231274287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,4,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,4,8,power_law_1.2,0.26856958866119385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,4,balanced,0.041519999504089355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,4,balanced,0.04773333172003428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,4,balanced,0.04178133110205332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,4,8,power_law_1.2,0.3854912042617798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,4,balanced,0.04553600152333578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,4,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,4,balanced,0.041562666495641075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,4,8,power_law_1.2,0.4892864227294922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,4,balanced,0.05183466772238413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,4,balanced,0.036464000741640724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,4,8,power_law_1.2,0.7652224063873291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,4,balanced,0.05557866891225179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,4,balanced,0.035429333647092186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,4,8,power_law_1.2,0.8833087921142578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,4,balanced,0.06410666803518932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,4,balanced,0.0334346666932106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,4,8,power_law_1.2,1.0931520462036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,4,balanced,0.06820266445477803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,4,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,4,8,power_law_1.2,1.6247615814208984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,4,balanced,0.03495999922355016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,4,balanced,0.09123733639717102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,4,balanced,0.035317334036032356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,4,8,power_law_1.2,3.303148651123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,4,balanced,0.11159466703732808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,4,balanced,0.04769066472848257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,4,balanced,0.049413333336512245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,4,balanced,0.14577066898345947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,4,balanced,0.06604266663392384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,4,balanced,0.17248000701268515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,4,balanced,0.07801066835721333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,4,balanced,0.09929600358009338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,4,balanced,0.229802668094635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,4,balanced,0.1227946678797404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,4,balanced,0.16738667090733847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,4,balanced,0.28755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,4,balanced,0.21079466740290323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,4,balanced,0.34537065029144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,4,balanced,0.29710400104522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,4,balanced,0.38514665762583417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,4,balanced,0.5171039899190267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,4,balanced,0.47327999273935956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,4,balanced,0.7382559776306152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,4,balanced,0.9688639640808105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,4,power_law_1.01,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,4,power_law_1.01,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,4,power_law_1.01,0.03357439935207367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,4,balanced,1.442394733428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,4,power_law_1.01,0.03573760092258453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,4,power_law_1.01,0.035743999481201175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,4,power_law_1.01,0.03671680092811584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,4,power_law_1.01,0.03639039993286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,4,power_law_1.01,0.03723520040512085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,4,power_law_1.01,0.03888640105724335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,4,power_law_1.01,0.03811840116977692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,4,power_law_1.01,0.03863680064678192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,4,power_law_1.01,0.04019840061664581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,4,power_law_1.01,0.03984639942646027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,4,power_law_1.01,0.04206080138683319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,4,power_law_1.01,0.0500927984714508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,4,power_law_1.01,0.05219200253486633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,4,power_law_1.01,0.05503360033035278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,4,power_law_1.01,0.0642624020576477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.01,0.07151359915733338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.01,0.08440319895744323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.01,0.10739840269088745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.01,0.14507520198822021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.01,0.17590399980545043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.01,0.21751680374145507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.01,0.2398144006729126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.01,0.45569281578063964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.01,0.5583744049072266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,4,power_law_1.01,0.7617216110229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,4,power_law_1.01,1.1126591682434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,4,power_law_1.01,1.7900415420532227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,4,power_law_1.01,0.01854719966650009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,4,power_law_1.01,0.01820800006389618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,4,power_law_1.01,0.018105599284172057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,4,power_law_1.01,0.018348799645900728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,4,power_law_1.01,0.017849600315093993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,4,power_law_1.01,0.018675200641155243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,4,power_law_1.01,0.021913599967956544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,4,power_law_1.01,0.02162559926509857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,4,power_law_1.01,0.03927040100097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,4,power_law_1.01,0.03918080031871796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,4,power_law_1.01,0.03948799967765808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,4,power_law_1.01,0.035104000568389894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,4,power_law_1.01,0.03340159952640533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,4,power_law_1.01,0.031385600566864014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,4,power_law_1.01,0.03701759874820709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,4,power_law_1.01,0.033055999875068666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,4,power_law_1.01,0.03472639918327332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,4,power_law_1.01,0.04604159891605377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.01,0.04819200038909912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.01,0.06267520189285278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.01,0.0748799979686737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.01,0.09715200066566468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,4,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.01,0.11985280513763427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,4,balanced,0.033701332906881966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.01,0.1641343951225281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,4,balanced,0.03580799947182337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.01,0.2063040018081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,4,balanced,0.03411199897527695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.01,0.29192318916320803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,4,balanced,0.03577066709597906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.01,0.38138880729675295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,4,balanced,0.03711466739575068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,4,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,4,power_law_1.01,0.4698495864868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,4,balanced,0.03571200122435888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,4,balanced,0.03775999943415324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,4,power_law_1.01,0.7343679904937744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,4,balanced,0.031685332457224526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,4,balanced,0.03578133384386698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,4,balanced,0.035631999373435974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,4,power_law_1.01,1.442751979827881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,4,balanced,0.036474667489528656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,4,balanced,0.03404266635576884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,4,balanced,0.037920000652472176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,4,balanced,0.03572266548871994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,4,balanced,0.03868266691764196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,4,balanced,0.03549333413441976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,4,balanced,0.039749334255854286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,4,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,4,balanced,0.0417546679576238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,4,balanced,0.050000001986821495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,4,balanced,0.03669866671164831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,4,balanced,0.049466664592425026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,4,balanced,0.03738133360942205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,4,power_law_1.2,0.047577598690986635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,4,balanced,0.055498664577802025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,4,balanced,0.03623999903599421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,4,balanced,0.06234666705131531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,4,balanced,0.03729599962631861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,4,power_law_1.2,0.0331712007522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,4,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,4,balanced,0.062421331803003945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,4,power_law_1.2,0.03298560082912445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,4,balanced,0.037530665596326195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,4,balanced,0.08065600196520488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,4,power_law_1.2,0.03402239978313446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,4,balanced,0.03948266555865606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,4,power_law_1.2,0.03480960130691528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,4,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,4,balanced,0.07867733140786488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,4,power_law_1.2,0.03542400002479553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,4,balanced,0.047695999344189964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,4,power_law_1.2,0.039904001355171206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,4,balanced,0.10802132884661357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,4,balanced,0.047600001096725464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,4,power_law_1.2,0.04322560131549835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,4,balanced,0.0581226646900177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,4,balanced,0.12136000394821167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,4,power_law_1.2,0.04658559858798981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,4,balanced,0.05817066629727682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,4,power_law_1.2,0.04776319861412048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,4,balanced,0.0786240001519521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,4,power_law_1.2,0.04792320132255554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,4,balanced,0.15681599577267966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,4,balanced,0.08755200107892354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,4,power_law_1.2,0.04813440144062042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,4,power_law_1.2,0.05138559937477112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,4,balanced,0.11339199542999268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,4,power_law_1.2,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,4,balanced,0.19311465819676718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,4,balanced,0.1358453333377838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,4,power_law_1.2,0.06299520134925843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,4,power_law_1.2,0.03139199912548065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,4,power_law_1.2,0.0654528021812439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,4,power_law_1.2,0.0657920002937317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,4,balanced,0.18246400356292725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,4,power_law_1.2,0.03187200129032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,4,power_law_1.2,0.07921919822692872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,4,power_law_1.2,0.03201920092105866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,4,balanced,0.2534880042076111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,4,power_law_1.2,0.0977728009223938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,4,power_law_1.2,0.03289600014686585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,4,power_law_1.2,0.11889280080795288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,4,balanced,0.21792000532150269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,4,power_law_1.2,0.032902398705482484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,4,power_law_1.2,0.13723520040512086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,4,power_law_1.2,0.1849727988243103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,4,power_law_1.2,0.034585601091384886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,4,power_law_1.2,0.20632319450378417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,4,power_law_1.2,0.03654400110244751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,4,balanced,0.3087573250134786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,4,power_law_1.2,0.3364351987838745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,4,power_law_1.2,0.03591040074825287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,4,balanced,0.30565333366394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,4,power_law_1.2,0.36004478931427003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,4,power_law_1.2,0.03680639863014221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,4,power_law_1.2,0.03713920116424561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,4,power_law_1.2,0.6194240093231201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,4,power_law_1.2,0.03854719996452331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,4,power_law_1.2,0.6428736209869385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,4,power_law_1.2,0.039801600575447085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,4,power_law_1.2,0.9149439811706543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,4,balanced,0.3595999876658122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,4,power_law_1.2,0.04345600008964538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,4,power_law_1.2,1.4071104049682617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,4,balanced,0.3911199967066447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,4,power_law_1.2,0.044300800561904906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,4,power_law_1.2,2.485638427734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,4,power_law_1.2,0.056531202793121335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,4,power_law_1.2,0.05846400260925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,4,power_law_1.2,0.06398720145225525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,4,power_law_1.2,0.0765824019908905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,4,power_law_1.2,0.08863999843597412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,4,balanced,0.4712373415629069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,4,power_law_1.2,0.11882239580154419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,4,power_law_1.2,0.12867200374603271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,4,balanced,0.5429973204930624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,4,power_law_1.2,0.16567039489746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,4,power_law_1.2,0.2066943883895874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,4,power_law_1.2,0.2736448049545288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,4,power_law_1.2,0.03592959940433502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,4,power_law_1.2,0.41077117919921874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,4,balanced,0.7227839628855387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,4,power_law_1.2,0.6540671825408936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,4,power_law_1.2,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,4,power_law_1.2,0.704531192779541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,4,power_law_1.2,0.033529600501060484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,4,power_law_1.2,0.9021632194519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,4,power_law_1.2,0.035667198896408084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,4,power_law_1.2,1.532249641418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,4,power_law_1.2,0.03568640053272247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,4,power_law_1.2,0.03656319975852966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,4,power_law_1.2,3.0976768493652345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,4,power_law_1.2,0.03663359880447388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,4,power_law_1.2,0.036671999096870425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,4,power_law_1.2,0.03839359879493713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,4,power_law_1.2,0.037651199102401736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,4,balanced,1.0079572995503743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,4,power_law_1.2,0.039263999462127684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,4,power_law_1.2,0.039776000380516055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,4,power_law_1.2,0.04166400134563446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,4,power_law_1.2,0.043424001336097716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,4,balanced,1.3681492805480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,4,power_law_1.2,0.05027840137481689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,4,power_law_1.2,0.0547327995300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,4,power_law_1.01,0.048179200291633605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,4,power_law_1.2,0.05372160077095032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,4,power_law_1.2,0.06289280056953431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,4,power_law_1.01,0.034560000896453856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.2,0.0648576021194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.2,0.08782079815864563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,4,power_law_1.01,0.03373439908027649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.2,0.10944639444351197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,4,power_law_1.01,0.0341376006603241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,4,power_law_1.01,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.2,0.15808639526367188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,4,power_law_1.01,0.034969601035118106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.2,0.15779839754104613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,4,power_law_1.01,0.03845120072364807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.2,0.25649280548095704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.2,0.3050112009048462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,4,power_law_1.01,0.04191359877586365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.2,0.5009984016418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,4,power_law_1.01,0.04419200122356415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,4,power_law_1.01,0.045817598700523376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.2,0.7153535842895508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,4,power_law_1.01,0.048851200938224794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,4,power_law_1.2,0.7357439994812012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,4,power_law_1.2,1.1478015899658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,4,power_law_1.01,0.04856959879398346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,4,power_law_1.01,0.05195519924163818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,4,power_law_1.2,2.1451711654663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,4,power_law_1.01,0.05428479909896851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,4,power_law_1.01,0.0636352002620697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,4,power_law_1.01,0.06862080097198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,4,power_law_1.01,0.07470080256462097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,4,power_law_1.01,0.08134400248527526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,4,power_law_1.01,0.08472959995269776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,4,power_law_1.01,0.11420799493789673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,4,power_law_1.01,0.14103039503097534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,4,power_law_1.01,0.1628224015235901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,4,power_law_1.01,0.21468799114227294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,4,power_law_1.01,0.29278719425201416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,4,power_law_1.01,0.03258239924907684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,4,power_law_1.01,0.3636928081512451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,4,power_law_1.01,0.03349120020866394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,4,power_law_1.01,0.3951936006546021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,4,power_law_1.01,0.03269760012626648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,4,power_law_1.2,0.018617600202560425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,4,power_law_1.01,0.645139217376709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,4,power_law_1.01,0.03275519907474518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,4,power_law_1.2,0.018367999792099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,4,power_law_1.01,0.8140352249145508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,4,power_law_1.2,0.018355199694633485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,4,power_law_1.01,0.03234559893608093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,4,power_law_1.01,1.3318079948425292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,4,power_law_1.2,0.018111999332904815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,4,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,4,power_law_1.2,0.01822720021009445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,4,power_law_1.01,2.3246015548706054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,4,power_law_1.01,0.034944000840187076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,4,power_law_1.2,0.01865600049495697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,4,power_law_1.2,0.02178560048341751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,4,power_law_1.01,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,4,power_law_1.2,0.021766400337219237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,4,power_law_1.01,0.03672960102558136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,4,power_law_1.2,0.03945600092411041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,4,power_law_1.01,0.03945600092411041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,4,power_law_1.2,0.0392767995595932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,4,power_law_1.01,0.03940480053424835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,4,power_law_1.2,0.039526399970054624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,4,power_law_1.01,0.04044159948825836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,4,power_law_1.2,0.035071998834609985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,4,power_law_1.2,0.03363839983940124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,4,power_law_1.01,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,4,power_law_1.2,0.031200000643730165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,4,power_law_1.01,0.046214398741722104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,4,power_law_1.2,0.03686400055885315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,4,power_law_1.01,0.055193597078323366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,4,power_law_1.2,0.03293440043926239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,4,power_law_1.2,0.034764799475669864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,4,power_law_1.01,0.06170880198478699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,4,power_law_1.2,0.04599680006504059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,4,power_law_1.01,0.061363202333450315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.2,0.048044800758361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,4,power_law_1.01,0.07031040191650391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.2,0.06217600107192993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.2,0.07505279779434204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,4,power_law_1.01,0.08352640271186829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.2,0.09708799719810486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,4,power_law_1.01,0.10295039415359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.2,0.12005120515823364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,4,power_law_1.01,0.12337919473648071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.2,0.163372802734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,4,power_law_1.01,0.17686400413513184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.2,0.20663039684295653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,4,power_law_1.01,0.20727040767669677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.2,0.292576003074646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.2,0.3815743923187256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,4,power_law_1.01,0.32632319927215575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,4,power_law_1.2,0.4695295810699463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,4,power_law_1.01,0.3920576095581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,4,power_law_1.2,0.7363967895507812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,4,power_law_1.01,0.5792255878448487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,4,power_law_1.2,1.441420841217041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,4,power_law_1.01,0.6893119812011719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,4,power_law_1.01,0.8288384437561035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,4,power_law_1.01,1.144819164276123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,4,power_law_1.01,2.804627227783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,2,balanced,0.035349334279696144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,2,balanced,0.03580799947182337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,2,balanced,0.03737599899371465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,2,balanced,0.019071999937295914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,2,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,2,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,2,balanced,0.019088000059127808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,2,balanced,0.03740799923737844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,2,balanced,0.018965333700180054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,2,balanced,0.03941333293914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,2,balanced,0.01926933353145917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,2,balanced,0.03753600021203359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,2,balanced,0.019215999792019527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,2,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,2,balanced,0.03958933303753535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,2,balanced,0.019205333044131596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,2,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,2,balanced,0.023050665855407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,2,balanced,0.03948266555865606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,2,balanced,0.021066665649414062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,2,balanced,0.03956799954175949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,2,balanced,0.0230880007147789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,2,balanced,0.04146133363246918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,2,balanced,0.023365333676338196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,2,balanced,0.047728002071380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,2,balanced,0.022991999983787537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,2,balanced,0.04764799773693085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,2,balanced,0.02514133354028066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,2,balanced,0.05002133548259735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,2,balanced,0.029317334294319153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,2,balanced,0.05630399783452352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,2,balanced,0.025978667040665943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,2,balanced,0.06241066753864288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,2,balanced,0.02717333287000656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,2,balanced,0.03025600065787633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,2,balanced,0.072202667593956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,2,balanced,0.03526933242877325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,2,balanced,0.08180266618728638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,2,balanced,0.03973866750796636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,2,balanced,0.11140799522399902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,2,balanced,0.04186666508515676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,2,balanced,0.12999999523162842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,2,balanced,0.05579199890295664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,2,power_law_1.01,0.03285120129585266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,2,power_law_1.01,0.01783040016889572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,2,balanced,0.06191466748714447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,2,balanced,0.17046932379404703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,2,power_law_1.01,0.03252480030059814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,2,balanced,0.08105066418647766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,2,power_law_1.01,0.032902398705482484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,2,power_law_1.01,0.017907199263572694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,2,power_law_1.01,0.03566080033779144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,2,power_law_1.01,0.0172992005944252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,2,balanced,0.2121386726697286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,2,balanced,0.09707732995351155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,2,power_law_1.01,0.03604480028152466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,2,power_law_1.01,0.018131199479103088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,2,power_law_1.01,0.036339199542999266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,2,power_law_1.01,0.018105599284172057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,2,power_law_1.01,0.03734399974346161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,2,balanced,0.12779733538627625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,2,power_law_1.01,0.018688000738620758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,2,balanced,0.30060267448425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,2,power_law_1.01,0.03761920034885406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,2,power_law_1.01,0.01908479928970337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,2,power_law_1.01,0.03797119855880737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,2,power_law_1.01,0.019545599818229675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,2,balanced,0.1613653302192688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,2,power_law_1.01,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,2,power_law_1.01,0.0215488001704216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,2,power_law_1.01,0.03986560106277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,2,power_law_1.01,0.021356800198554994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,2,balanced,0.3879520098368327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,2,power_law_1.01,0.04040960073471069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,2,power_law_1.01,0.02220800071954727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,2,power_law_1.01,0.04219520092010498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,2,balanced,0.2241386572519938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,2,power_law_1.01,0.040294399857521056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,2,power_law_1.01,0.043270400166511534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,2,power_law_1.01,0.04014720022678375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,2,power_law_1.01,0.049318400025367734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,2,power_law_1.01,0.04023680090904236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,2,power_law_1.01,0.051660799980163576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,2,power_law_1.01,0.03313280045986176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,2,balanced,0.4739946524302165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,2,power_law_1.01,0.05299839973449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,2,power_law_1.01,0.03252480030059814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,2,balanced,0.288266658782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,2,power_law_1.01,0.06175360083580017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,2,power_law_1.01,0.033164799213409424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,2,power_law_1.01,0.03875199854373932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.01,0.0658240020275116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.01,0.04084480106830597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.01,0.09117439985275269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.01,0.053491199016571046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,2,balanced,0.35282135009765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.01,0.11078399419784546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.01,0.05950719714164734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.01,0.12124799489974976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.01,0.07927039861679078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,2,balanced,0.7300000190734863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.01,0.1527232050895691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.01,0.09534080028533935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.01,0.20616960525512695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.01,0.12521599531173705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.01,0.26537599563598635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.01,0.15752320289611815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.01,0.39587199687957764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.01,0.22067201137542725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.01,0.48277759552001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,2,balanced,0.5448319911956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.01,0.2846719980239868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,2,power_law_1.01,0.6037119865417481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,2,power_law_1.01,0.34844160079956055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,2,power_law_1.01,1.0209792137145997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,2,power_law_1.01,0.5414527893066406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,2,power_law_1.01,1.9664703369140626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,2,power_law_1.01,1.0544063568115234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,2,balanced,1.4236960411071777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,2,balanced,1.0563039779663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,2,balanced,0.03151999910672506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,2,balanced,0.03482133398453394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,2,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,2,balanced,0.03532266616821289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,2,balanced,0.0356480007370313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,2,balanced,0.037317333122094475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,2,balanced,0.03740799923737844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,2,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,2,balanced,0.037402667105197906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,2,balanced,0.03734933336575826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,2,balanced,0.03933866570393244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,2,balanced,0.032325332363446556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,2,balanced,0.039247999588648476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,2,power_law_1.2,0.030425599217414855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,2,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,2,balanced,0.043568000197410583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,2,power_law_1.2,0.03322240114212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,2,balanced,0.050527999798456825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,2,balanced,0.031685332457224526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,2,power_law_1.2,0.031123200058937074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,2,balanced,0.052928000688552856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,2,balanced,0.03374933451414108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,2,power_law_1.2,0.03246079981327057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,2,power_law_1.2,0.0302592009305954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,2,balanced,0.03390933324893316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,2,balanced,0.05399466554323832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,2,power_law_1.2,0.03276160061359405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,2,power_law_1.2,0.03381119966506958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,2,balanced,0.03416533271471659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,2,power_law_1.2,0.03529599905014038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,2,balanced,0.06670933465162913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,2,power_law_1.2,0.034380799531936644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,2,balanced,0.03573333223660787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,2,power_law_1.2,0.035436800122261046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,2,balanced,0.06833600004514058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,2,balanced,0.03452266752719879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,2,power_law_1.2,0.0352512001991272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,2,power_law_1.2,0.03680639863014221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,2,balanced,0.08475200335184734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,2,balanced,0.035760000348091125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,2,power_law_1.2,0.039468801021575926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,2,power_law_1.2,0.017657600343227386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,2,power_law_1.2,0.03694719970226288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,2,balanced,0.03801066676775614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,2,power_law_1.2,0.040012800693511964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,2,balanced,0.09331199526786804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,2,power_law_1.2,0.037478399276733396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,2,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,2,power_law_1.2,0.04444800019264221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,2,power_law_1.2,0.03743999898433685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,2,power_law_1.2,0.017875200510025023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,2,balanced,0.037834666669368744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,2,balanced,0.11757866541544597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,2,power_law_1.2,0.03831680119037628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,2,power_law_1.2,0.043609601259231565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,2,balanced,0.03774933268626531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,2,power_law_1.2,0.018035200238227845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,2,power_law_1.2,0.03898879885673523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,2,power_law_1.2,0.04271360039710999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,2,balanced,0.03977066775163015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,2,power_law_1.2,0.018329599499702455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,2,balanced,0.14249599973360697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,2,power_law_1.2,0.039987200498580934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,2,power_law_1.2,0.04876160025596619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,2,balanced,0.039706667264302574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,2,power_law_1.2,0.018240000307559966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,2,power_law_1.2,0.04314239919185638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,2,balanced,0.046725332736968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,2,power_law_1.2,0.047142401337623596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,2,power_law_1.2,0.018662400543689728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,2,power_law_1.2,0.044147199392318724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,2,balanced,0.04804266492525736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,2,balanced,0.18699200948079428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,2,power_law_1.2,0.0505407989025116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,2,power_law_1.2,0.019679999351501463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,2,power_law_1.2,0.04932479858398438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,2,power_law_1.2,0.05931519865989685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,2,balanced,0.04990399877230326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,2,power_law_1.2,0.019564799964427948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,2,power_law_1.2,0.050732797384262084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,2,power_law_1.2,0.06162559986114502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,2,balanced,0.06035199761390686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,2,power_law_1.2,0.05441920161247253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,2,power_law_1.2,0.02181120067834854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,2,balanced,0.22577067216237387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,2,power_law_1.2,0.06462720036506653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,2,power_law_1.2,0.06371840238571166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,2,balanced,0.0626453310251236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,2,power_law_1.2,0.02192640006542206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.2,0.0714303970336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,2,power_law_1.2,0.07571840286254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,2,power_law_1.2,0.0226623997092247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,2,balanced,0.085999995470047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.2,0.08648319840431214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,2,power_law_1.2,0.08492799997329711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,2,power_law_1.2,0.04021120071411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,2,balanced,0.09968533118565877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.2,0.10580480098724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,2,power_law_1.2,0.0967423975467682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,2,balanced,0.31194667021433514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,2,power_law_1.2,0.04038400053977966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.2,0.14575999975204468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,2,power_law_1.2,0.1196671962738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,2,balanced,0.1360319952170054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,2,power_law_1.2,0.04081279933452606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.2,0.16255359649658202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,2,power_law_1.2,0.15507199764251708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,2,power_law_1.2,0.03343360126018524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.2,0.21853439807891845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,2,power_law_1.2,0.19891200065612794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,2,balanced,0.15146666765213013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,2,power_law_1.2,0.03226880133152008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.2,0.3257791996002197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,2,power_law_1.2,0.2534336090087891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,2,power_law_1.2,0.03298560082912445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,2,balanced,0.3867786725362142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.2,0.42328319549560545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,2,balanced,0.21187732617060342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,2,power_law_1.2,0.36285440921783446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,2,power_law_1.2,0.03863680064678192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.2,0.4938816070556641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,2,power_law_1.2,0.4348927974700928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.2,0.040755200386047366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,2,power_law_1.2,0.6931968212127686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,2,balanced,0.2653119961420695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,2,power_law_1.2,0.6085504055023193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.2,0.0541055977344513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,2,power_law_1.2,0.9752127647399902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,2,power_law_1.2,0.6167103767395019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.2,0.059462398290634155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,2,balanced,0.4570986827214559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,2,power_law_1.2,2.1883007049560548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.2,0.07961599826812744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,2,power_law_1.2,1.0619903564453126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,2,balanced,0.3752853473027547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.2,0.0954688012599945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,2,power_law_1.2,2.1830272674560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.2,0.12477439641952515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.2,0.15839999914169312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,2,balanced,0.4969173272450765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.2,0.22125439643859862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,2,balanced,0.7019893328348795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,2,power_law_1.01,0.03218559920787811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.2,0.2863296031951904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,2,power_law_1.2,0.34973440170288084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,2,power_law_1.01,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,2,balanced,0.61080535252889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,2,power_law_1.2,0.5414207935333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,2,power_law_1.01,0.03290880024433136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,2,power_law_1.2,1.0554752349853516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,2,power_law_1.01,0.034822401404380796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,2,power_law_1.01,0.03150080144405365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,2,power_law_1.01,0.033939200639724734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,2,power_law_1.01,0.031615999341011045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,2,power_law_1.01,0.03443840146064758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,2,power_law_1.01,0.031302401423454286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,2,balanced,0.9301813443501791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,2,power_law_1.01,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,2,power_law_1.01,0.033081600069999696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,2,balanced,1.3234879970550537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,2,power_law_1.01,0.03934719860553741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,2,power_law_1.01,0.033081600069999696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,2,power_law_1.01,0.04439679980278015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,2,power_law_1.01,0.03464959859848023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,2,power_law_1.01,0.0450111985206604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,2,power_law_1.01,0.0364544004201889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,2,power_law_1.01,0.04243200123310089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,2,power_law_1.01,0.046374401450157164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,2,power_law_1.01,0.03544960021972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,2,power_law_1.01,0.05002880096435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,2,power_law_1.01,0.03694719970226288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,2,balanced,1.7910559972127278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,2,power_law_1.01,0.0513152003288269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,2,power_law_1.01,0.03834240138530731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,2,power_law_1.01,0.06146559715270996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,2,power_law_1.01,0.037785598635673524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,2,power_law_1.01,0.04037759900093078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,2,power_law_1.01,0.063155198097229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,2,power_law_1.01,0.04007039964199066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,2,power_law_1.01,0.06543999910354614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,2,power_law_1.01,0.0426367998123169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,2,power_law_1.01,0.05021439790725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,2,power_law_1.01,0.07056000232696533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,2,power_law_1.01,0.05655679702758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,2,power_law_1.01,0.08083199858665466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,2,power_law_1.01,0.05537279844284058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,2,power_law_1.01,0.09971839785575867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,2,power_law_1.01,0.07432960271835327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,2,power_law_1.01,0.11912959814071655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,2,power_law_1.01,0.14406399726867675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,2,power_law_1.01,0.0726144015789032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,2,power_law_1.01,0.1807487964630127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,2,power_law_1.01,0.099481600522995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,2,power_law_1.01,0.25627520084381106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,2,power_law_1.01,0.11867519617080688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,2,power_law_1.01,0.30794239044189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,2,power_law_1.01,0.162009596824646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,2,power_law_1.01,0.43754239082336427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,2,power_law_1.01,0.17342079877853395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,2,power_law_1.01,0.5164735794067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,2,power_law_1.01,0.2710400104522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,2,power_law_1.01,0.7966784000396728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,2,power_law_1.01,0.3330303907394409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,2,power_law_1.01,1.2706048011779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,2,power_law_1.01,0.5115007877349853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,2,power_law_1.01,2.173849678039551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,2,power_law_1.01,0.6955391883850097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,2,power_law_1.01,0.7873087882995605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,2,power_law_1.01,1.3513664245605468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,2,power_law_1.01,2.4433664321899413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,2,power_law_1.2,0.03155199885368347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,2,power_law_1.2,0.031385600566864014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,2,power_law_1.2,0.03157120048999786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,2,power_law_1.2,0.03249920010566711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,2,power_law_1.2,0.03264000117778778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,2,power_law_1.2,0.034220799803733826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,2,power_law_1.2,0.03617280125617981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,2,power_law_1.2,0.03578239977359772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,2,power_law_1.2,0.03764480054378509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,2,power_law_1.2,0.03811840116977692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,2,power_law_1.2,0.0377920001745224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,2,power_law_1.2,0.03935999870300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,2,power_law_1.2,0.04021120071411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,2,power_law_1.2,0.042316800355911253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,2,power_law_1.2,0.04924800097942352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,2,power_law_1.2,0.05335680246353149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,2,power_law_1.2,0.05441280007362366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,2,power_law_1.2,0.06820480227470398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,2,power_law_1.2,0.0761023998260498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,2,power_law_1.2,0.09945600032806397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,2,power_law_1.2,0.11774719953536987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,2,power_law_1.2,0.1703871965408325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,2,power_law_1.2,0.186080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,2,power_law_1.2,0.28017919063568114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,2,power_law_1.2,0.3282815933227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,2,power_law_1.2,0.5135871887207031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,2,power_law_1.2,0.644870376586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,2,power_law_1.2,0.7405824184417724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,2,power_law_1.2,1.2901503562927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,2,power_law_1.2,2.6518592834472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,1,balanced,0.03164266546567281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,1,balanced,0.03150933235883713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,1,balanced,0.035589332381884255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,1,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,1,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,1,power_law_1.01,0.030169600248336793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,1,balanced,0.03746666759252548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,1,power_law_1.01,0.03139199912548065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,1,balanced,0.0296426663796107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,1,balanced,0.038047999143600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,1,power_law_1.01,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,1,power_law_1.2,0.030079999566078187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,1,balanced,0.03164800008138021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,1,power_law_1.01,0.030937600135803222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,1,power_law_1.01,0.03592959940433502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,1,balanced,0.037402667105197906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,1,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,1,power_law_1.2,0.03178240060806274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,1,power_law_1.01,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,1,power_law_1.01,0.03084160089492798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,1,balanced,0.03875733415285746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,1,balanced,0.034346667428811394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,1,power_law_1.2,0.033292800188064575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,1,power_law_1.01,0.036032000184059144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,1,power_law_1.01,0.03214719891548157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,1,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,1,balanced,0.03578133384386698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,1,power_law_1.2,0.03607679903507233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,1,power_law_1.01,0.03900800049304962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,1,power_law_1.01,0.03415679931640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,1,balanced,0.039674667020638786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,1,balanced,0.03590933233499527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,1,power_law_1.2,0.03678719997406006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,1,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,1,power_law_1.01,0.03866879940032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,1,power_law_1.01,0.03430399894714355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,1,balanced,0.03788266579310099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,1,power_law_1.2,0.03548159897327423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,1,balanced,0.04168533285458883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,1,power_law_1.01,0.0455487996339798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,1,power_law_1.01,0.03624959886074066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,1,balanced,0.03794133414824804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,1,power_law_1.2,0.03824000060558319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,1,balanced,0.043882668018341064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,1,power_law_1.01,0.04791040122509003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,1,power_law_1.01,0.037363201379776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,1,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,1,power_law_1.2,0.04346239864826203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,1,balanced,0.059418668349583946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,1,power_law_1.01,0.04314880073070526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,1,power_law_1.01,0.03880960047245026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,1,balanced,0.0377866675456365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,1,power_law_1.2,0.04795520007610321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,1,power_law_1.2,0.030355200171470642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,1,balanced,0.03953066716591517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,1,power_law_1.2,0.047628799080848695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,1,balanced,0.059808000922203064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,1,power_law_1.01,0.04520959854125976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,1,power_law_1.01,0.03937920033931732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,1,balanced,0.04190400242805481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,1,power_law_1.2,0.030924800038337707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,1,power_law_1.2,0.0441536009311676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,1,balanced,0.06197333335876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,1,power_law_1.01,0.048537600040435794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,1,power_law_1.01,0.039392000436782836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,1,balanced,0.04200000067551931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,1,power_law_1.2,0.03047040104866028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,1,power_law_1.2,0.04635519981384277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,1,balanced,0.0786240001519521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,1,power_law_1.01,0.053363198041915895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,1,power_law_1.01,0.03978239893913269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,1,balanced,0.08028266827265422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,1,balanced,0.044122666120529175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,1,power_law_1.2,0.03389439880847931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,1,power_law_1.2,0.047219198942184445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,1,power_law_1.01,0.06169599890708923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,8,balanced,0.05813866853713989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,1,power_law_1.01,0.041715198755264284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,1,balanced,0.10539733370145161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,1,balanced,0.054144000013669334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,1,power_law_1.2,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,1,power_law_1.2,0.05010560154914856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,1,power_law_1.01,0.06288639903068542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,8,balanced,0.05973333120346069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,1,power_law_1.01,0.044531199336051944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,1,balanced,0.11143466830253601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,1,balanced,0.05596800148487091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,8,balanced,0.05658133327960968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,1,power_law_1.2,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,1,power_law_1.2,0.06144639849662781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,1,power_law_1.01,0.06386560201644897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,1,power_law_1.01,0.04746240079402923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,1,balanced,0.15597866972287497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,1,balanced,0.05619733532269796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,8,balanced,0.05572799841562907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,1,power_law_1.2,0.03779839873313904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,1,power_law_1.2,0.06395519971847534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,1,power_law_1.01,0.07795199751853943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,1,power_law_1.01,0.05683839917182922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,1,balanced,0.07074666519959767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,1,balanced,0.1678239901860555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,8,balanced,0.05788266658782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,1,power_law_1.2,0.03749119937419891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,1,power_law_1.2,0.0662335991859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,1,power_law_1.01,0.08683519959449768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,1,power_law_1.01,0.058323198556900026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,1,balanced,0.07916266719500224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,1,balanced,0.23797865708669028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,8,balanced,0.05799466868241628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,1,power_law_1.2,0.040652799606323245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,1,power_law_1.2,0.07783679962158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,1,power_law_1.01,0.10424959659576416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,8,balanced,0.05622399846712748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,1,balanced,0.3004586696624756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,1,power_law_1.01,0.05912960171699524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,1,balanced,0.10746133327484131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,1,power_law_1.2,0.0394239991903305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,1,power_law_1.2,0.08323839902877808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,1,balanced,0.12138133247693379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,1,power_law_1.01,0.11525119543075561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,8,balanced,0.05821866790453593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,1,power_law_1.01,0.07465599775314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,1,power_law_1.01,0.14809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,1,balanced,0.42687467734018963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,1,power_law_1.2,0.039743998646736146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,1,power_law_1.2,0.10421760082244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,1,balanced,0.16685332854588827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,8,balanced,0.05649066468079885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,1,power_law_1.01,0.08261119723320007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,1,power_law_1.01,0.17464959621429443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,1,balanced,0.5375039974848429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,1,power_law_1.2,0.04360319972038269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,1,power_law_1.2,0.11228159666061402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,8,balanced,0.056559999783833824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,1,balanced,0.1951520045598348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,1,power_law_1.01,0.10726399421691894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,1,power_law_1.01,0.23713281154632568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,1,power_law_1.2,0.04392319917678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,1,power_law_1.2,0.15344640016555786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,8,balanced,0.06038933495680491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,1,power_law_1.01,0.30212481021881105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,1,power_law_1.01,0.12155519723892212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,1,balanced,0.2792159914970398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,1,power_law_1.2,0.04602240025997162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,1,balanced,0.6605706612269083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,1,power_law_1.2,0.05392640233039856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,1,power_law_1.2,0.17348480224609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,8,balanced,0.05930666625499725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,1,power_law_1.01,0.42592639923095704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,1,power_law_1.01,0.16565120220184326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,1,balanced,0.359658678372701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,1,power_law_1.2,0.05669119954109192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,1,power_law_1.2,0.23658239841461182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,8,balanced,0.05839466551939646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,1,power_law_1.01,0.5514431953430176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,1,power_law_1.01,0.19674880504608155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,1,balanced,1.0204213460286458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,1,power_law_1.2,0.06109439730644226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,1,power_law_1.2,0.31059200763702394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,1,balanced,0.5202186504999796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,1,power_law_1.01,0.6763135910034179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,8,balanced,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,1,power_law_1.01,0.2796799898147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,1,power_law_1.2,0.07429119944572449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,1,power_law_1.2,0.4266496181488037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,1,power_law_1.01,1.0349056243896484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,8,balanced,0.06765333314736684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,1,power_law_1.01,0.3560767889022827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,1,power_law_1.2,0.08293120265007019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,1,balanced,0.6743893623352051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,1,power_law_1.2,0.5504703998565674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,1,power_law_1.01,1.9975040435791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,8,balanced,0.07027733325958252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,1,power_law_1.01,0.5203519821166992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,1,power_law_1.2,0.10670720338821411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,1,power_law_1.2,0.6746943950653076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,1,balanced,1.9896480242411296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,8,balanced,0.0702400008837382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,1,power_law_1.01,0.6803520202636719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,1,balanced,0.8363626797993978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,1,power_law_1.2,0.12079999446868897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,1,power_law_1.2,1.0345855712890626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,8,balanced,0.08042666812737782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,1,power_law_1.01,0.8359295845031738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,1,power_law_1.2,0.16599680185317994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,8,balanced,0.08058133224646251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,1,power_law_1.2,2.0078271865844726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,1,power_law_1.01,1.3057663917541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,1,power_law_1.2,0.19506560564041137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,8,balanced,0.09718400239944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,1,balanced,1.307098627090454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,1,power_law_1.01,2.5517120361328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,1,power_law_1.2,0.2791872024536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,8,balanced,0.10903466741243999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,1,power_law_1.2,0.36024320125579834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,8,balanced,0.12991467118263245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,1,power_law_1.2,0.5215104103088379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,1,power_law_1.2,0.6830207824707031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,8,balanced,0.15829333662986755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,1,power_law_1.2,0.8411647796630859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,8,balanced,0.20292800664901733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,1,power_law_1.2,1.3108287811279298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,8,balanced,0.24221867322921753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,1,balanced,2.540837287902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,1,power_law_1.2,2.554220771789551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,8,balanced,0.3409493366877238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,8,balanced,0.43090665340423584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,8,balanced,0.5334240198135376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,8,balanced,0.7997653484344482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,8,balanced,1.5207626024882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,8,balanced,0.037632000943024956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,8,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,8,balanced,0.03579200059175491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,8,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,8,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,8,balanced,0.03899733225504557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,8,balanced,0.03939733405907949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,8,balanced,0.03979733337958654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,8,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,8,balanced,0.03977066775163015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,8,balanced,0.04172799984614054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,8,balanced,0.0417546679576238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,8,balanced,0.041893333196640015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,8,balanced,0.04504533112049103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,8,balanced,0.051962668697039284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,8,balanced,0.053898667295773826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,8,balanced,0.062309334675470986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,8,balanced,0.07178666690985362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,8,balanced,0.07849599917729695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,8,balanced,0.09700266520182292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,8,balanced,0.10815466443697612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,8,balanced,0.13407466808954874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,8,balanced,0.16687999169031778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,8,balanced,0.2223680019378662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,8,balanced,0.27532800038655597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,8,balanced,0.38234134515126544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,8,balanced,0.47578132152557373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,8,balanced,0.5826773246129354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,8,power_law_1.01,0.03420799970626831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,8,power_law_1.01,0.05631999969482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,8,balanced,0.8993493715922037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,8,power_law_1.01,0.03461759984493255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,8,power_law_1.01,0.054028797149658206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,8,power_law_1.01,0.03380480110645294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,8,power_law_1.01,0.05306879878044128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,8,power_law_1.01,0.036735999584198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,8,power_law_1.01,0.05619199872016907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,8,power_law_1.01,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,4,8,power_law_1.2,0.055327999591827395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,8,power_law_1.01,0.055980801582336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,8,power_law_1.01,0.03775359988212586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,8,power_law_1.01,0.05598719716072083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,8,power_law_1.01,0.042310398817062375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,4,8,power_law_1.2,0.05363839864730835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,8,power_law_1.01,0.057734400033950806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,8,power_law_1.01,0.04552960097789764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,4,8,power_law_1.2,0.05252479910850525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,8,power_law_1.01,0.06074879765510559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,8,power_law_1.01,0.04734080135822296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,4,8,power_law_1.2,0.05576320290565491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,8,power_law_1.01,0.06348800063133239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,8,power_law_1.01,0.04935039877891541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,4,8,power_law_1.2,0.055327999591827395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,8,power_law_1.01,0.06302080154418946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,8,power_law_1.01,0.04776960015296936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,4,8,power_law_1.2,0.05568640232086182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,8,power_law_1.01,0.0648576021194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,8,power_law_1.01,0.0564736008644104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,4,8,power_law_1.2,0.05872640013694763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,8,balanced,1.727450688680013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,8,power_law_1.01,0.06880639791488648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,8,power_law_1.01,0.05319679975509643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,4,8,power_law_1.2,0.0616320013999939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,8,power_law_1.01,0.06984320282936096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,8,power_law_1.01,0.061273598670959474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,4,8,power_law_1.2,0.06422399878501892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,8,power_law_1.01,0.07580159902572632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,8,power_law_1.01,0.07493759989738465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,4,8,power_law_1.2,0.06293759942054748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,8,power_law_1.01,0.0870464026927948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,8,power_law_1.01,0.07577599883079529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,4,8,power_law_1.2,0.06515840291976929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,8,power_law_1.01,0.0889792025089264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,8,power_law_1.01,0.08547840118408204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,4,8,power_law_1.2,0.06844159960746765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,8,power_law_1.01,0.1017024040222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,8,power_law_1.01,0.10113919973373413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,4,8,power_law_1.2,0.0694208025932312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,8,power_law_1.01,0.12563199996948243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,8,power_law_1.01,0.11734399795532227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,4,8,power_law_1.2,0.07525119781494141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,8,power_law_1.01,0.15045759677886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,8,power_law_1.01,0.1463871955871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,4,8,power_law_1.2,0.08431360125541687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,8,power_law_1.01,0.185862398147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,8,power_law_1.01,0.179967999458313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,4,8,power_law_1.2,0.08732799887657165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,8,power_law_1.01,0.23816959857940673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,4,8,power_law_1.2,0.0972544014453888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,8,power_law_1.01,0.27578880786895754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,8,power_law_1.01,0.23169920444488526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,4,8,power_law_1.2,0.1343808054924011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,8,power_law_1.01,0.4019968032836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,8,power_law_1.01,0.30500481128692625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,4,8,power_law_1.2,0.14075520038604736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,8,power_law_1.01,0.6360576152801514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,8,power_law_1.01,0.4571904182434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,4,8,power_law_1.2,0.1969599962234497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,8,power_law_1.01,0.7867136001586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,8,power_law_1.01,0.5418560028076171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,4,8,power_law_1.2,0.23267199993133544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,8,power_law_1.01,1.1938176155090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,8,power_law_1.01,0.8150976181030274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,4,8,power_law_1.2,0.293939208984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,8,balanced,0.023061332603295643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,8,power_law_1.01,1.5184767723083497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,8,power_law_1.01,0.994758415222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,4,8,power_law_1.2,0.4154816150665283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,8,power_law_1.01,1.8718975067138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,8,balanced,0.023226665953795116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,8,power_law_1.01,1.1738240242004394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,4,8,power_law_1.2,0.6169856071472168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,8,power_law_1.01,2.77589111328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,8,balanced,0.021327999730904896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,8,power_law_1.01,1.993894386291504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,4,8,power_law_1.2,0.863475227355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,8,balanced,0.039749334255854286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,8,power_law_1.01,4.86835823059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,8,balanced,0.021141332884629566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,8,power_law_1.01,3.8374080657958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,4,8,power_law_1.2,1.3712512016296388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,8,power_law_1.01,0.020908799767494202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,8,balanced,0.04165866722663244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,8,balanced,0.02293866624434789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,4,8,power_law_1.2,1.531942367553711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,8,power_law_1.01,0.020691199600696562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,8,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,8,balanced,0.02346133440732956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,8,power_law_1.01,0.02077440023422241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,4,8,power_law_1.2,1.743929672241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,8,balanced,0.03982399900754293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,8,power_law_1.01,0.019827200472354888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,8,balanced,0.031328000128269196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,8,balanced,0.041536000867684685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,4,8,power_law_1.2,2.7787071228027345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,8,power_law_1.01,0.020876799523830415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,8,balanced,0.02604266752799352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,8,balanced,0.04125866790612539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,8,power_law_1.01,0.022841599583625794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,4,8,power_law_1.2,4.745836639404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,8,balanced,0.03362133353948593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,8,balanced,0.041759997606277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,8,power_law_1.01,0.029849600791931153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,8,balanced,0.031701333820819855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,8,balanced,0.041536000867684685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,8,power_law_1.01,0.02568320035934448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,8,power_law_1.01,0.03338240087032318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,8,balanced,0.04155199974775314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,8,balanced,0.03937600056330363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,8,power_law_1.01,0.030950400233268737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,8,balanced,0.0425546665986379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,8,power_law_1.01,0.03711360096931458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,8,balanced,0.05259733398755392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,8,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,8,power_law_1.01,0.037920001149177554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,8,power_law_1.01,0.05100160241127014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,8,balanced,0.050144001841545105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,8,balanced,0.04354133208592733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,8,power_law_1.01,0.04913919866085052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,8,power_law_1.01,0.03850879967212677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,8,balanced,0.04095466683308283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,8,balanced,0.043194666504859924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,8,power_law_1.01,0.03914879858493805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,8,power_law_1.01,0.038412800431251524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,8,balanced,0.0602400004863739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,8,balanced,0.04388799766699473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,8,power_law_1.01,0.058713597059249875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,8,power_law_1.01,0.03808000087738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,8,balanced,0.053904001911481224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,8,balanced,0.05183466772238413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,8,power_law_1.01,0.05251200199127197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,8,power_law_1.01,0.03818880021572113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,8,balanced,0.06038933495680491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,8,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,8,power_law_1.01,0.058432000875473025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,8,power_law_1.01,0.039289599657058714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,8,balanced,0.07455466687679291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,8,balanced,0.05208000044027964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,8,power_law_1.01,0.07301120162010193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,8,power_law_1.01,0.03932799994945526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,8,balanced,0.09293333689371745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.01,0.09031040072441102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,8,balanced,0.05821866790453593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,8,power_law_1.01,0.04005120098590851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.01,0.1242751955986023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,8,balanced,0.12625599900881448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.01,0.1494271993637085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,8,balanced,0.06198399762312571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,8,power_law_1.01,0.040934398770332336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.01,0.2096640110015869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,8,balanced,0.15110933780670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,8,balanced,0.0767146646976471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,8,power_law_1.01,0.043315199017524716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.01,0.2649856090545654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,8,balanced,0.21312000354131064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,8,balanced,0.08547199765841167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.01,0.38122239112854006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,8,power_law_1.01,0.044064000248909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,8,balanced,0.2685386737187703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.01,0.49772157669067385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,8,balanced,0.1058240036169688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,8,power_law_1.01,0.048492801189422605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,8,balanced,0.38444264729817706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.01,0.7281216144561767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,8,balanced,0.12086932857831319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,8,power_law_1.01,0.04893440008163452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.01,0.9580608367919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,8,balanced,0.49961598714192706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,8,power_law_1.01,1.1893312454223632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,8,balanced,0.15121066570281982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,8,power_law_1.01,0.0526528000831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,8,power_law_1.01,1.8824192047119142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,8,balanced,0.7317439715067545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,8,power_law_1.01,0.060288000106811526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,8,balanced,0.18292800585428873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,8,power_law_1.01,3.7338497161865236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,8,power_law_1.01,0.06381440162658691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,8,balanced,0.9609493414560953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,8,balanced,0.24658666054407755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,8,power_law_1.01,0.07301759719848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,8,power_law_1.01,0.08492799997329711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,8,balanced,0.30703999598821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.01,0.09329280257225037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,8,balanced,1.1910026868184407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.01,0.11431679725646973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,8,balanced,0.3729013204574585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.01,0.13519359827041627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,8,balanced,1.8833920160929363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.01,0.1796223998069763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.01,0.2140415906906128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,8,balanced,0.5653226772944132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.01,0.3220031976699829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.01,0.39801599979400637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.01,0.5803647994995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.01,0.7130239963531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,8,balanced,3.7250134150187173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,8,power_law_1.01,0.8771072387695312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,8,balanced,1.0693333148956299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,8,power_law_1.01,1.381593608856201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,8,power_law_1.01,2.753881645202637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,4,8,power_law_1.2,0.034118399024009705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,4,8,power_law_1.2,0.03480960130691528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,4,8,power_law_1.2,0.03331199884414673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,4,8,power_law_1.2,0.03603839874267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,4,8,power_law_1.2,0.03614720106124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,4,8,power_law_1.2,0.03702400028705597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,4,8,power_law_1.2,0.04168320000171662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,4,8,power_law_1.2,0.044915199279785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,4,8,power_law_1.2,0.05287039875984192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,4,8,power_law_1.2,0.04867840111255646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,4,8,power_law_1.2,0.04947839975357056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,4,8,power_law_1.2,0.05361279845237732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,4,8,power_law_1.2,0.04713599979877472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,4,8,power_law_1.2,0.052902400493621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,4,8,power_law_1.2,0.05772799849510193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,4,8,power_law_1.2,0.05505920052528381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,4,8,power_law_1.2,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,4,8,power_law_1.2,0.05449600219726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,4,8,power_law_1.2,0.06149759888648987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,4,8,power_law_1.2,0.05377280116081238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,4,8,power_law_1.2,0.07208319902420043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,4,8,power_law_1.2,0.052697598934173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,4,8,power_law_1.2,0.07517439723014832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,4,8,power_law_1.2,0.05264000296592712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,4,8,power_law_1.2,0.02190079987049103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,4,8,power_law_1.2,0.0843455970287323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,4,8,power_law_1.2,0.05077120065689087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,4,8,power_law_1.2,0.021580800414085388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,4,8,power_law_1.2,0.10661120414733886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,4,8,power_law_1.2,0.05123199820518494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,4,8,power_law_1.2,0.020857599377632142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,4,8,power_law_1.2,0.12126719951629639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,4,8,power_law_1.2,0.049055999517440795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,4,8,power_law_1.2,0.020883199572563172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,4,8,power_law_1.2,0.15202560424804687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,4,8,power_law_1.2,0.05310720205307007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,4,8,power_law_1.2,0.02157440036535263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,4,8,power_law_1.2,0.1851583957672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,4,8,power_law_1.2,0.060864001512527466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,4,8,power_law_1.2,0.022681599855422972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,4,8,power_law_1.2,0.24164481163024903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,4,8,power_law_1.2,0.05932160019874573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,4,8,power_law_1.2,0.029151999950408937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,4,8,power_law_1.2,0.3167680025100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,4,8,power_law_1.2,0.0713536024093628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,4,8,power_law_1.2,0.02558079957962036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,4,8,power_law_1.2,0.4420032024383545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,4,8,power_law_1.2,0.06839039921760559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,4,8,power_law_1.2,0.03208320140838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,4,8,power_law_1.2,0.5813759803771973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,4,8,power_law_1.2,0.07701759934425353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,4,8,power_law_1.2,0.03052160143852234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,4,8,power_law_1.2,0.837343978881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,4,8,power_law_1.2,0.09658240079879761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,4,8,power_law_1.2,0.03703039884567261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,4,8,power_law_1.2,1.0490176200866699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.2,0.11344000101089477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,4,8,power_law_1.2,0.050496000051498416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,4,8,power_law_1.2,1.2221504211425782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.2,0.14492160081863403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,4,8,power_law_1.2,0.04932479858398438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,4,8,power_law_1.2,1.9672832489013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.2,0.1783679962158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,4,8,power_law_1.2,0.039987200498580934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,4,8,power_law_1.2,3.8995777130126954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.2,0.2465280055999756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,4,8,power_law_1.2,0.058329600095748904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.2,0.2909312009811401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,4,8,power_law_1.2,0.05175679922103882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.2,0.44977917671203616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,4,8,power_law_1.2,0.05852159857749939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.2,0.5662015914916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,4,8,power_law_1.2,0.07322239875793457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.2,0.8522303581237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.2,0.09044479727745056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.2,1.0904831886291504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.2,0.12563199996948243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,4,8,power_law_1.2,1.3928832054138183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.2,0.15061119794845582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,4,8,power_law_1.2,2.217849540710449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.2,0.21013119220733642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,4,8,power_law_1.2,4.3854015350341795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.2,0.2665024042129517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.2,0.3818367958068848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.2,0.49621119499206545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.2,0.7279232025146485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.2,0.9580991744995118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,4,8,power_law_1.2,1.189139175415039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,4,8,power_law_1.2,1.878214454650879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,4,8,power_law_1.2,3.7223934173583983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,4,balanced,0.021029333273569744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,4,balanced,0.020975999534130096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,4,balanced,0.021125334004561108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,4,balanced,0.020992000897725422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,4,balanced,0.03756266583998998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,4,balanced,0.02107733239730199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,4,balanced,0.037717332442601524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,4,balanced,0.02163733293612798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,4,balanced,0.03980266551176707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,4,balanced,0.039674667020638786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,4,balanced,0.022997332115968067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,4,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,4,balanced,0.03753600021203359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,4,balanced,0.02513066679239273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,4,balanced,0.05560000240802765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,4,balanced,0.039434666434923805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,4,balanced,0.05394133428732554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,4,balanced,0.031685332457224526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,4,balanced,0.0394400010506312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,4,balanced,0.05413866539796194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,4,balanced,0.031632001201311745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,4,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,4,balanced,0.05585066477457682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,4,balanced,0.029535998900731403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,4,balanced,0.03968533376852671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,4,balanced,0.0558186670144399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,4,balanced,0.031541332602500916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,4,balanced,0.03934400031963984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,4,balanced,0.057445332407951355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,4,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,4,balanced,0.041450666884581246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,4,balanced,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,4,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,4,balanced,0.041562666495641075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,4,balanced,0.056176001826922096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,4,balanced,0.044165333112080894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,4,balanced,0.04178133110205332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,4,balanced,0.05985599756240845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,4,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,4,balanced,0.04176533222198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,4,balanced,0.05825600028038025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,4,balanced,0.043605332573254905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,4,balanced,0.04211199780305227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,4,balanced,0.058549334605534874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,4,balanced,0.052789335449536644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,4,balanced,0.055914665261904396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,4,balanced,0.06229866544405619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,4,balanced,0.050474668542544045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,4,balanced,0.060346667965253196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,4,balanced,0.0684853345155716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,4,balanced,0.051674668987592064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,4,balanced,0.08277866741021474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,4,balanced,0.06870933373769124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,4,balanced,0.05991466840108236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,4,balanced,0.10328533252080281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,4,balanced,0.07082666456699371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,4,balanced,0.06598400076230367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,4,balanced,0.07899733384450276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,4,balanced,0.07801066835721333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,4,balanced,0.13804266850153604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,4,balanced,0.0817386656999588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,4,balanced,0.09019200007120769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,4,balanced,0.16644799709320068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,4,balanced,0.09714667002360027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,4,balanced,0.11034133036931355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,4,balanced,0.10726400216420491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,4,balanced,0.2369226614634196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,4,balanced,0.13782399892807007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,4,balanced,0.13397866487503052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,4,balanced,0.30191999673843384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,4,balanced,0.17627199490865073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,4,balanced,0.1606773336728414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,4,balanced,0.43805332978566486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,4,balanced,0.2118613322575887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,4,balanced,0.20815465847651163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,4,balanced,0.5763466755549113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,4,balanced,0.29365867376327515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,4,balanced,0.2642773389816284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,4,balanced,0.7094293435414633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,4,balanced,0.3728533188501994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,4,balanced,0.3622026840845744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,4,balanced,0.45125333468119305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,4,balanced,0.4386826753616333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,4,balanced,1.1140159765879314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,4,balanced,0.5352426767349243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,4,balanced,0.6871946652730306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,4,balanced,0.03677866607904434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,4,balanced,0.035786665976047516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,4,balanced,0.03579200059175491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,4,balanced,0.8067626953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,4,balanced,2.1921812693277993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,4,balanced,0.0376800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,4,balanced,0.0378506655494372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,4,balanced,1.3033866882324219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,4,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,4,balanced,0.038480001191298165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,4,balanced,0.037717332442601524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,4,power_law_1.01,0.03488639891147614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,4,balanced,1.5546773274739583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,4,balanced,0.03982399900754293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,4,power_law_1.01,0.034041601419448855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,4,balanced,0.039887999494870506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,4,power_law_1.01,0.03333759903907776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,4,balanced,0.041749333341916404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,4,power_law_1.01,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,4,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,4,power_law_1.01,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,4,balanced,0.04234133164087931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,4,power_law_1.01,0.037561601400375365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,4,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,4,power_law_1.01,0.041152000427246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,4,balanced,0.0540533314148585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,4,power_law_1.01,0.040863999724388124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,4,balanced,0.05628266433874766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,4,power_law_1.01,0.044198399782180785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,4,balanced,0.05966933568318685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,4,power_law_1.01,0.044607999920845035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,4,balanced,0.07225599884986877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,4,power_law_1.01,0.04347519874572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,4,balanced,0.081194669008255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,4,power_law_1.01,0.04764800071716309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,4,power_law_1.01,0.05268480181694031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,4,balanced,0.1011199951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,4,power_law_1.01,0.049568000435829165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,4,power_law_1.01,0.04871039986610413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,4,power_law_1.01,0.05666559934616089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,4,balanced,0.11338667074839275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,4,power_law_1.01,0.066348797082901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,4,power_law_1.01,0.047814399003982544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,4,balanced,0.14518933494885763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,4,power_law_1.01,0.07236480116844177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,4,power_law_1.01,0.05383679866790771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,4,power_law_1.01,0.020422400534152986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,4,power_law_1.01,0.07281280159950257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,4,balanced,0.17675199111302695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,4,power_law_1.01,0.05534719824790955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,4,power_law_1.01,0.05504000186920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,4,power_law_1.01,0.08789119720458985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,4,power_law_1.01,0.01968639940023422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,4,balanced,0.23946666717529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,4,power_law_1.01,0.052211201190948485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,4,power_law_1.01,0.05422719717025757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,4,power_law_1.01,0.01955839991569519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,4,power_law_1.01,0.10887680053710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,4,power_law_1.01,0.05244799852371216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,4,power_law_1.01,0.05470079779624939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,4,power_law_1.01,0.01955839991569519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,4,balanced,0.3036800026893616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,4,power_law_1.01,0.12984319925308227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,4,power_law_1.01,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,4,power_law_1.01,0.054073601961135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,4,power_law_1.01,0.020211200416088104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,4,power_law_1.01,0.1529088020324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,4,balanced,0.42331735293070477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,4,power_law_1.01,0.05322880148887634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,4,power_law_1.01,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,4,power_law_1.01,0.02078080028295517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,4,power_law_1.01,0.21806080341339112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,4,power_law_1.01,0.05377920269966126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,4,power_law_1.01,0.05446400046348572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,4,power_law_1.01,0.02102400064468384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,4,power_law_1.01,0.2891200065612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,4,balanced,0.5411253372828165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,4,power_law_1.01,0.05621119737625122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,4,power_law_1.01,0.05390080213546753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,4,power_law_1.01,0.3739583969116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,4,power_law_1.01,0.024294400215148927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,4,power_law_1.01,0.0574783980846405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,4,power_law_1.01,0.5033472061157227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,4,power_law_1.01,0.05671039819717407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,4,power_law_1.01,0.029407998919486998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,4,power_law_1.01,0.05902079939842224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,4,balanced,0.6648053328196207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,4,power_law_1.01,0.7127423763275147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,4,power_law_1.01,0.057708799839019775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,4,power_law_1.01,0.030265599489212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,4,power_law_1.01,0.05988479852676391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,4,power_law_1.01,1.0028223991394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,4,power_law_1.01,0.061862397193908694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,4,power_law_1.01,0.02773120105266571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,4,power_law_1.01,0.06419839859008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,4,power_law_1.01,1.2374208450317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,4,power_law_1.01,0.06487039923667907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,4,power_law_1.01,0.03054719865322113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,4,power_law_1.01,0.06876800060272217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,4,balanced,1.0954079627990723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,4,power_law_1.01,2.05865592956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,4,power_law_1.01,0.06991999745368957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,4,power_law_1.01,0.03234559893608093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,4,power_law_1.01,0.06451839804649354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,4,power_law_1.01,4.1760704040527346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,4,power_law_1.01,0.07366399765014649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,4,power_law_1.01,0.03720960021018982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,4,power_law_1.01,0.07095040082931518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,4,power_law_1.01,0.08460800051689148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,4,power_law_1.01,0.04287999868392944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,4,power_law_1.01,0.07888640165328979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.01,0.10072959661483764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,4,power_law_1.01,0.03886080086231232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,4,power_law_1.01,0.08169599771499633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.01,0.12012799978256225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,4,power_law_1.01,0.0426367998123169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.01,0.16092159748077392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,4,power_law_1.01,0.08821120262145996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,4,power_law_1.01,0.05509759783744812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,4,balanced,1.9580532709757488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.01,0.17976959943771362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,4,power_law_1.01,0.10454399585723877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.01,0.06053119897842407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.01,0.2468928098678589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,4,power_law_1.01,0.1211583971977234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.01,0.081632000207901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.01,0.39054720401763915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,4,power_law_1.01,0.13441280126571656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.01,0.10144000053405762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.01,0.48830080032348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,4,power_law_1.01,0.18199679851531983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.01,0.13612159490585327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.01,0.7652863979339599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,4,power_law_1.01,0.25740799903869627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.01,0.1654911994934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.01,0.9621439933776855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,4,power_law_1.01,0.27591040134429934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.01,0.23698561191558837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,4,power_law_1.01,1.0699263572692872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.01,0.300927996635437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,4,power_law_1.01,0.4121920108795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,4,power_law_1.01,2.0063039779663088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.01,0.43744640350341796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,4,power_law_1.01,0.6136832237243652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,4,power_law_1.01,3.918668746948242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.01,0.5741568088531495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,4,power_law_1.01,0.8833536148071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,4,power_law_1.01,0.7070591926574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,4,power_law_1.01,1.0089920043945313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,4,power_law_1.01,1.1138879776000976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,4,power_law_1.01,1.0885696411132812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,4,power_law_1.01,2.1927104949951173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,4,power_law_1.01,2.2693824768066406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,4,power_law_1.01,4.00458869934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,4,power_law_1.2,0.03371520042419433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,4,power_law_1.2,0.048979198932647704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,4,power_law_1.2,0.034720000624656674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,4,power_law_1.2,0.05119360089302063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,4,power_law_1.2,0.05495039820671081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,4,power_law_1.2,0.033670398592948916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,4,power_law_1.2,0.04408960044384003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,4,power_law_1.2,0.049132800102233885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,4,power_law_1.2,0.03444480001926422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,4,power_law_1.2,0.052198398113250735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,4,power_law_1.2,0.0506496012210846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,4,power_law_1.2,0.03610239923000336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,4,power_law_1.2,0.05275520086288452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,4,power_law_1.2,0.0532800018787384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,4,power_law_1.2,0.037510401010513304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,4,power_law_1.2,0.051481598615646364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,4,power_law_1.2,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,4,power_law_1.2,0.04116480052471161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,4,power_law_1.2,0.05616000294685364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,4,power_law_1.2,0.05470719933509827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,4,power_law_1.2,0.04245119988918304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,4,power_law_1.2,0.05567359924316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,4,power_law_1.2,0.05724800229072571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,4,power_law_1.2,0.04411520063877106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,4,power_law_1.2,0.05299199819564819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,4,power_law_1.2,0.055206400156021115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,4,power_law_1.2,0.05330560207366943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,4,power_law_1.2,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,4,power_law_1.2,0.054630398750305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,4,power_law_1.2,0.06130560040473938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,4,power_law_1.2,0.04493440091609955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,4,power_law_1.2,0.05295360088348389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,4,power_law_1.2,0.06268159747123718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,4,power_law_1.2,0.05187199711799621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,4,power_law_1.2,0.06220160126686096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,4,power_law_1.2,0.06910719871520996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,4,power_law_1.2,0.05719040036201477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,4,power_law_1.2,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,4,power_law_1.2,0.06636800169944763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,4,power_law_1.2,0.0605567991733551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,4,power_law_1.2,0.0674560010433197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,4,power_law_1.2,0.07031040191650391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,4,power_law_1.2,0.056403201818466184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,4,power_law_1.2,0.0644864022731781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,4,power_law_1.2,0.08055679798126221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,4,power_law_1.2,0.06524159908294677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,4,power_law_1.2,0.07519360184669495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,4,power_law_1.2,0.08085759878158569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,4,power_law_1.2,0.07340160012245178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,4,power_law_1.2,0.09076480269432068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,4,power_law_1.2,0.09233279824256897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,4,power_law_1.2,0.07442560195922851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.2,0.0955839991569519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,4,power_law_1.2,0.10510720014572143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.2,0.13722879886627198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,4,power_law_1.2,0.09201279878616334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,4,power_law_1.2,0.12692480087280272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.2,0.14237439632415771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,4,power_law_1.2,0.10871039628982544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.2,0.2135551929473877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,4,power_law_1.2,0.1538879990577698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,4,power_law_1.2,0.1413632035255432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.2,0.26019840240478515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,4,power_law_1.2,0.20156800746917725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,4,power_law_1.2,0.1786687970161438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.2,0.3978624105453491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,4,power_law_1.2,0.2985663890838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,4,power_law_1.2,0.2305216073989868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.2,0.5208255767822265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,4,power_law_1.2,0.31624319553375246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,4,power_law_1.2,0.3048320055007935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.2,0.7035903930664062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,4,power_law_1.2,0.3710911989212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,4,power_law_1.2,0.43402881622314454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.2,1.0211647987365722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,4,power_law_1.2,0.5902080059051513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,4,power_law_1.2,0.5616767883300782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,4,power_law_1.2,1.1661248207092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,4,power_law_1.2,0.835264015197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,4,power_law_1.2,0.7299967765808105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,4,power_law_1.2,1.8790143966674804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,4,power_law_1.2,1.1527551651000976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,4,power_law_1.2,0.8694463729858398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,4,power_law_1.2,4.150003051757812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,4,power_law_1.2,1.42673921585083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,4,power_law_1.2,1.0358336448669434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,4,power_law_1.2,2.336089515686035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,4,power_law_1.2,2.437811279296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,4,power_law_1.2,0.020703999698162077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,4,power_law_1.2,3.7950782775878906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,4,power_law_1.2,4.115801620483398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,4,power_law_1.2,0.019519999623298645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,4,power_law_1.2,0.019475199282169342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,4,power_law_1.2,0.019513599574565887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,4,power_law_1.2,0.020364800095558168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,4,power_law_1.2,0.020403200387954713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,4,power_law_1.2,0.02091519981622696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,4,power_law_1.2,0.024076800048351287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,4,power_law_1.2,0.02942720055580139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,4,power_law_1.2,0.03025279939174652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,4,power_law_1.2,0.027091199159622194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,4,power_law_1.2,0.030988800525665283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,4,power_law_1.2,0.031974399089813234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,4,power_law_1.2,0.036908799409866334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,4,power_law_1.2,0.042752000689506534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,4,power_law_1.2,0.03837440013885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,4,power_law_1.2,0.04221439957618713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,4,power_law_1.2,0.055103999376296994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.2,0.05985280275344849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.2,0.0815936028957367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.2,0.10075520277023316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.2,0.13608319759368898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.2,0.16506880521774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.2,0.2357503890991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.2,0.30132479667663575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.2,0.43836798667907717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.2,0.5727615833282471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,4,power_law_1.2,0.7078271865844726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,4,power_law_1.2,1.1132543563842774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,4,power_law_1.2,2.193606376647949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,2,balanced,0.020058666666348774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,2,balanced,0.01964266722400983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,2,balanced,0.020917333662509918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,2,balanced,0.020106667031844456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,2,balanced,0.021375998854637146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,2,balanced,0.021370666722456615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,2,balanced,0.022346665461858112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,2,balanced,0.0236160010099411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,2,balanced,0.023397333920001984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,2,balanced,0.02382933348417282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,2,balanced,0.025439999997615814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,2,balanced,0.03165333221356074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,2,balanced,0.03215999901294708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,2,balanced,0.0296426663796107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,2,balanced,0.0335359995563825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,2,balanced,0.03577066709597906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,2,balanced,0.04190933207670847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,2,balanced,0.045610666275024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,2,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,2,balanced,0.06860266625881195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,2,balanced,0.07835199932257335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,2,balanced,0.037871999045213066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,2,balanced,0.10109333197275798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,2,balanced,0.037834666669368744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,2,balanced,0.12679466605186462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,2,balanced,0.03978666663169861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,2,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,2,balanced,0.17399466037750244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,2,balanced,0.039808000127474465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,2,balanced,0.040864000717798867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,2,balanced,0.04174399872620901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,2,balanced,0.21308799584706625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,2,balanced,0.041850666205088295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,2,balanced,0.04189866781234741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,2,balanced,0.041450666884581246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,2,balanced,0.30929599205652875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,2,balanced,0.043552001317342125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,2,balanced,0.04400533437728882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,2,balanced,0.0436160018046697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,2,balanced,0.396565318107605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,2,balanced,0.04558399816354116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,2,balanced,0.05153066913286845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,2,balanced,0.05379733443260193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,2,power_law_1.01,0.043635201454162595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,2,balanced,0.05587733288606008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,2,balanced,0.06449066599210103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,2,balanced,0.4920533498128255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,2,power_law_1.01,0.04333440065383911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,2,balanced,0.07261333366235097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,2,power_law_1.01,0.04613119959831238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,2,balanced,0.08901866277058919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,2,power_law_1.01,0.05237759947776795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,2,power_law_1.01,0.05172479748725891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,2,balanced,0.10326932867368062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,2,power_law_1.01,0.05236480236053467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,2,balanced,0.12754133343696594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,2,power_law_1.01,0.052908802032470705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,2,power_law_1.01,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,2,balanced,0.7622293631235758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,2,balanced,0.15214932958285013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,2,power_law_1.01,0.05265920162200928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,2,power_law_1.01,0.05204480290412903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,2,balanced,0.20578134059906006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,2,power_law_1.01,0.0529151976108551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,2,power_law_1.01,0.05354239940643311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,2,power_law_1.01,0.05576320290565491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,2,balanced,0.2630560000737508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,2,power_law_1.01,0.019731199741363524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,2,power_law_1.01,0.05916159749031067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,2,power_law_1.01,0.06601600050926208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,2,power_law_1.01,0.06788480281829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,2,power_law_1.01,0.018700799345970152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,2,power_law_1.01,0.0725055992603302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,2,power_law_1.01,0.018643200397491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,2,balanced,0.37189332644144696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,2,power_law_1.01,0.08749439716339111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,2,power_law_1.01,0.018681600689888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.01,0.09701120257377624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,2,power_law_1.01,0.01984640061855316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.01,0.12378879785537719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,2,power_law_1.01,0.01937279999256134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.01,0.14191999435424804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,2,power_law_1.01,0.020396800339221956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,2,balanced,1.4888854026794434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.01,0.19022719860076903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,2,balanced,0.4776959816614787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,2,power_law_1.01,0.021695999801158904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.01,0.19857280254364013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,2,power_law_1.01,0.021766400337219237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.01,0.32454400062561034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,2,power_law_1.01,0.022252799570560457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.01,0.4017216205596924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,2,power_law_1.01,0.024345600605010988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.01,0.6164671897888183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,2,power_law_1.01,0.02958720028400421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.01,0.8959936141967774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,2,balanced,0.5873226722081503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,2,power_law_1.01,1.0112256050109862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,2,power_law_1.01,0.030681601166725157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,2,power_law_1.01,1.680339241027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,2,power_law_1.01,0.028064000606536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,2,power_law_1.01,0.031513598561286923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,2,power_law_1.01,3.2330368041992186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,2,power_law_1.01,0.03407999873161316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,2,power_law_1.01,0.04115839898586273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,2,power_law_1.01,0.04357120096683502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.01,0.04616959989070892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,2,balanced,0.9058506488800049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.01,0.06661760210990905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.01,0.07501440048217774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.01,0.0992255985736847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.01,0.12446719408035278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.01,0.17111040353775026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.01,0.20936961174011232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.01,0.3061311960220337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.01,0.3922368049621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,2,power_law_1.01,0.48769278526306153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,2,power_law_1.01,0.7586304187774658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,2,power_law_1.01,1.4830207824707031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,2,balanced,1.7559092839558919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,2,balanced,0.035445332527160645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,2,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,2,balanced,0.037530665596326195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,2,balanced,0.03577066709597906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,2,balanced,0.05402666827042898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,2,balanced,0.03781333317359289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,2,balanced,0.054901331663131714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,2,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,2,balanced,0.05555733541647593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,2,balanced,0.037674665451049805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,2,balanced,0.055205335219701133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,2,balanced,0.037615999579429626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,2,balanced,0.0554666668176651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,2,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,2,balanced,0.054101333022117615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,2,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,2,balanced,0.05606933434804281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,2,balanced,0.039520000418027244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,2,balanced,0.05651199817657471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,2,balanced,0.04126933217048645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,2,balanced,0.05831466615200043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,2,balanced,0.04219200213750204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,2,balanced,0.058143998185793556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,2,balanced,0.04231466849644979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,2,balanced,0.06016000111897787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,2,balanced,0.0436160018046697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,2,balanced,0.06187200049559275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,2,balanced,0.06942399839560191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,2,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,2,power_law_1.2,0.031385600566864014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,2,power_law_1.2,0.034944000840187076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,2,balanced,0.07025599976380666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,2,balanced,0.062319998939832054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,2,power_law_1.2,0.03162879943847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,2,balanced,0.07237866520881653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,2,balanced,0.06250666578610738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,2,power_law_1.2,0.0318336009979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,2,balanced,0.08238933483759563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,2,power_law_1.2,0.03717760145664215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,2,balanced,0.0612960010766983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,2,balanced,0.08591999610265096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,2,power_law_1.2,0.03546879887580871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,2,power_law_1.2,0.03699840009212494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,2,balanced,0.07667199770609538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,2,balanced,0.10325866937637329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,2,power_law_1.2,0.035872000455856326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,2,power_law_1.2,0.05319039821624756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,2,balanced,0.12956800063451132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,2,power_law_1.2,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,2,balanced,0.08761599659919739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,2,power_law_1.2,0.053350400924682614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,2,balanced,0.16269333163897196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,2,power_law_1.2,0.04066559970378876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,2,power_law_1.2,0.053472000360488894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,2,balanced,0.10971200466156006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,2,balanced,0.18453333775202432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,2,power_law_1.2,0.041254401206970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,2,power_law_1.2,0.05628799796104431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,2,balanced,0.1334719955921173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,2,power_law_1.2,0.04521600008010864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,2,power_law_1.2,0.05742719769477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,2,balanced,0.2449280023574829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,2,power_law_1.2,0.046079999208450316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,2,power_law_1.2,0.05849599838256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,2,balanced,0.18017599980036417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,2,power_law_1.2,0.04824959933757782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,2,power_law_1.2,0.05868800282478333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,2,power_law_1.2,0.04339199960231781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,2,power_law_1.2,0.01903360038995743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,2,power_law_1.2,0.04903680086135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,2,balanced,0.30766934156417847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,2,power_law_1.2,0.06064000129699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,2,balanced,0.19956799348195395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,2,power_law_1.2,0.050969600677490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,2,power_law_1.2,0.04716159999370575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,2,power_law_1.2,0.06159999966621399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,2,power_law_1.2,0.018873600661754607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,2,power_law_1.2,0.05486080050468445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,2,power_law_1.2,0.04802559912204742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,2,power_law_1.2,0.06177279949188232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,2,power_law_1.2,0.018572799861431122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,2,power_law_1.2,0.0642624020576477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,2,balanced,0.2768266598383586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,2,power_law_1.2,0.05253120064735413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,2,power_law_1.2,0.07050880193710327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,2,balanced,0.430458664894104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,2,power_law_1.2,0.01905920058488846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,2,power_law_1.2,0.06804479956626892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,2,power_law_1.2,0.052070397138595584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,2,power_law_1.2,0.07850880026817322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,2,power_law_1.2,0.019980800151824952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,2,power_law_1.2,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,2,power_law_1.2,0.0802623987197876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,2,power_law_1.2,0.01932799965143204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,2,power_law_1.2,0.052127999067306516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,2,balanced,0.3574666579564412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,2,power_law_1.2,0.08865280151367187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,2,power_law_1.2,0.08092799782752991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,2,power_law_1.2,0.020364800095558168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,2,power_law_1.2,0.05384960174560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,2,power_law_1.2,0.10028159618377686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,2,power_law_1.2,0.10005760192871094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,2,power_law_1.2,0.05370240211486817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,2,balanced,0.5302613178888956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,2,power_law_1.2,0.021715199947357176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,2,power_law_1.2,0.1106112003326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,2,power_law_1.2,0.1349951982498169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,2,power_law_1.2,0.052223998308181765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,2,power_law_1.2,0.13912960290908813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,2,power_law_1.2,0.05303040146827698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,2,power_law_1.2,0.021984000504016877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,2,power_law_1.2,0.16810879707336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,2,balanced,0.5159306526184082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,2,power_law_1.2,0.17557120323181152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,2,power_law_1.2,0.052934402227401735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,2,power_law_1.2,0.21521279811859131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,2,power_law_1.2,0.022489599883556366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,2,power_law_1.2,0.2251904010772705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,2,power_law_1.2,0.057062399387359616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,2,balanced,0.6534666617711385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,2,power_law_1.2,0.244704008102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,2,power_law_1.2,0.05826560258865356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,2,power_law_1.2,0.02428800016641617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,2,power_law_1.2,0.25665280818939207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,2,power_law_1.2,0.05923200249671936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,2,power_law_1.2,0.3716928005218506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,2,power_law_1.2,0.02983039915561676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,2,power_law_1.2,0.3564800024032593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,2,power_law_1.2,0.06682239770889283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,2,power_law_1.2,0.03059839904308319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,2,power_law_1.2,0.44048638343811036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,2,power_law_1.2,0.4702911853790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,2,balanced,0.6645653247833252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,2,power_law_1.2,0.06956160068511963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,2,power_law_1.2,0.028038400411605834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,2,power_law_1.2,0.6149824142456055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,2,power_law_1.2,0.6645311832427978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,2,power_law_1.2,0.07276800274848938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,2,power_law_1.2,0.03155199885368347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,2,power_law_1.2,0.8662655830383301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,2,power_law_1.2,0.9091903686523437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,2,power_law_1.2,0.08447999954223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,2,power_law_1.2,0.03424000144004822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,2,power_law_1.2,0.9333503723144532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,2,power_law_1.2,1.0781760215759277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.2,0.10127999782562255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,2,power_law_1.2,0.04058879911899567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,2,balanced,0.995306650797526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.2,0.1223423957824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,2,power_law_1.2,1.6936960220336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,2,power_law_1.2,0.043635201454162595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,2,power_law_1.2,1.6913087844848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.2,0.14203519821166993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.2,0.04647679924964905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,2,balanced,0.8215893109639486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,2,power_law_1.2,3.2775295257568358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,2,power_law_1.2,3.439846420288086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.2,0.18686720132827758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.2,0.06664959788322448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.2,0.22721281051635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.2,0.07504640221595764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.2,0.35207040309906007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.2,0.09939200282096863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.2,0.4391808032989502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.2,0.12387839555740357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.2,0.7008448123931885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.2,0.1709247946739197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.2,0.20933759212493896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.2,0.8490303993225098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.2,0.3061631917953491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,2,power_law_1.2,0.9841216087341309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,2,balanced,1.2695573170979817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.2,0.3925312042236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,2,power_law_1.2,1.762950325012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,2,power_law_1.2,0.48839678764343264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,2,power_law_1.2,3.5468414306640623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,2,power_law_1.2,0.7587584018707275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,2,balanced,1.9327306747436523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,2,power_law_1.2,1.4838080406188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,2,power_law_1.01,0.0328000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,2,power_law_1.01,0.03282560110092163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,2,power_law_1.01,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,2,power_law_1.01,0.03544960021972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,2,power_law_1.01,0.03555839955806732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,2,power_law_1.01,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,2,power_law_1.01,0.04097279906272888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,2,power_law_1.01,0.04317440092563629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,2,balanced,2.4710453351338706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,2,power_law_1.01,0.042182400822639465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,2,power_law_1.01,0.045151999592781066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,2,power_law_1.01,0.04769279956817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,2,power_law_1.01,0.03857280015945434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,2,power_law_1.01,0.0494592010974884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,2,power_law_1.01,0.040550398826599124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,2,power_law_1.01,0.05132799744606018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,2,power_law_1.01,0.044012799859046936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,2,power_law_1.01,0.05392000079154968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,2,power_law_1.01,0.05118079781532288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,2,power_law_1.01,0.06105599999427795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,2,power_law_1.01,0.05333120226860046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,2,power_law_1.01,0.06692479848861695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,2,power_law_1.01,0.05560320019721985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,2,power_law_1.01,0.07052800059318542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,2,power_law_1.01,0.056831997632980344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,2,power_law_1.01,0.08846719861030579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,2,power_law_1.01,0.05854079723358154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,2,power_law_1.01,0.05809919834136963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,2,power_law_1.01,0.0991807997226715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,2,power_law_1.01,0.05911039710044861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,2,power_law_1.01,0.12905600070953369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,2,power_law_1.01,0.059961599111557004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,2,power_law_1.01,0.15783679485321045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,2,power_law_1.01,0.06156799793243408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,2,power_law_1.01,0.21751039028167723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,2,power_law_1.01,0.06344320178031922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,2,power_law_1.01,0.23968639373779296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,2,power_law_1.01,0.06730239987373351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,2,power_law_1.01,0.3734911918640137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,2,power_law_1.01,0.0763584017753601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,2,power_law_1.01,0.4746367931365967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,2,power_law_1.01,0.07812479734420777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,2,power_law_1.01,0.6650688171386718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,2,power_law_1.01,0.07699840068817139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,2,power_law_1.01,0.09786880016326904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,2,power_law_1.01,0.9155903816223144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,2,power_law_1.01,0.11061760187149047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,2,power_law_1.01,0.12741119861602784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,2,power_law_1.01,1.1876352310180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,2,power_law_1.01,0.14866559505462645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,2,power_law_1.01,1.5763456344604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,2,power_law_1.01,0.22327039241790772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,2,power_law_1.01,2.9264448165893553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,2,power_law_1.01,0.23160319328308104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,2,power_law_1.01,0.3245503902435303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,2,power_law_1.01,0.3968384027481079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,2,power_law_1.01,0.5891071796417237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,2,power_law_1.01,0.7785727977752686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,2,power_law_1.01,0.8592512130737304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,2,power_law_1.01,1.4001279830932618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,2,power_law_1.01,2.9737600326538085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,1,balanced,0.019093333433071773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,1,balanced,0.019248000035683315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,1,balanced,0.020367999871571858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,1,balanced,0.01950399950146675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,1,balanced,0.039642666776975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,1,balanced,0.019445333629846573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,1,balanced,0.021125334004561108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,1,balanced,0.03955200066169103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,1,balanced,0.02120000123977661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,1,balanced,0.04179200033346812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,1,balanced,0.021082667013009388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,1,power_law_1.01,0.017510400712490083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,1,balanced,0.04301866888999939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,1,balanced,0.023034666975339253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,1,balanced,0.04367466767628988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,1,power_law_1.01,0.017689600586891174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,1,balanced,0.04390933116277059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,1,balanced,0.02317333221435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,1,power_law_1.01,0.01857919991016388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,1,balanced,0.04377066592375437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,1,balanced,0.02510400116443634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,1,power_law_1.01,0.018182399868965148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,1,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,1,power_law_1.01,0.01845120042562485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,1,balanced,0.02517866591612498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,1,balanced,0.045738667249679565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,1,power_law_1.01,0.019462400674819948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,1,balanced,0.025194667279720306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,1,balanced,0.04417066772778829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,1,power_law_1.01,0.020204800367355346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,1,balanced,0.045941332976023354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,1,power_law_1.01,0.020051200687885285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,1,balanced,0.027552001178264618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,1,balanced,0.047151997685432434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,1,power_law_1.01,0.02162559926509857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,1,balanced,0.033333333830038704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,1,balanced,0.049728001157442726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,1,power_law_1.01,0.021516799926757812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,1,balanced,0.04993600149949392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,1,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,1,power_law_1.01,0.022944000363349915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,1,balanced,0.059952000776926674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,1,balanced,0.033285332222779594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,1,power_law_1.01,0.02398719936609268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,1,balanced,0.05958933134873708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,1,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,1,power_law_1.01,0.02428800016641617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,1,balanced,0.06410133341948192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,1,balanced,0.04788800080617269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,1,power_law_1.01,0.02629759907722473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,1,balanced,0.0742986649274826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,1,power_law_1.01,0.031865599751472476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,1,balanced,0.055861334005991616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,1,balanced,0.09072533249855042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,1,power_law_1.01,0.03412480056285858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,1,balanced,0.06372799972693126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,1,balanced,0.1095199982325236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,1,power_law_1.01,0.031193599104881287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,1,balanced,0.08784533540407817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,1,balanced,0.13089600205421448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,1,power_law_1.01,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,1,balanced,0.10178132851918538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,1,balanced,0.17603733142217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.01,0.04601599872112274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.01,0.05308160185813904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,1,balanced,0.20977065960566202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,1,balanced,0.13850133617719015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.01,0.061887997388839724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.01,0.0866815984249115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,1,power_law_1.01,0.03715200126171112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,1,balanced,0.29341334104537964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.01,0.09902719855308532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,1,balanced,0.17399466037750244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,1,power_law_1.01,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.01,0.1351359963417053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,1,power_law_1.01,0.038438400626182555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.01,0.17134079933166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,1,power_law_1.01,0.041440001130104064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,1,balanced,0.3801013231277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.01,0.24304640293121338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,1,power_law_1.01,0.04200319945812225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,1,balanced,0.2473120093345642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.01,0.30519039630889894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,1,power_law_1.01,0.0430976003408432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,1,power_law_1.01,0.04325119853019714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.01,0.37833600044250487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,1,power_law_1.01,0.043136000633239746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,1,power_law_1.01,0.5855167865753174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,1,balanced,0.5442399978637695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,1,power_law_1.01,0.04410879909992218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,1,power_law_1.01,1.1424320220947266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,1,power_law_1.01,0.04540160000324249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,1,balanced,0.3086293339729309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,1,power_law_1.01,0.04672000110149384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,1,power_law_1.01,0.05057280063629151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,1,power_law_1.01,0.05086719989776611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,1,balanced,0.7014933427174886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,1,power_law_1.01,0.05322880148887634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,1,power_law_1.01,0.05985280275344849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,1,power_law_1.01,0.06267520189285278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,1,balanced,0.3816479841868083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,1,power_law_1.01,0.06751360297203064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,1,power_law_1.01,0.07674880027770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.01,0.09217919707298279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,1,balanced,0.8649280071258545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.01,0.11074559688568116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.01,0.133951997756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.01,0.17520639896392823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.01,0.21062400341033935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.01,0.2931328058242798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,1,balanced,0.5887200037638346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.01,0.3764480113983154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.01,0.5392064094543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,1,balanced,1.3393492698669434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.01,0.7027328014373779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.01,0.865715217590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,1,power_law_1.01,1.3352831840515136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,1,power_law_1.01,2.614476776123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,1,balanced,1.1468746662139893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,1,balanced,2.6235146522521973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,1,power_law_1.2,0.0190528005361557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,1,power_law_1.2,0.04104959964752197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,1,power_law_1.2,0.018617600202560425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,1,power_law_1.2,0.043673598766326906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,1,power_law_1.2,0.01831679940223694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,1,power_law_1.2,0.044198399782180785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,1,power_law_1.2,0.019865599274635316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,1,power_law_1.2,0.053958398103713986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,1,power_law_1.2,0.019526399672031403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,1,power_law_1.2,0.05588480234146118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,1,power_law_1.2,0.01953279972076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,1,power_law_1.2,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,1,power_law_1.2,0.02033279985189438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,1,power_law_1.2,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,1,power_law_1.2,0.02072319984436035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,1,power_law_1.2,0.054816001653671266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,1,power_law_1.2,0.021081599593162536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,1,power_law_1.2,0.05539839863777161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,1,power_law_1.2,0.02147199958562851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,1,power_law_1.2,0.05479679703712463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,1,power_law_1.2,0.023206399381160737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,1,power_law_1.2,0.05553280115127564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,1,power_law_1.2,0.023027199506759643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,1,power_law_1.2,0.058323198556900026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,1,power_law_1.2,0.024006399512290954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,1,power_law_1.2,0.05864319801330566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,1,power_law_1.2,0.02573440074920654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,1,balanced,0.03376533339420954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,1,power_law_1.2,0.06120960116386413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,1,power_law_1.2,0.03163520097732544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,1,balanced,0.033285332222779594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,1,power_law_1.2,0.06807039976119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,1,power_law_1.2,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,1,balanced,0.035530666510264076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,1,power_law_1.2,0.031513598561286923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,1,power_law_1.2,0.0701312005519867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,1,balanced,0.0353973334034284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,1,power_law_1.2,0.03825919926166534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,1,balanced,0.053786665201187134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,1,power_law_1.2,0.07770879864692688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.2,0.0452672004699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,1,balanced,0.056159997979799904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,1,power_law_1.2,0.08852480053901672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,1,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.2,0.053439998626708986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,1,balanced,0.055813332398732506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,1,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.2,0.06247680187225342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.2,0.10259840488433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.2,0.08591359853744507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,1,balanced,0.05607999861240387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,1,balanced,0.03786666691303253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.2,0.12340480089187622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.2,0.09932799935340882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,1,balanced,0.05622933308283488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,1,balanced,0.03966933240493139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.2,0.1456447958946228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.2,0.1351807951927185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,1,balanced,0.03961600114901861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,1,balanced,0.05607999861240387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.2,0.1715648055076599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.2,0.19160959720611573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,1,balanced,0.0395359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,1,balanced,0.05729599793752035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.2,0.24405760765075685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.2,0.2403264045715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,1,balanced,0.04164800047874451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,1,balanced,0.0580213318268458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.2,0.3052160024642944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,1,balanced,0.043621331453323364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.2,0.37849600315093995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,1,balanced,0.05648533503214518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.2,0.3332672119140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,1,power_law_1.2,0.5853631973266602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,1,balanced,0.0436160018046697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,1,balanced,0.05830933153629303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.2,0.4251584053039551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,1,power_law_1.2,1.143123245239258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,1,balanced,0.045696000258127846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,1,balanced,0.059749335050582886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.2,0.6086463928222656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,1,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,1,balanced,0.06358399987220764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.2,0.7927487850189209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,1,balanced,0.0517493337392807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,1,balanced,0.07222400108973186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.2,0.9831295967102051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,1,balanced,0.06379733482996623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,1,balanced,0.07282133400440216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,1,power_law_1.2,1.5426176071166993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,1,balanced,0.08043733239173889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,1,balanced,0.06642666459083557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,1,power_law_1.2,3.0162368774414063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,1,power_law_1.01,0.03235200047492981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,1,balanced,0.0942133367061615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,1,balanced,0.07663466533025105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,1,balanced,0.10566932956377666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,1,balanced,0.09181333581606548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,1,power_law_1.01,0.032915198802948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,1,balanced,0.1381333371003469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,1,power_law_1.01,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,1,balanced,0.10322667161623637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,1,power_law_1.01,0.0374208003282547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,1,balanced,0.1325920025507609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,1,balanced,0.14429333806037903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,1,power_law_1.01,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,1,power_law_1.01,0.03988479971885681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,1,balanced,0.15314666430155435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,1,balanced,0.19289066394170126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,1,power_law_1.01,0.040966400504112245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,1,power_law_1.01,0.042847999930381776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,1,power_law_1.01,0.04322560131549835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,1,balanced,0.22915732860565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,1,balanced,0.21572800477345785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,1,power_law_1.01,0.045952001214027406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,1,balanced,0.3173919916152954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,1,balanced,0.258517324924469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,1,power_law_1.01,0.046598398685455324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,1,balanced,0.4075253407160441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,1,power_law_1.01,0.049030399322509764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,1,power_law_1.01,0.05189120173454285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,1,balanced,0.3654559850692749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,1,power_law_1.01,0.05495039820671081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,1,power_law_1.01,0.06325119733810425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,1,balanced,0.5824533303578695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,1,power_law_1.01,0.06622719764709473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,1,power_law_1.01,0.034092798829078674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,1,power_law_1.01,0.0766975998878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,1,balanced,0.4667786757151286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,1,power_law_1.01,0.09074559807777405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,1,power_law_1.01,0.03726719915866852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,1,power_law_1.01,0.04036479890346527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,1,power_law_1.01,0.10586240291595458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,1,balanced,0.733125368754069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,1,power_law_1.01,0.05359359979629517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,1,power_law_1.01,0.05306239724159241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,1,balanced,0.67740265528361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,1,power_law_1.01,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,1,power_law_1.01,0.13390719890594482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,1,power_law_1.01,0.055027198791503903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,1,power_law_1.01,0.05711359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,1,power_law_1.01,0.15166720151901245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,1,power_law_1.01,0.058316802978515624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,1,balanced,0.9061493078867594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,1,power_law_1.01,0.21742079257965088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,1,power_law_1.01,0.0596671998500824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,1,balanced,0.874901294708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,1,power_law_1.01,0.05902720093727112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,1,power_law_1.01,0.25768959522247314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,1,power_law_1.01,0.06279039978981019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,1,power_law_1.01,0.3616192102432251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,1,power_law_1.01,0.06335999965667724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,1,power_law_1.01,0.0675711989402771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,1,power_law_1.01,0.4676032066345215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,1,power_law_1.01,0.07828480005264282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,1,power_law_1.01,0.6746560096740722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,1,power_law_1.01,0.08152959942817688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,1,power_law_1.01,0.8828607559204101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,1,balanced,1.0836053689320881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,1,power_law_1.01,0.08142079710960388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,1,balanced,1.4046986897786458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,1,power_law_1.01,1.082476806640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,1,power_law_1.01,0.09228159785270691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,1,power_law_1.01,1.7018560409545898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,1,power_law_1.01,0.11030399799346924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,1,power_law_1.01,0.13770240545272827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,1,power_law_1.01,3.3204929351806642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,1,power_law_1.01,0.14876799583435057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,1,power_law_1.01,0.19806720018386842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,1,power_law_1.01,0.2360896110534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,1,power_law_1.01,0.31957120895385743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,1,balanced,1.6876160303751628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,1,power_law_1.01,0.4406015872955322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,1,power_law_1.01,0.5943808078765869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,1,power_law_1.01,0.768064022064209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,1,power_law_1.01,0.9373824119567871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,1,power_law_1.01,1.5460479736328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,1,balanced,2.7550719579060874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,1,power_law_1.01,3.1182655334472655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,1,balanced,3.321040153503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,1,power_law_1.2,0.033907198905944826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,1,power_law_1.2,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,1,power_law_1.2,0.03935360014438629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,1,power_law_1.2,0.05317760109901428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,1,power_law_1.2,0.053504002094268796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,1,power_law_1.2,0.05482879877090454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,1,power_law_1.2,0.05491840243339539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,1,power_law_1.2,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,1,power_law_1.2,0.057811200618743896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,1,power_law_1.2,0.05841919779777527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,1,power_law_1.2,0.0592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,1,power_law_1.2,0.061440002918243405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,1,power_law_1.2,0.0627839982509613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,1,power_law_1.2,0.06849920153617858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,1,power_law_1.2,0.07650560140609741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,1,power_law_1.2,0.07868160009384155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,1,power_law_1.2,0.08065279722213745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,1,power_law_1.2,0.0921280026435852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,1,power_law_1.2,0.11118079423904419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,1,power_law_1.2,0.13527040481567382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,1,power_law_1.2,0.03204480111598969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,1,power_law_1.2,0.1492735981941223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,1,power_law_1.2,0.1941375970840454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,1,power_law_1.2,0.03294079899787903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,1,power_law_1.2,0.23141119480133057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,1,power_law_1.2,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,1,power_law_1.2,0.3192512035369873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,1,power_law_1.2,0.03648000061511993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,1,power_law_1.2,0.03747200071811676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,1,power_law_1.2,0.40633602142333985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,1,power_law_1.2,0.040524798631668094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,1,power_law_1.2,0.5818111896514893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,1,power_law_1.2,0.04245760142803192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,1,power_law_1.2,0.7566976070404052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,1,power_law_1.2,0.04106239974498749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,1,power_law_1.2,0.9063360214233398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,1,power_law_1.2,0.04429439902305603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,1,power_law_1.2,1.4236415863037108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,1,power_law_1.2,0.047167998552322385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,1,power_law_1.2,2.752992057800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,1,power_law_1.2,0.04613119959831238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,1,power_law_1.2,0.04935039877891541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,1,power_law_1.2,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,1,power_law_1.2,0.053388798236846925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,1,power_law_1.2,0.06282240152359009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,1,power_law_1.2,0.06519039869308471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,32,balanced,0.04985066751639048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,1,power_law_1.2,0.07753599882125854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,1,power_law_1.2,0.0905023992061615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,32,balanced,0.053861334919929504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,1,power_law_1.2,0.10501760244369507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,32,balanced,0.05154666801293691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,1,power_law_1.2,0.13181439638137818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,32,balanced,0.053871999184290566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,1,power_law_1.2,0.153766405582428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,32,balanced,0.06218666831652323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,32,balanced,0.08410132924715678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,1,power_law_1.2,0.21322240829467773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,32,balanced,0.08620267113049825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,1,power_law_1.2,0.2578752040863037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,32,balanced,0.08661866188049316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,1,power_law_1.2,0.3621887922286987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,32,balanced,0.08683199683825175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,1,power_law_1.2,0.4662911891937256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,32,balanced,0.08690667152404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,1,power_law_1.2,0.6745984077453613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,32,balanced,0.08888000249862671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,1,power_law_1.2,0.8858688354492188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,32,balanced,0.08758933345476787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,1,power_law_1.2,1.0832703590393067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,32,balanced,0.09034132957458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,1,power_law_1.2,1.6989952087402345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,32,balanced,0.0915786623954773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,32,balanced,0.09671466549237569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,1,power_law_1.2,3.325356674194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,32,balanced,0.09641066193580627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,32,balanced,0.1002293328444163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,32,balanced,0.10774399836858113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,32,balanced,0.1122773289680481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,32,balanced,0.045882667104403176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,32,balanced,0.1274773379166921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,32,balanced,0.04692266881465912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,32,balanced,0.041759997606277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,32,balanced,0.14181333780288696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,32,balanced,0.042912001411120095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,32,balanced,0.0465280016263326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,32,balanced,0.05411200225353241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,32,balanced,0.16699733336766562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,32,balanced,0.058517331878344216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,32,balanced,0.06062399844328562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,32,balanced,0.06740800042947133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,32,balanced,0.19103999932607016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,32,balanced,0.06705066561698914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,32,balanced,0.07100266714890797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,32,balanced,0.08286400139331818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,32,balanced,0.08509332935015361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,32,balanced,0.12777066230773926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,32,balanced,0.24680533011754355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,32,balanced,0.16022400061289468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,32,balanced,0.16064533591270447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,32,balanced,0.19341333707173666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,32,balanced,0.26065067450205487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,32,balanced,0.3087679942448934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,32,balanced,0.29656000932057697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,32,balanced,0.44087998072306317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,32,balanced,0.24065599838892618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,32,balanced,0.31006399790445965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,32,balanced,0.4036639928817749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,32,balanced,0.39908266067504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,32,balanced,0.5547786553700765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,32,balanced,0.7285066445668539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,32,balanced,0.5079520146052042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,32,balanced,1.0929653644561768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,32,balanced,0.6206186612447103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,32,balanced,1.4254825909932454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,32,balanced,1.7875359853108723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,32,balanced,0.9277919928232828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,32,balanced,0.04948266843954722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,32,balanced,0.047637333472569786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,32,balanced,0.04789333542188009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,32,balanced,2.8102238972981772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,32,balanced,0.05178666611512502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,32,power_law_1.01,0.08164479732513427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,32,balanced,0.07363733152548473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,32,balanced,0.10090133547782898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,32,power_law_1.01,0.0779263973236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,32,balanced,0.10288000106811523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,32,power_law_1.01,0.06492159962654113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,32,balanced,0.10085866848627727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,32,power_law_1.01,0.061337602138519284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,32,balanced,0.10291199882825215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,32,balanced,0.06982933481534322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,32,balanced,0.10430933038393657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,32,power_law_1.01,0.07191680073738098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,32,balanced,0.10494933525721233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,32,power_law_1.01,0.07616000175476074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,32,balanced,0.1069546639919281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,32,power_law_1.01,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,32,balanced,0.05314666529496511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,32,balanced,0.10929066936175029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,32,power_law_1.01,0.08204159736633301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,32,balanced,0.050154666105906166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,32,balanced,0.1114026705423991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,32,power_law_1.01,0.08118399977684021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,32,balanced,0.07337066531181335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,32,balanced,0.11512000362078349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,32,power_law_1.01,0.08249599933624267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,32,balanced,0.09862400094668071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,32,balanced,0.11924800276756287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,32,power_law_1.01,0.0828927993774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,32,balanced,0.14917332927385965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,32,balanced,0.1202133297920227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,32,power_law_1.01,0.08615040183067321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,32,power_law_1.01,0.04848639965057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,32,balanced,0.15010133385658264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,32,power_law_1.01,0.08569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,32,balanced,0.1336426635583242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,32,balanced,0.15340800086657205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,32,power_law_1.01,0.08862079977989197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,32,power_law_1.01,0.04933120012283325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,32,balanced,0.15127467115720114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,32,balanced,0.13871999581654867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,32,power_law_1.01,0.09403520226478576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,32,balanced,0.15465600291887918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,32,power_law_1.01,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,32,balanced,0.15153066317240396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,32,power_law_1.01,0.09572479724884034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,32,power_law_1.01,0.03984639942646027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,32,balanced,0.160863995552063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,32,balanced,0.15310933192571005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,32,power_law_1.01,0.10128639936447144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,32,power_law_1.01,0.04462080001831055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,32,balanced,0.1537493367989858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,32,balanced,1.7913653055826824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,32,power_law_1.01,0.11390719413757325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,32,power_law_1.01,0.05073919892311096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,32,balanced,0.1564906636873881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,32,balanced,0.17609065771102905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.01,0.11845120191574096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,32,balanced,0.15970133741696677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,32,power_law_1.01,0.06539520025253295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.01,0.14421759843826293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,32,balanced,0.1602826714515686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,32,power_law_1.01,0.06738560199737549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,32,balanced,0.1627253293991089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,32,balanced,0.21308799584706625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.01,0.16293760538101196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,32,power_law_1.01,0.09555839896202087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,32,balanced,0.16954133907953897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,32,power_law_1.01,0.0956928014755249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.01,0.1968384027481079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,32,power_law_1.01,0.09573760032653808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,32,balanced,0.17385600010553995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.01,0.23238399028778076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,32,balanced,0.2500373323758443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,32,power_law_1.01,0.10586240291595458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.01,0.31722240447998046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,32,balanced,0.1845653255780538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,32,power_law_1.01,0.10355199575424194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.01,0.41690239906311033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,32,power_law_1.01,0.10458879470825196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,32,balanced,0.1970026691754659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.01,0.5964416027069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,32,power_law_1.01,0.10764800310134888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,32,balanced,0.3259626626968384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.01,0.7163008213043213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,32,power_law_1.01,0.09845759868621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,32,balanced,0.22032533089319864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,32,power_law_1.01,0.8895423889160157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,32,power_law_1.01,0.10618239641189575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,32,power_law_1.01,0.11125760078430176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,32,power_law_1.01,1.3744064331054688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.01,0.1086527943611145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,32,balanced,0.24734399716059366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.01,0.11808639764785767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,32,balanced,0.3912320137023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.01,0.1352959990501404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,32,power_law_1.01,2.6689983367919923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.01,0.17932159900665284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.01,0.22506239414215087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,32,balanced,0.30513066053390503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.01,0.29925758838653566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,32,balanced,5.581658681233724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.01,0.4081727981567383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.01,0.6136896133422851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.01,0.7545983791351318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,32,power_law_1.01,0.09767040014266967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,32,balanced,0.5438666741053263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,32,power_law_1.01,0.9513343811035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,32,power_law_1.01,0.10687359571456909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,32,balanced,0.3429653247197469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,32,power_law_1.01,0.07386879920959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,32,power_law_1.01,1.5824064254760741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,32,power_law_1.01,0.06983680129051209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,32,power_law_1.01,0.08106880187988282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,32,power_law_1.01,3.20880012512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,32,power_law_1.01,0.08364160060882568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,32,power_law_1.01,0.09317119717597962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,32,power_law_1.01,0.09535999894142151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,32,balanced,0.4516959985097249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,32,balanced,0.6841066678365072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,32,power_law_1.01,0.09997439980506898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,32,power_law_1.01,0.09686400294303894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,32,power_law_1.01,0.09932159781455993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,32,power_law_1.01,0.10265599489212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,32,power_law_1.01,0.10601600408554077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,32,power_law_1.01,0.10788480043411255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,32,power_law_1.01,0.11549439430236816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,32,power_law_1.01,0.11980160474777221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,32,power_law_1.01,0.12297600507736206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,32,power_law_1.01,0.14865920543670655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,32,balanced,0.5419040123621622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,32,power_law_1.01,0.15101439952850343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,32,power_law_1.01,0.1982143998146057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,32,balanced,0.8543039957682291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,32,power_law_1.01,0.2153536081314087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,32,power_law_1.01,0.2776639938354492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,32,power_law_1.01,0.30884480476379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,32,power_law_1.01,0.16637439727783204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,32,power_law_1.01,0.4384511947631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,32,power_law_1.01,0.5476287841796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,32,power_law_1.01,0.14762239456176757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,32,power_law_1.01,0.07831680178642272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,32,power_law_1.01,0.7275455951690674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,32,power_law_1.01,0.09689599871635438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,32,power_law_1.01,0.944268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,32,power_law_1.01,0.10674560070037842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,32,balanced,0.6849760214487711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,32,power_law_1.01,0.12424319982528687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,32,power_law_1.01,1.1918272018432616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,32,power_law_1.01,0.13111679553985595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,32,power_law_1.01,1.8317632675170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,32,power_law_1.01,0.13804160356521605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,32,power_law_1.01,0.14433280229568482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,32,power_law_1.01,3.5867198944091796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,32,power_law_1.01,0.13849600553512573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,32,power_law_1.01,0.13148800134658814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,32,power_law_1.01,0.14614399671554565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,32,balanced,1.2583146890004475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,32,power_law_1.01,0.14574719667434693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,32,power_law_1.01,0.1535423994064331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,32,power_law_1.01,0.1609984040260315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,32,power_law_1.01,0.160697603225708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,32,power_law_1.01,0.16837120056152344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,32,power_law_1.01,0.19726719856262206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,32,power_law_1.01,0.17843199968338014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,32,power_law_1.01,0.2300031900405884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,32,power_law_1.01,0.24197759628295898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,32,power_law_1.01,0.306278395652771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,32,power_law_1.01,0.3318527936935425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,32,balanced,0.9358932971954346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,32,power_law_1.01,0.46681599617004393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,32,power_law_1.01,0.5434879779815673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,32,power_law_1.01,0.7702911853790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,32,power_law_1.01,0.9925375938415527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,32,power_law_1.01,1.1635647773742677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,32,power_law_1.01,2.0060543060302733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,32,power_law_1.01,3.5895233154296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,32,balanced,2.4619040489196777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,32,power_law_1.2,0.14434560537338256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,32,power_law_1.2,0.14249600172042848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,32,power_law_1.2,0.07524480223655701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,32,power_law_1.2,0.09793279767036438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,32,power_law_1.2,0.10255999565124511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,32,power_law_1.2,0.1013759970664978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,32,power_law_1.2,0.12200959920883178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,32,power_law_1.2,0.13125120401382445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,32,power_law_1.2,0.14126720428466796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,32,power_law_1.2,0.1410815954208374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,32,power_law_1.2,0.1365504026412964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,32,power_law_1.2,0.14454400539398193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,32,power_law_1.2,0.14856319427490233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,32,power_law_1.2,0.1499008059501648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,32,balanced,1.8208319346110027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,32,power_law_1.2,0.15764479637145995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,32,power_law_1.2,0.09771519899368286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,32,power_law_1.2,0.16005760431289673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,32,power_law_1.2,0.18725759983062745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,32,power_law_1.2,0.10245120525360107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,32,power_law_1.2,0.2155519962310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,32,power_law_1.2,0.08110079765319825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,32,power_law_1.2,0.05457280278205871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,32,power_law_1.2,0.2089855909347534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,32,power_law_1.2,0.06569600105285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,32,power_law_1.2,0.25978879928588866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,32,power_law_1.2,0.07822080254554749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,32,power_law_1.2,0.07853440046310425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,32,power_law_1.2,0.2794111967086792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,32,power_law_1.2,0.09075199961662292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,32,power_law_1.2,0.051072001457214355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,32,power_law_1.2,0.09433599710464477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,32,power_law_1.2,0.3847872018814087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,32,power_law_1.2,0.09091839790344239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,32,power_law_1.2,0.06499840021133423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,32,power_law_1.2,0.411513614654541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,32,power_law_1.2,0.0946175992488861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,32,power_law_1.2,0.06819199919700622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,32,power_law_1.2,0.09932799935340882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,32,power_law_1.2,0.07434880137443542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,32,power_law_1.2,0.09953280091285706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,32,power_law_1.2,0.5417535781860352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,32,power_law_1.2,0.07832319736480713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,32,power_law_1.2,0.10075520277023316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,32,power_law_1.2,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,32,power_law_1.2,0.7029823780059814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,32,power_law_1.2,0.10200320482254029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,32,power_law_1.2,0.04855040013790131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,32,power_law_1.2,0.07957760095596314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,32,power_law_1.2,0.8599360466003418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,32,power_law_1.2,0.1128767967224121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,32,power_law_1.2,0.11610239744186401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,32,power_law_1.2,0.08186240196228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,32,power_law_1.2,0.049516800045967105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,32,power_law_1.2,1.2108799934387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,32,power_law_1.2,0.12336000204086303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,32,power_law_1.2,0.08390399813652039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,32,power_law_1.2,0.03610239923000336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,32,power_law_1.2,0.13124480247497558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,32,power_law_1.2,1.5971455574035645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,32,power_law_1.2,0.0855679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,32,power_law_1.2,0.15886080265045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,32,power_law_1.2,0.042828801274299624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,32,power_law_1.2,0.16081279516220093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,32,power_law_1.2,2.6638784408569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,32,power_law_1.2,0.041228801012039185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,32,power_law_1.2,0.08706560134887695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,32,power_law_1.2,0.20439040660858154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,32,power_law_1.2,0.048230400681495665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,32,power_law_1.2,0.22649600505828857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,32,power_law_1.2,0.07329919934272766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,32,power_law_1.2,0.08954240083694458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,32,power_law_1.2,4.391987228393555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,32,power_law_1.2,0.3227839946746826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,32,power_law_1.2,0.0713536024093628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,32,power_law_1.2,0.3456768035888672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,32,power_law_1.2,0.09596160054206848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,32,power_law_1.2,0.09544960260391236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,32,power_law_1.2,0.48227200508117674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,32,power_law_1.2,0.6430016040802002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,32,power_law_1.2,0.09537280201911927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,32,power_law_1.2,0.09829760193824769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,32,power_law_1.2,0.9305472373962402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,32,power_law_1.2,0.09605759978294373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,32,power_law_1.2,1.0955391883850099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,32,power_law_1.2,0.107315194606781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,32,power_law_1.2,0.10151040554046631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,32,power_law_1.2,0.10508160591125489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,32,power_law_1.2,1.3857088088989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,32,power_law_1.2,0.10140160322189332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,32,power_law_1.2,0.11841280460357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,32,power_law_1.2,2.175148773193359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,32,power_law_1.2,0.10290559530258178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.2,0.13251839876174926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,32,power_law_1.2,0.10314240455627441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,32,power_law_1.2,4.3956352233886715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.2,0.15348479747772217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,32,power_law_1.2,0.10115840435028076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.2,0.17515519857406617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,32,power_law_1.2,0.11052800416946411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.2,0.2132159948348999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.2,0.10232959985733033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.2,0.27820160388946535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.2,0.1359231948852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.2,0.3750272035598755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.2,0.14883840084075928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.2,0.4611328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.2,0.1982848048210144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.2,0.7039231777191162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.2,0.22985599040985108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.2,0.8882304191589355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.2,0.34471681118011477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.2,0.4388927936553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,32,power_law_1.2,1.1630528450012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.2,0.6110208034515381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,32,power_law_1.2,1.8594175338745118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.2,0.9278016090393066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,32,power_law_1.2,3.5073856353759765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,32,power_law_1.2,1.1219840049743652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,32,power_law_1.2,1.8353792190551759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,32,power_law_1.2,3.73372802734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,16,balanced,0.05187733471393585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,16,balanced,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,16,balanced,0.05197866757710775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,16,balanced,0.05589866638183594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,16,balanced,0.06482666730880737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,16,balanced,0.08515733480453491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,16,balanced,0.02550400048494339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,16,balanced,0.08846933643023173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,16,balanced,0.08875200152397156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,16,balanced,0.02516799916823705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,16,balanced,0.08934932947158813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,16,balanced,0.02565866708755493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,16,balanced,0.08890133102734883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,16,balanced,0.027061333258946735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,16,balanced,0.040591999888420105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,16,balanced,0.09212266405423482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,16,balanced,0.05710400144259135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,16,balanced,0.09081600109736125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,16,balanced,0.06200533111890157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,16,balanced,0.09434133768081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,16,balanced,0.06632000207901001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,16,balanced,0.09633066256841023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,16,balanced,0.06737599770228068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,16,balanced,0.10084799925486247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,16,balanced,0.07029866675535838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,16,balanced,0.10064533352851868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,16,balanced,0.07443200051784515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,16,balanced,0.1053546667098999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,16,balanced,0.07784000039100647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,16,balanced,0.08230400085449219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,16,balanced,0.11500799655914307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,16,balanced,0.08284799754619598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,16,balanced,0.12071999907493591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,16,balanced,0.08845333258310954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,16,balanced,0.09337600072224934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,16,balanced,0.13567466537157694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,16,balanced,0.15575466553370157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,16,balanced,0.15432000160217285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,16,balanced,0.1740000049273173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,16,balanced,0.21205333868662515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,16,balanced,0.18175999323527017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,16,balanced,0.2851039965947469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,16,balanced,0.19298666715621948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,16,balanced,0.2113920052846273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,16,power_law_1.01,0.060499197244644164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,16,power_law_1.01,0.027820798754692077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,16,balanced,0.23882132768630981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,16,power_law_1.01,0.08876799941062927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,16,power_law_1.01,0.04962559938430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,16,balanced,0.27963199218114215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,16,power_law_1.01,0.03412480056285858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,16,power_law_1.01,0.05500159859657287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,16,balanced,0.2965173323949178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,16,power_law_1.01,0.029523199796676634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,16,power_law_1.01,0.05996800065040588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,16,power_law_1.01,0.03943040072917938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,16,power_law_1.01,0.06498559713363647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,16,power_law_1.01,0.04094719886779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,16,power_law_1.01,0.07454079985618592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,16,power_law_1.01,0.0490880012512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,16,power_law_1.01,0.08069760203361512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,16,power_law_1.01,0.055315202474594115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,16,power_law_1.01,0.07999359965324401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,16,power_law_1.01,0.06926079988479614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,16,power_law_1.01,0.08195840120315552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,16,balanced,0.3845706780751546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,16,balanced,0.337066650390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,16,power_law_1.01,0.0736191987991333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,16,power_law_1.01,0.08564479947090149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,16,power_law_1.01,0.06774399876594543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,16,power_law_1.01,0.08501759767532349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,16,power_law_1.01,0.09538559913635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,16,power_law_1.01,0.08938239812850952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,16,power_law_1.01,0.09632639884948731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,16,power_law_1.01,0.08979200124740601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,16,power_law_1.01,0.09669119715690613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,16,power_law_1.01,0.0913919985294342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,16,power_law_1.01,0.09921919703483581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,16,power_law_1.01,0.09572479724884034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,16,balanced,0.4857493241628011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,16,power_law_1.01,0.1030784010887146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,16,power_law_1.01,0.09869440197944641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,16,power_law_1.01,0.10177919864654542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,16,power_law_1.01,0.10546560287475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,16,power_law_1.01,0.10734720230102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,16,power_law_1.01,0.1212928056716919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.01,0.10733439922332763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,16,balanced,0.4620693524678548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.01,0.11562240123748779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.01,0.13258880376815796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.01,0.1259328007698059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.01,0.15815680027008056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.01,0.15947519540786742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.01,0.1713152050971985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.01,0.19913599491119385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.01,0.2541248083114624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,16,balanced,0.7495413621266683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.01,0.21860480308532715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.01,0.31201920509338377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.01,0.26813440322875975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.01,0.4265791893005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.01,0.3563136100769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.01,0.5926655769348145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.01,0.4346367835998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,16,power_law_1.01,0.7303743839263916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,16,balanced,0.5841600100199381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.01,0.5655295848846436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,16,power_law_1.01,1.1184255599975585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.01,0.7865087985992432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,16,power_law_1.01,0.967347240447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,16,power_law_1.01,2.312870407104492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,16,power_law_1.01,1.6486400604248046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,16,balanced,0.9547466437021891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,16,power_law_1.01,2.997331237792969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,16,balanced,0.7120693524678549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,16,balanced,1.2067999839782715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,16,balanced,1.08516263961792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,16,balanced,1.8742720286051433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,16,balanced,0.04612799982229868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,16,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,16,balanced,0.052111998200416565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,16,balanced,0.0466186652580897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,16,balanced,0.07075199981530507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,16,balanced,0.09710933764775594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,16,balanced,0.047877331574757896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,16,balanced,0.15109333395957947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,16,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,16,balanced,0.152346670627594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,16,balanced,0.04979733129342397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,16,balanced,0.1492959956328074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,16,balanced,0.07453333338101704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,16,balanced,0.14975466330846152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,16,balanced,0.10307733217875163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,16,balanced,0.15195733308792114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,16,balanced,0.10292266805966695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,16,balanced,0.15054933230082193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,16,balanced,0.10297066966692607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,16,balanced,0.15307199954986572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,16,balanced,0.1050933301448822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,16,balanced,0.1543839971224467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,16,balanced,0.10600533088048299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,16,balanced,0.15588800112406412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,16,balanced,0.10830932855606079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,16,balanced,0.1609760026137034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,16,balanced,0.1091306706269582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,16,balanced,2.109210650126139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,16,balanced,0.16328533490498862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,16,balanced,0.11241066455841064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,16,balanced,0.11135466893513997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,16,balanced,0.16473066806793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,16,balanced,0.11686933040618896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,16,balanced,0.1715786655743917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,16,balanced,0.12008532881736755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,16,balanced,0.17942400773366293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,16,balanced,0.12339199582735698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,16,balanced,3.712597211201986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,16,balanced,0.1359946628411611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,16,balanced,0.1913493275642395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,16,balanced,0.14330666263898215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,16,balanced,0.20600533485412598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,16,balanced,0.16389866669972739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,16,balanced,0.23427200317382812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,16,balanced,0.18872000773747763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,16,balanced,0.22748267650604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,16,balanced,0.2565973401069641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,16,balanced,0.26364266872406006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,16,balanced,0.3256266713142395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,16,balanced,0.3545973300933838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,16,balanced,0.36498133341471356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,16,balanced,0.4342133204142253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,16,balanced,0.5192960103352865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,16,balanced,0.6051733493804932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,16,balanced,0.5952586730321249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,16,balanced,0.7802293300628662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,16,power_law_1.2,0.0911616027355194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,16,power_law_1.2,0.07141759991645813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,16,power_law_1.2,0.13467520475387573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,16,power_law_1.2,0.09255679845809936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,16,power_law_1.2,0.05944960117340088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,16,power_law_1.2,0.05181440114974976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,16,power_law_1.2,0.071424001455307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,16,power_law_1.2,0.06190720200538635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,16,power_law_1.2,0.08856959939002991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,16,power_law_1.2,0.07091839909553528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,16,power_law_1.2,0.09685119986534119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,16,power_law_1.2,0.07584000229835511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,16,power_law_1.2,0.12767360210418702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,16,power_law_1.2,0.08714879751205444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,16,power_law_1.2,0.1248960018157959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,16,power_law_1.2,0.09590399861335755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,16,power_law_1.2,0.12908799648284913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,16,power_law_1.01,0.09548159837722778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,16,power_law_1.2,0.09468799829483032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,16,power_law_1.2,0.13857920169830323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,16,power_law_1.2,0.09605759978294373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,16,power_law_1.2,0.14298239946365357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,16,power_law_1.01,0.1471743941307068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,16,power_law_1.2,0.10088319778442383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,16,power_law_1.2,0.14165760278701783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,16,power_law_1.01,0.11918079853057861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,16,balanced,0.9503520329793295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,16,power_law_1.2,0.09928320050239563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,16,power_law_1.2,0.143993604183197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,16,power_law_1.01,0.08659840226173401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,16,balanced,0.760591983795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,16,power_law_1.01,0.07461760044097901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,16,power_law_1.2,0.10439039468765259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,16,power_law_1.2,0.15219839811325073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,16,power_law_1.01,0.091430401802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,16,power_law_1.2,0.15697280168533326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,16,power_law_1.01,0.10102399587631225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,16,power_law_1.2,0.11026560068130493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,16,power_law_1.01,0.10334080457687378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,16,power_law_1.2,0.15036159753799438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,16,power_law_1.01,0.12438399791717529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,16,power_law_1.2,0.11650559902191163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,16,power_law_1.01,0.06321920156478882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,16,power_law_1.2,0.1709247946739197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,16,power_law_1.01,0.13775999546051027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,16,power_law_1.2,0.1209663987159729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,16,power_law_1.2,0.20088319778442382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,16,power_law_1.01,0.06108160018920898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,16,power_law_1.01,0.12469760179519654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,16,power_law_1.2,0.21111679077148438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,16,power_law_1.2,0.12924799919128419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,16,power_law_1.01,0.06816639900207519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,16,power_law_1.2,0.27143039703369143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,16,power_law_1.01,0.13857280015945433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,16,power_law_1.2,0.144051194190979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,16,power_law_1.01,0.07833600044250488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,16,power_law_1.2,0.25824000835418703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,16,power_law_1.01,0.14415359497070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,16,power_law_1.2,0.15825920104980468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,16,power_law_1.2,0.32872960567474363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,16,power_law_1.01,0.09267200231552124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,16,power_law_1.01,0.14510079622268676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,16,power_law_1.2,0.18928639888763427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,16,power_law_1.2,0.3660032033920288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,16,power_law_1.01,0.09465600252151489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,16,power_law_1.01,0.14156800508499146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,16,power_law_1.2,0.22916479110717775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,16,power_law_1.2,0.467468786239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,16,power_law_1.01,0.09635840058326721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,16,power_law_1.01,0.1450943946838379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,16,power_law_1.2,0.2684351921081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,16,power_law_1.2,0.6234816074371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,16,power_law_1.01,0.09729920029640197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,16,power_law_1.01,0.1536960005760193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,16,power_law_1.2,0.7901375770568848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,16,power_law_1.01,0.09840000271797181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,16,power_law_1.2,0.3451263904571533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,16,power_law_1.01,0.16127359867095947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,16,power_law_1.2,1.1061504364013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,16,power_law_1.01,0.10172159671783447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,16,power_law_1.01,0.16365439891815187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,16,power_law_1.2,0.4954944133758545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,16,power_law_1.2,1.3709759712219238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,16,power_law_1.01,0.10673279762268066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,16,power_law_1.01,0.18559999465942384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,16,power_law_1.2,0.6404928207397461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,16,power_law_1.01,0.10991359949111938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,16,power_law_1.01,0.19116159677505493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,16,power_law_1.2,1.8033792495727539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,16,power_law_1.2,0.8621184349060058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,16,power_law_1.01,0.11880960464477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,16,power_law_1.01,0.22001919746398926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,16,power_law_1.2,4.6517375946044925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,16,power_law_1.01,0.12176640033721924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,16,power_law_1.2,1.084390354156494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,16,power_law_1.01,0.23593599796295167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,16,power_law_1.01,0.12337280511856079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,16,balanced,1.4132800102233887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,16,power_law_1.2,1.4507007598876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,16,power_law_1.01,0.2934655904769897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,16,power_law_1.01,0.14224640130996705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,16,power_law_1.01,0.15331200361251832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,16,power_law_1.01,0.3397631883621216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,16,balanced,1.0446879863739014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,16,power_law_1.2,2.2268672943115235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,16,power_law_1.01,0.18395520448684693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,16,power_law_1.01,0.4027071952819824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,16,power_law_1.01,0.20499839782714843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,16,power_law_1.01,0.5295872211456298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,16,power_law_1.2,4.118867111206055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,16,power_law_1.01,0.2612607955932617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,16,power_law_1.01,0.6960576057434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,16,power_law_1.01,0.3115583896636963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,16,power_law_1.01,0.9418239593505859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,16,power_law_1.01,0.42362241744995116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,16,power_law_1.01,1.067244815826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,16,power_law_1.01,0.5421184062957763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,16,power_law_1.01,0.7582655906677246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,16,power_law_1.01,1.7966400146484376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,16,power_law_1.01,0.9741120338439941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,16,power_law_1.01,1.2399104118347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,16,power_law_1.01,3.4861377716064452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,16,power_law_1.01,1.8786495208740235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,16,power_law_1.01,3.7442047119140627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,16,power_law_1.2,0.06169599890708923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,16,power_law_1.2,0.09018239974975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,16,power_law_1.2,0.05470719933509827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,16,power_law_1.2,0.061260801553726194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,16,power_law_1.2,0.06354560256004334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,16,power_law_1.2,0.06670719981193543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,16,power_law_1.2,0.08324480056762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,16,power_law_1.2,0.08284159898757934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,16,power_law_1.2,0.08460800051689148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,16,power_law_1.2,0.08298879861831665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,16,power_law_1.2,0.08655999898910523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,16,power_law_1.2,0.08933759927749634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,16,power_law_1.2,0.028569599986076354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,16,power_law_1.2,0.08825600147247314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,16,power_law_1.2,0.04813440144062042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,16,power_law_1.2,0.09268479943275451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,16,balanced,2.796112060546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,16,power_law_1.2,0.029465600848197937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,16,power_law_1.2,0.09927039742469787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,16,power_law_1.2,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,16,power_law_1.2,0.10333440303802491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,16,power_law_1.2,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,16,power_law_1.2,0.1099392056465149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,16,power_law_1.2,0.039577600359916684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,16,power_law_1.2,0.12836480140686035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,16,balanced,2.0327839851379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,16,power_law_1.2,0.053439998626708986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.2,0.13649280071258546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.2,0.17702399492263793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,16,power_law_1.2,0.05006080269813538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.2,0.18483840227127074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,16,power_law_1.2,0.06298239827156067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.2,0.23583359718322755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,16,power_law_1.2,0.06286720037460328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,16,power_law_1.2,0.06606079936027527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.2,0.2915008068084717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,16,power_law_1.2,0.09475200176239014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.2,0.38760321140289306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.2,0.46945919990539553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,16,power_law_1.2,0.09601920247077941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,16,power_law_1.2,0.0955839991569519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.2,0.7282239913940429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,16,power_law_1.2,0.10844160318374634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.2,0.8820096015930176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,16,power_law_1.2,0.10385279655456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,16,power_law_1.2,1.3276736259460449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,16,power_law_1.2,0.1054144024848938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,16,power_law_1.2,1.9994688034057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,16,power_law_1.2,0.10175360441207885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.2,0.1048192024230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.2,0.117740797996521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,16,power_law_1.2,3.5557952880859376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.2,0.12917120456695558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.2,0.16922240257263182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.2,0.20813438892364503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.2,0.2773119926452637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.2,0.3513472080230713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.2,0.5061439990997314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.2,0.6828735828399658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,16,power_law_1.2,0.753772783279419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,16,power_law_1.2,1.3339455604553223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,16,power_law_1.2,2.629696083068848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,8,balanced,0.02312533309062322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,8,balanced,0.02333866556485494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,8,balanced,0.025573333104451496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,8,balanced,0.027258666853109997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,8,balanced,0.039477333426475525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,8,balanced,0.05580266813437144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,8,balanced,0.0584799995024999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,8,balanced,0.061466669042905174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,8,balanced,0.06152533491452535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,8,balanced,0.06307733555634816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,8,balanced,0.06451199948787689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,8,balanced,0.06731200218200684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,8,balanced,0.0684799998998642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,8,balanced,0.0763626645008723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,8,balanced,0.05100266635417938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,8,balanced,0.08098133405049641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,8,balanced,0.08991466959317525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,8,balanced,0.05403733253479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,8,balanced,0.053957333167394005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,8,balanced,0.0944640040397644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,8,balanced,0.05776533484458923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,8,balanced,0.1053600013256073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,8,balanced,0.07039466500282288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,8,power_law_1.01,0.0818560004234314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,8,balanced,0.09908800323804219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,8,balanced,0.18112534284591675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,8,balanced,0.09731200337409973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,8,power_law_1.01,0.10544639825820923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,8,balanced,0.0993386705716451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,8,power_law_1.01,0.07790079712867737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,8,balanced,0.09967466195424397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,8,balanced,0.22438400983810425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,8,power_law_1.01,0.0780672013759613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,8,balanced,0.10052800178527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,8,balanced,0.10247466961542766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,8,power_law_1.01,0.09726719856262207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,8,balanced,0.10310932993888855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,8,balanced,0.1869706710179647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,8,power_law_1.01,0.1058303952217102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,8,balanced,0.10280533631642659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,8,power_law_1.01,0.12654080390930175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,8,balanced,0.10587199529012044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,8,power_law_1.01,0.1224511981010437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,8,balanced,0.1130400002002716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,8,power_law_1.01,0.12685439586639405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,8,balanced,0.2193173368771871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,8,balanced,0.11322666207949321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,8,power_law_1.01,0.12926080226898193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,8,balanced,0.12146666646003723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,8,power_law_1.01,0.12986880540847778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,8,power_law_1.01,0.13147519826889037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,8,balanced,0.13085866967837015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,8,balanced,0.2548746665318807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,8,power_law_1.01,0.13516160249710082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,8,balanced,0.13949333628018698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,8,power_law_1.01,0.025216001272201537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,8,power_law_1.01,0.13388160467147828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,8,balanced,0.16145066420237222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,8,power_law_1.01,0.14231040477752685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,8,power_law_1.01,0.028889599442481994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,8,power_law_1.01,0.14101120233535766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,8,power_law_1.01,0.025887998938560485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,8,balanced,0.1751306653022766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,8,power_law_1.01,0.14740480184555055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,8,power_law_1.01,0.027871999144554137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,8,balanced,0.3171360095342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,8,power_law_1.01,0.15866880416870116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,8,power_law_1.01,0.031699201464653014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.01,0.17386239767074585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,8,balanced,0.21243733167648315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,8,power_law_1.01,0.04293760061264038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.01,0.2005311965942383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,8,power_law_1.01,0.04633600115776062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.01,0.23431038856506348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,8,power_law_1.01,0.04917759895324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.01,0.2843071937561035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,8,balanced,0.24872533480326334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,8,power_law_1.01,0.050316798686981204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,8,balanced,0.3859519958496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.01,0.3483392000198364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,8,power_law_1.01,0.05145599842071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.01,0.4648064136505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,8,power_law_1.01,0.053011202812194826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.01,0.6178688049316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,8,power_law_1.01,0.05294079780578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,8,power_law_1.01,0.05749120116233826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,8,balanced,0.34359999497731525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.01,0.870751953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,8,power_law_1.01,0.05911679863929749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.01,1.0027135848999023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,8,power_law_1.01,0.06462720036506653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,8,power_law_1.01,0.0664896011352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,8,power_law_1.01,1.2501952171325683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,8,power_law_1.01,0.06746879816055298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,8,power_law_1.01,1.9878400802612304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,8,balanced,0.5996319850285848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,8,power_law_1.01,0.07848320007324219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,8,balanced,0.42081065972646076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.01,0.14001920223236083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,8,power_law_1.01,3.9929214477539063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.01,0.16090240478515624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.01,0.13616000413894652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.01,0.15965440273284912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.01,0.18253439664840698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.01,0.25257599353790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.01,0.30352001190185546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.01,0.41460480690002444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,8,balanced,0.59170134862264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.01,0.53951997756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,8,balanced,0.7540960311889648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,8,power_law_1.01,0.6376128196716309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,8,power_law_1.01,1.0228544235229493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,8,power_law_1.01,1.9837760925292969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,8,balanced,0.7546026706695557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,8,balanced,0.9593706925710043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,8,balanced,0.9247039953867594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,8,balanced,1.4652800559997559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,8,balanced,1.4253387451171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,8,balanced,2.8981653849283853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,8,balanced,2.8014399210611978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,8,balanced,0.04604266583919525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,8,balanced,0.04641066491603851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,8,balanced,0.04574933151404063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,8,balanced,0.049813335140546165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,8,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,8,balanced,0.05373866856098175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,8,balanced,0.05009600023428599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,8,balanced,0.0747519979874293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,8,balanced,0.07112533350785573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,8,balanced,0.10697600245475769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,8,balanced,0.09944533308347066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,8,balanced,0.10643733541170756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,8,balanced,0.14865066607793173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,8,balanced,0.10904533664385478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,8,balanced,0.14904000361760458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,8,balanced,0.1090773344039917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,8,balanced,0.15310399731000265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,8,balanced,0.11050132910410564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,8,balanced,0.1520639955997467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,8,balanced,0.11343999703725179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,8,balanced,0.15267200271288553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,8,balanced,0.11384532848993938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,8,balanced,0.15305599570274353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,8,balanced,0.11550933122634888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,8,balanced,0.155130664507548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,8,balanced,0.1541920006275177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,8,balanced,0.11806399623552959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,8,balanced,0.15733866890271506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,8,power_law_1.2,0.04880000054836273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,8,balanced,0.12563199798266092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,8,balanced,0.1632960041364034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,8,balanced,0.1253439982732137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,8,balanced,0.16366933782895407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,8,power_law_1.2,0.07006080150604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,8,balanced,0.1691840092341105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,8,balanced,0.1316266655921936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,8,power_law_1.2,0.04794879853725433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,8,balanced,0.1783519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,8,balanced,0.14844266573588052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,8,power_law_1.2,0.05694079995155334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,8,power_law_1.2,0.06628479957580566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,8,balanced,0.18487467368443808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,8,balanced,0.15478932857513428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,8,power_law_1.2,0.07521920204162598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,8,balanced,0.20326934258143106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,8,balanced,0.18259199460347494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,8,power_law_1.2,0.09463040232658386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,8,balanced,0.23508799076080322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,8,power_law_1.2,0.097952002286911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,8,balanced,0.20638932784398398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,8,power_law_1.2,0.09980159997940063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,8,power_law_1.2,0.10069119930267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,8,balanced,0.2651946743329366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,8,power_law_1.01,0.05146239995956421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,8,power_law_1.2,0.10414719581604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,8,balanced,0.25998934110005695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,8,power_law_1.01,0.07429119944572449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,8,power_law_1.2,0.10614399909973145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,8,power_law_1.01,0.05312640070915222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,8,power_law_1.2,0.10995199680328369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,8,balanced,0.2988426685333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,8,power_law_1.01,0.059699201583862306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,8,power_law_1.2,0.11297279596328735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,8,power_law_1.01,0.07116159796714783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,8,balanced,0.30688534180323285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,8,power_law_1.01,0.07984640002250672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,8,power_law_1.2,0.12370560169219971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,8,power_law_1.01,0.0951744019985199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,8,power_law_1.2,0.1271615982055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,8,power_law_1.01,0.096697598695755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,8,power_law_1.2,0.1357759952545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,8,balanced,0.3858613173166911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,8,power_law_1.01,0.09811840057373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,8,power_law_1.2,0.1606528043746948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,8,power_law_1.01,0.0979968011379242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,8,balanced,0.41209598382314044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,8,power_law_1.2,0.17415679693222047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,8,power_law_1.01,0.10279680490493774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,8,power_law_1.01,0.10453120470046998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,8,power_law_1.2,0.20347518920898439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,8,power_law_1.01,0.10817279815673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,8,power_law_1.2,0.24725759029388428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,8,balanced,0.43194135030110675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,8,power_law_1.01,0.1131775975227356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,8,power_law_1.2,0.32673919200897217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,8,power_law_1.01,0.12104320526123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,8,power_law_1.01,0.07491840124130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,8,power_law_1.2,0.3705535888671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,8,balanced,0.5121546586354574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,8,power_law_1.01,0.12303999662399293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,8,power_law_1.2,0.540019178390503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,8,power_law_1.01,0.09654399752616882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,8,power_law_1.01,0.13452800512313842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,8,power_law_1.2,0.6744895935058594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,8,power_law_1.01,0.06803200244903565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,8,power_law_1.01,0.14999680519104003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,8,power_law_1.01,0.07146239876747132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,8,power_law_1.2,0.9358016014099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,8,power_law_1.01,0.1673087954521179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,8,power_law_1.01,0.0896511971950531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,8,power_law_1.01,0.19633920192718507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,8,power_law_1.2,1.2141440391540528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,8,power_law_1.01,0.10509439706802368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,8,balanced,0.6122613350550333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,8,power_law_1.01,0.22844159603118896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,8,power_law_1.01,0.12005759477615356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,8,power_law_1.2,1.7416831970214843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,8,power_law_1.01,0.2950783967971802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,8,power_law_1.01,0.1293887972831726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,8,power_law_1.2,2.352627182006836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,8,power_law_1.01,0.34985599517822263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,8,balanced,0.7204426924387614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,8,power_law_1.01,0.1363968014717102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,8,power_law_1.01,0.517907190322876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,8,power_law_1.01,0.13479679822921753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,8,power_law_1.2,4.8606208801269535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,8,power_law_1.01,0.6087168216705322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,8,power_law_1.01,0.1409600019454956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,8,power_law_1.01,0.13804800510406495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,8,power_law_1.01,0.8790143966674805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,8,power_law_1.01,0.14266239404678344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,8,power_law_1.01,1.0991488456726075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,8,power_law_1.01,0.1450111985206604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,8,power_law_1.01,0.1555199980735779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,8,power_law_1.01,1.3474495887756348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,8,balanced,0.7408160368601481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,8,power_law_1.01,0.1614400029182434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,8,power_law_1.01,0.16376320123672486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,8,power_law_1.01,2.1482751846313475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,8,power_law_1.01,0.18932479619979858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,8,power_law_1.01,0.19468799829483033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,8,balanced,0.9130187034606934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,8,power_law_1.01,0.21451520919799805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,8,power_law_1.01,4.251334381103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,8,power_law_1.01,0.2503103971481323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,8,power_law_1.01,0.3213376045227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,8,power_law_1.01,0.3610624074935913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,8,power_law_1.01,0.49338879585266116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,8,power_law_1.01,0.5435455799102783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,8,power_law_1.01,0.7986688137054443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,8,power_law_1.01,0.9832063674926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,8,balanced,0.9195306301116943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,8,power_law_1.01,1.2840512275695801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,8,power_law_1.01,1.887379264831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,8,power_law_1.01,3.437516784667969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,8,balanced,1.128111998240153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,8,power_law_1.2,0.02396800071001053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,8,power_law_1.2,0.08170239925384522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,8,power_law_1.2,0.02730880081653595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,8,power_law_1.2,0.1041152000427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,8,power_law_1.2,0.06615679860115051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,8,power_law_1.2,0.02540160119533539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,8,power_law_1.2,0.07554559707641602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,8,power_law_1.2,0.02685439884662628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,8,power_law_1.2,0.095769602060318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,8,power_law_1.2,0.03059839904308319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,8,power_law_1.2,0.10390399694442749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,8,power_law_1.2,0.03885439932346344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,8,power_law_1.2,0.12429440021514893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,8,power_law_1.2,0.04673919975757599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,8,power_law_1.2,0.12480000257492066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,8,balanced,1.3404587109883626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,8,power_law_1.2,0.047737601399421695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,8,power_law_1.2,0.12621439695358277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,8,power_law_1.2,0.04911360144615173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,8,power_law_1.2,0.12821760177612304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,8,power_law_1.2,0.05102720260620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,8,power_law_1.2,0.13224320411682128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,8,power_law_1.2,0.05257599949836731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,8,balanced,1.7158400217692058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,8,power_law_1.2,0.13645440340042114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,8,power_law_1.2,0.05185920000076294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,8,power_law_1.2,0.13148800134658814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,8,power_law_1.2,0.05413119792938233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,8,power_law_1.2,0.06265599727630615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,8,power_law_1.2,0.13861119747161865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,8,power_law_1.2,0.06572800278663635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,8,power_law_1.2,0.14469120502471924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,8,power_law_1.2,0.06949120163917541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,8,power_law_1.2,0.14769279956817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,8,power_law_1.2,0.06944640278816223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,8,power_law_1.2,0.15247999429702758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,8,power_law_1.2,0.07790079712867737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,8,power_law_1.2,0.1657088041305542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.2,0.14300800561904908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.2,0.17909120321273803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.2,0.17048959732055663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.2,0.21380479335784913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.2,0.1387392044067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.2,0.24727039337158202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.2,0.16354559659957885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.2,0.3159104108810425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.2,0.36579198837280275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.2,0.19875839948654175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.2,0.49530878067016604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.2,0.2557055950164795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.2,0.3202239990234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.2,0.6508224010467529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.2,0.46209278106689455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.2,0.9117312431335449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.2,0.578707218170166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.2,1.1542847633361817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,8,power_law_1.2,1.374073600769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,8,power_law_1.2,0.7029183864593506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,8,power_law_1.2,1.1329664230346679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,8,power_law_1.2,2.0969791412353516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,8,power_law_1.2,2.151206398010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,8,power_law_1.2,4.430815887451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,8,balanced,2.5889600118001304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,8,balanced,3.363061269124349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,8,power_law_1.2,0.07184640169143677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,8,power_law_1.2,0.09340800046920776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,8,power_law_1.2,0.06224640011787415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,8,power_law_1.2,0.07066879868507385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,8,power_law_1.2,0.08654720187187195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,8,power_law_1.2,0.09592959880828858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,8,power_law_1.2,0.11781760454177856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,8,power_law_1.2,0.1224511981010437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,8,power_law_1.2,0.12617599964141846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,8,power_law_1.2,0.13111679553985595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,8,power_law_1.2,0.1377408027648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,8,power_law_1.2,0.14384000301361083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,8,power_law_1.2,0.1436735987663269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,8,power_law_1.2,0.150271999835968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,8,power_law_1.2,0.15623680353164673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,8,power_law_1.2,0.15813119411468507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,8,power_law_1.2,0.17337599992752076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,8,power_law_1.2,0.1955839991569519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,8,power_law_1.2,0.20490880012512208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,8,power_law_1.2,0.24984960556030272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,8,power_law_1.2,0.25767040252685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,8,power_law_1.2,0.31111679077148435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,8,power_law_1.2,0.38930559158325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,8,power_law_1.2,0.5461887836456298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,8,power_law_1.2,0.6094336032867431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,8,power_law_1.2,0.8268927574157715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,8,power_law_1.2,1.1022272109985352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,8,power_law_1.2,1.3516736030578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,8,power_law_1.2,2.1943424224853514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,8,power_law_1.2,3.952384185791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,4,balanced,0.05597866574923197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,4,balanced,0.05797333518664042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,4,balanced,0.060133333007494606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,4,balanced,0.02313599983851115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,4,balanced,0.06611733138561249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,4,balanced,0.08250666658083598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,4,balanced,0.02334933231274287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,4,balanced,0.11939733227094014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,4,balanced,0.02550400048494339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,4,balanced,0.12302933136622111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,4,balanced,0.030245333909988403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,4,balanced,0.12377066413561504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,4,balanced,0.041834667325019836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,4,balanced,0.12381866574287415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,4,balanced,0.05858133236567179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,4,balanced,0.12658666570981345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,4,balanced,0.0602400004863739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,4,balanced,0.12358400225639343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,4,balanced,0.06282133360703786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,4,balanced,0.12566933035850525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,4,balanced,0.06196266909440359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,4,balanced,0.12914666533470154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,4,balanced,0.06477866570154826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,4,balanced,0.13201066851615906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,4,balanced,0.0680320014556249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,4,balanced,0.13991467157999674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,4,balanced,0.0688266654809316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,4,balanced,0.14018666744232178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,4,balanced,0.07195200026035309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,4,balanced,0.14292266964912415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,4,balanced,0.07421333094437917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,4,balanced,0.09531733393669128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,4,balanced,0.15958399573961893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,4,balanced,0.09616532921791077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,4,balanced,0.16742932796478271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,4,balanced,0.09708266456921895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,4,power_law_1.01,0.05609599947929382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,4,balanced,0.15083733201026917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,4,balanced,0.19409600893656412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,4,power_law_1.01,0.062118399143219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,4,balanced,0.15820800264676413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,4,power_law_1.01,0.05944960117340088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,4,power_law_1.01,0.06848639845848084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,4,balanced,0.21996800104777017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,4,balanced,0.18090667327245077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,4,power_law_1.01,0.0779263973236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,4,power_law_1.01,0.0915008008480072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,4,power_law_1.01,0.10296319723129273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,4,balanced,0.26817067464192706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,4,balanced,0.20571200052897134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,4,power_law_1.01,0.10576640367507935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,4,power_law_1.01,0.11041920185089112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,4,balanced,0.21357866128285727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,4,power_law_1.01,0.11119999885559081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,4,balanced,0.32758400837580365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,4,power_law_1.01,0.11847679615020752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,4,balanced,0.255130668481191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,4,power_law_1.01,0.11854720115661621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,4,power_law_1.01,0.12133760452270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,4,balanced,0.4680639902750651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,4,power_law_1.01,0.12833280563354493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,4,power_law_1.01,0.13325439691543578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,4,power_law_1.01,0.1360383987426758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,4,balanced,0.30754133065541583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,4,power_law_1.01,0.13865599632263184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,4,power_law_1.01,0.15843199491500853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.01,0.17909120321273803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.01,0.20369279384613037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,4,balanced,0.5828160047531128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.01,0.2458240032196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.01,0.29825921058654786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.01,0.36960639953613283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,4,balanced,0.36234132448832196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.01,0.5260479927062989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.01,0.6620800018310546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.01,0.9018624305725098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.01,1.2722944259643554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,4,power_law_1.01,1.485804843902588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,4,balanced,0.8393599987030029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,4,power_law_1.01,2.2422719955444337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,4,balanced,0.5571146806081136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,4,power_law_1.01,4.812384033203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,4,balanced,1.0924159685770671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,4,balanced,0.6794506708780924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,4,balanced,1.346757411956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,4,balanced,0.8859679698944092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,4,balanced,2.0539627075195312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,4,balanced,1.3357812563578289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,4,power_law_1.01,0.021958400309085847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,4,power_law_1.01,0.023871999979019166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,4,power_law_1.01,0.024582399427890776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,4,power_law_1.01,0.029894399642944335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,4,power_law_1.01,0.03467519879341126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,4,power_law_1.01,0.043033599853515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,4,power_law_1.01,0.049132800102233885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,4,power_law_1.01,0.05195519924163818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,4,power_law_1.01,0.051148802042007446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,4,power_law_1.01,0.05340800285339355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,4,power_law_1.01,0.055795198678970336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,4,power_law_1.01,0.056415998935699464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,4,power_law_1.01,0.05996800065040588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,4,power_law_1.01,0.059308797121047974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,4,power_law_1.01,0.07167360186576843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,4,power_law_1.01,0.07260159850120544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,4,power_law_1.01,0.0747648000717163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,4,power_law_1.01,0.11523840427398682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.01,0.13725440502166747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.01,0.1437440037727356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.01,0.14852479696273804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.01,0.17638399600982665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.01,0.19944319725036622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.01,0.2605247974395752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.01,0.3147455930709839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.01,0.44568958282470705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.01,0.5612095832824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,4,power_law_1.01,0.6735424041748047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,4,power_law_1.01,1.0656703948974608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,4,power_law_1.01,2.009433555603027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,4,balanced,3.981727917989095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,4,balanced,2.6294399897257485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,4,balanced,0.05009066561857859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,4,balanced,0.04971733192602793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,4,balanced,0.05671466886997223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,4,balanced,0.07399466633796692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,4,balanced,0.10381866494814555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,4,balanced,0.15613333384195963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,4,balanced,0.15636266271273294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,4,balanced,0.1588213344415029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,4,balanced,0.16160533825556436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,4,balanced,0.16048533717791238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,4,balanced,0.16224533319473267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,4,balanced,0.1633333365122477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,4,balanced,0.04562666515509287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,4,balanced,0.16668800512949625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,4,balanced,0.16641066471735635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,4,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,4,balanced,0.04987200101216634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,4,balanced,0.17652799685796103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,4,balanced,0.056143999099731445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,4,balanced,0.18051733573277792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,4,balanced,0.08090133468310039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,4,balanced,0.11148800452550252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,4,balanced,0.18197333812713623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,4,balanced,0.11386133233706157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,4,balanced,0.117658664782842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,4,balanced,0.20150933663050333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,4,balanced,0.11854933698972066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,4,balanced,0.11803733309110005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,4,balanced,0.21027199427286783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,4,balanced,0.11712533235549927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,4,balanced,0.11917866269747417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,4,balanced,0.12144533793131511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,4,balanced,0.23566399017969766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,4,balanced,0.12900267044703165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,4,balanced,0.13780267039934793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,4,balanced,0.14262400070826212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,4,balanced,0.26206399997075397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,4,balanced,0.15220800042152405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,4,balanced,0.16989866892496744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,4,balanced,0.32449066638946533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,4,balanced,0.1823413372039795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,4,balanced,0.2130826711654663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,4,balanced,0.37559465567270917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,4,balanced,0.2429386576016744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,4,balanced,0.48747201760609943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,4,balanced,0.32255999247233075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,4,balanced,0.38813332716623944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,4,balanced,0.5861013333002726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,4,balanced,0.5403840144475301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,4,balanced,0.8288479646046957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,4,balanced,0.6663839817047119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,4,balanced,0.9546613693237305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,4,balanced,1.0383893648783367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,4,balanced,1.2263360023498535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,4,power_law_1.01,0.05244799852371216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,4,power_law_1.01,0.07006719708442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,4,power_law_1.01,0.056480002403259275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,4,balanced,1.2904373009999592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,4,power_law_1.01,0.06659839749336242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,4,power_law_1.01,0.04930559992790222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,4,power_law_1.01,0.08764799833297729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,4,power_law_1.01,0.10198400020599366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,4,power_law_1.01,0.05618559718132019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,4,power_law_1.01,0.12762880325317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,4,power_law_1.01,0.05375360250473023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,4,power_law_1.01,0.13214720487594606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,4,power_law_1.01,0.05939840078353882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,4,power_law_1.01,0.13163520097732545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,4,power_law_1.01,0.06999040246009827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,4,power_law_1.01,0.13598079681396485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,4,power_law_1.01,0.08442879915237426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,4,power_law_1.01,0.13893120288848876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,4,power_law_1.01,0.09765120148658753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,4,power_law_1.01,0.143014395236969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,4,power_law_1.01,0.10206719636917114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,4,power_law_1.01,0.1455423951148987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,4,balanced,1.513962745666504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,4,power_law_1.01,0.10457600355148315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,4,power_law_1.01,0.15447039604187013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,4,power_law_1.01,0.10499199628829955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,4,power_law_1.01,0.16522239446640014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,4,power_law_1.01,0.11107200384140015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,4,power_law_1.01,0.17102080583572388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,4,power_law_1.01,0.11479040384292602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,4,power_law_1.01,0.17944960594177245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,4,power_law_1.01,0.12125439643859863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,4,power_law_1.01,0.2016832113265991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,4,power_law_1.01,0.13006080389022828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,4,power_law_1.01,0.21997439861297607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,4,power_law_1.2,0.05120000243186951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,4,power_law_1.01,0.14444799423217775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,4,power_law_1.01,0.2498944044113159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,4,power_law_1.2,0.057030397653579715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,4,power_law_1.2,0.021721599996089934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,4,power_law_1.01,0.14970240592956544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,4,power_law_1.01,0.3006848096847534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,4,power_law_1.2,0.05616000294685364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,4,power_law_1.01,0.1504639983177185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,4,power_law_1.2,0.06283519864082336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,4,power_law_1.01,0.35482239723205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,4,power_law_1.2,0.023654399812221526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,4,power_law_1.2,0.04902400076389313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,4,power_law_1.01,0.1718016028404236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,4,power_law_1.2,0.05685120224952698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,4,balanced,1.9304854075113933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,4,power_law_1.2,0.05860480070114136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,4,power_law_1.2,0.025740799307823182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,4,power_law_1.01,0.454911994934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,4,power_law_1.01,0.18855680227279664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,4,power_law_1.2,0.06792320013046264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,4,power_law_1.2,0.06865919828414917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,4,power_law_1.2,0.030048000812530517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,4,power_law_1.01,0.2364288091659546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,4,power_law_1.2,0.07456640005111695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,4,power_law_1.01,0.5689343929290771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,4,power_law_1.2,0.07798399925231933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,4,power_law_1.2,0.033606401085853575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,4,power_law_1.2,0.08278399705886841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,4,power_law_1.01,0.2760447978973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,4,power_law_1.2,0.09356799721717834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,4,power_law_1.01,0.7102528095245362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,4,power_law_1.2,0.040556800365447995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,4,power_law_1.2,0.10508160591125489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,4,power_law_1.01,0.34617600440979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,4,power_law_1.2,0.09902080297470092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,4,power_law_1.01,0.9652288436889649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,4,power_law_1.2,0.10321279764175414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,4,power_law_1.2,0.0484607994556427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,4,power_law_1.2,0.1031999945640564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,4,power_law_1.01,0.4452864170074463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,4,power_law_1.2,0.1091007947921753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,4,power_law_1.2,0.05043839812278748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,4,power_law_1.01,1.2519935607910155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,4,power_law_1.2,0.10462080240249634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,4,power_law_1.01,0.6105216026306153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,4,power_law_1.2,0.11388159990310669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,4,power_law_1.2,0.050835198163986205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,4,power_law_1.2,0.11152640581130982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,4,power_law_1.01,0.8012543678283691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,4,power_law_1.01,1.5320320129394531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,4,power_law_1.2,0.11670399904251098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,4,power_law_1.2,0.05423359870910645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,4,power_law_1.2,0.11851520538330078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,4,power_law_1.2,0.1213312029838562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,4,power_law_1.01,1.0999232292175294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,4,power_law_1.2,0.12206720113754273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,4,power_law_1.2,0.05691519975662231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,4,power_law_1.01,2.2947391510009765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,4,power_law_1.2,0.1238592028617859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,4,balanced,2.3398292859395347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,4,power_law_1.2,0.13335039615631103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,4,power_law_1.01,1.4073472023010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,4,power_law_1.2,0.057548797130584715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,4,power_law_1.2,0.12842240333557128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,4,power_law_1.2,0.14357119798660278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,4,power_law_1.2,0.05865600109100342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,4,power_law_1.01,1.7566463470458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,4,power_law_1.2,0.13653119802474975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,4,power_law_1.01,4.602182388305664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,4,power_law_1.2,0.15189119577407836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,4,power_law_1.2,0.061894398927688596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,4,power_law_1.2,0.13797760009765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,4,power_law_1.2,0.15429120063781737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,4,power_law_1.2,0.07070080041885377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,4,power_law_1.01,2.8674943923950194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,4,power_law_1.2,0.14611200094223023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,4,power_law_1.2,0.17345919609069824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,4,power_law_1.2,0.07204480171203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,4,power_law_1.2,0.16440320014953613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,4,power_law_1.2,0.19685759544372558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,4,power_law_1.2,0.0804095983505249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.2,0.18249599933624266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,4,power_law_1.01,5.197785568237305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,4,power_law_1.2,0.12117760181427002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,4,power_law_1.2,0.23932158946990967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.2,0.21913599967956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.2,0.1492416024208069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,4,power_law_1.2,0.29475200176239014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.2,0.24293758869171142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.2,0.3240511894226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,4,power_law_1.2,0.3756799936294556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.2,0.15375360250473022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.2,0.38973441123962405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,4,power_law_1.2,0.445849609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.2,0.15520000457763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.2,0.5465343952178955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,4,power_law_1.2,0.6232384204864502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.2,0.18102400302886962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.2,0.6932672023773193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,4,power_law_1.2,0.788159990310669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.2,0.20381441116333007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,4,power_law_1.2,1.1470720291137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.2,1.0148223876953124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.2,0.2649280071258545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,4,power_law_1.2,1.5233344078063964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.2,1.310431957244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.2,0.3392575979232788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,4,power_law_1.2,1.511673641204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,4,power_law_1.2,1.8558528900146485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.2,0.45442562103271483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.2,0.5841983795166016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,4,power_law_1.2,2.8418432235717774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,4,power_law_1.2,2.810419273376465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,4,power_law_1.2,0.7086592197418213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,4,power_law_1.2,4.950239944458008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,4,power_law_1.2,5.4495361328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,4,power_law_1.2,1.1175359725952148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,4,power_law_1.2,2.2865856170654295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,4,balanced,3.7549705505371094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,4,balanced,4.5295359293619795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,4,power_law_1.2,0.05729920268058777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,4,power_law_1.2,0.06938239932060242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,4,power_law_1.2,0.05984640121459961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,4,power_law_1.2,0.06721919775009155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,4,power_law_1.2,0.07879040241241456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,4,power_law_1.2,0.09718400239944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,4,power_law_1.2,0.12535680532455445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,4,power_law_1.2,0.12881920337677003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,4,power_law_1.2,0.13299839496612548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,4,power_law_1.2,0.13540480136871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,4,power_law_1.2,0.14415359497070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,4,power_law_1.2,0.1499519944190979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,4,power_law_1.2,0.1550655961036682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,4,power_law_1.2,0.15834239721298218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,4,power_law_1.2,0.1641983985900879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,4,power_law_1.2,0.17340799570083618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,4,power_law_1.2,0.18755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,4,power_law_1.2,0.21795198917388917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,4,power_law_1.2,0.2280128002166748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,4,power_law_1.2,0.26474239826202395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,4,power_law_1.2,0.30056960582733155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,4,power_law_1.2,0.38968958854675295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,4,power_law_1.2,0.46289920806884766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,4,power_law_1.2,0.6571008205413819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,4,power_law_1.2,0.7141568183898925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,4,power_law_1.2,1.0178560256958007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,4,power_law_1.2,1.3662655830383301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,4,power_law_1.2,1.8434944152832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,4,power_law_1.2,2.4621824264526366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,4,power_law_1.2,5.000844955444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,2,balanced,0.02327999969323476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,2,balanced,0.023215999205907185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,2,balanced,0.025221332907676697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,2,balanced,0.0312266672650973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,2,balanced,0.04605866471926371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,2,balanced,0.06634666522343953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,2,balanced,0.06992533306280772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,2,balanced,0.06899199883143108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,2,balanced,0.07011733452479045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,2,balanced,0.07191999753316243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,2,balanced,0.07830399771531422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,2,balanced,0.08030400176843007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,2,balanced,0.08272533118724823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,2,balanced,0.08571199576059978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,2,balanced,0.08922132849693298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,2,balanced,0.09259733557701111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,2,balanced,0.09870933492978413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,2,balanced,0.11915733416875203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,2,balanced,0.133050670226415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,2,balanced,0.15682133038838705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,2,balanced,0.18607467412948608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,2,balanced,0.26845333973566693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,2,balanced,0.3012426694234212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,2,balanced,0.06611200173695882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,2,balanced,0.3440693219502767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,2,balanced,0.06860266625881195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,2,balanced,0.07457600037256877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,2,balanced,0.08679466446240743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,2,balanced,0.11252267162005107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,2,balanced,0.17283199230829874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,2,balanced,0.17289066314697266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,2,balanced,0.1702293356259664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,2,balanced,0.39132265249888104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,2,balanced,0.17215466499328613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,2,balanced,0.1744640072186788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,2,balanced,0.1746079921722412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,2,balanced,0.1792746583620707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,2,balanced,0.18100800116856894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,2,balanced,0.18371733029683432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,2,balanced,0.19171732664108276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,2,balanced,0.1956640084584554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,2,balanced,0.20334400733311972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,2,power_law_1.01,0.06589440107345582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,2,balanced,0.22642666101455688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,2,balanced,0.623082677523295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,2,power_law_1.01,0.07139840126037597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,2,balanced,0.24580266078313193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,2,power_law_1.01,0.07107200026512146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,2,power_law_1.01,0.08313599824905396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,2,balanced,0.28628265857696533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,2,power_law_1.01,0.10002559423446655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,2,power_law_1.01,0.11367039680480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,2,power_law_1.01,0.1384384036064148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,2,balanced,0.32832000652949017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,2,power_law_1.01,0.14613120555877684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,2,power_law_1.01,0.15240319967269897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,2,power_law_1.01,0.15134079456329347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,2,balanced,0.4222613175710042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,2,power_law_1.01,0.1592319965362549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,2,power_law_1.01,0.1621567964553833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,2,power_law_1.01,0.16672639846801757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,2,power_law_1.01,0.17288320064544677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,2,balanced,0.5144213438034058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,2,balanced,0.7326239744822184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,2,power_law_1.01,0.178220796585083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,2,power_law_1.01,0.18623360395431518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,2,power_law_1.01,0.19626879692077637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,2,power_law_1.01,0.22295680046081542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.01,0.2457727909088135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,2,balanced,0.7532587051391602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.01,0.3113152027130127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.01,0.3561984062194824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.01,0.470630407333374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.01,0.5838719844818115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.01,0.7952256202697754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.01,1.042796802520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,2,balanced,0.9581600030263265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.01,1.4436223983764649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.01,1.9045375823974608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,2,balanced,0.9678506851196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,2,power_law_1.01,2.3919231414794924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,2,power_law_1.01,3.744371032714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,2,power_law_1.01,6.944966125488281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,2,balanced,1.3923145929972331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,2,balanced,1.816186745961507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,2,balanced,1.4550399780273438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,2,balanced,2.248213291168213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,2,power_law_1.01,0.02078080028295517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,2,power_law_1.01,0.022368000447750093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,2,power_law_1.01,0.024371199309825897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,2,power_law_1.01,0.029203200340270997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,2,power_law_1.01,0.04022400081157684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,2,power_law_1.01,0.04696959853172302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,2,power_law_1.01,0.05658239722251892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,2,power_law_1.01,0.059571200609207155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,2,power_law_1.01,0.061324799060821535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,2,power_law_1.01,0.06303359866142273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,2,power_law_1.01,0.0652351975440979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,2,power_law_1.01,0.06832000017166137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,2,power_law_1.01,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,2,power_law_1.01,0.07153919935226441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,2,balanced,3.4263413747151694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,2,power_law_1.01,0.07511039972305297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,2,power_law_1.01,0.07834240198135375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,2,power_law_1.01,0.08931840062141419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,2,power_law_1.01,0.10355839729309083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.01,0.1164736032485962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.01,0.14335999488830567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.01,0.16339839696884156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.01,0.21074559688568115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.01,0.24897921085357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.01,0.30796799659729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.01,0.37100160121917725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.01,0.5149824142456054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,2,balanced,2.8822720845540366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.01,0.6333888053894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,2,power_law_1.01,0.7723711967468262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,2,power_law_1.01,1.243507194519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,2,power_law_1.01,2.3105087280273438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,2,balanced,6.7673600514729815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,2,balanced,0.051818668842315674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,2,balanced,0.053930665055910744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,2,balanced,0.0602453351020813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,2,balanced,0.08137066662311554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,2,balanced,0.1120746632417043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,2,balanced,0.17571200927098593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,2,balanced,0.17614400386810303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,2,balanced,0.17957866191864014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,2,balanced,0.18376533190409342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,2,balanced,0.1883253256479899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,2,balanced,0.18171199162801108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,2,balanced,0.18782933553059897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,2,balanced,0.18925867478052774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,2,balanced,0.19510932763417563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,2,balanced,0.202074666817983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,2,balanced,0.04763199885686239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,2,balanced,0.20721065998077393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,2,balanced,0.0498986691236496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,2,balanced,0.2150986591974894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,2,balanced,0.054272000988324486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,2,balanced,0.06653866668542226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,2,balanced,0.23748799165089926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,2,balanced,0.09499200185139973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,2,balanced,0.13798399766286215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,2,balanced,0.26224533716837567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,2,balanced,0.1420799990495046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,2,balanced,0.13796266913414001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,2,balanced,0.2983253399531047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,2,balanced,0.14206399520238241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,2,balanced,0.14220800002415976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,2,balanced,0.144186665614446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,2,balanced,0.3336373170216878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,2,balanced,0.14827733238538107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,2,balanced,0.15031466881434122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,2,balanced,0.1551359991232554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,2,balanced,0.44998931884765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,2,balanced,0.16430933276812235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,2,balanced,0.16833599408467612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,2,balanced,0.17919999361038208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,2,balanced,0.5073813199996948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,2,balanced,0.2071146567662557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,2,power_law_1.01,0.05191680192947388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,2,balanced,0.2250773310661316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,2,power_law_1.01,0.05601919889450073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,2,power_law_1.01,0.061875200271606444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,2,balanced,0.2741706569989522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,2,power_law_1.01,0.07269759774208069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,2,balanced,0.7327893575032552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,2,power_law_1.01,0.09366400241851806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,2,balanced,0.31462399164835614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,2,power_law_1.01,0.10974719524383544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,2,power_law_1.01,0.14048000574111938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,2,power_law_1.01,0.1449087977409363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,2,balanced,0.439733346303304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,2,power_law_1.01,0.1499392032623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,2,power_law_1.2,0.04602240025997162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,2,power_law_1.2,0.08115839958190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,2,power_law_1.01,0.15593600273132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,2,power_law_1.01,0.16392960548400878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,2,power_law_1.2,0.05043839812278748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,2,balanced,0.8811039924621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,2,power_law_1.2,0.09248639941215515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,2,power_law_1.01,0.1726591944694519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,2,balanced,0.5185546477635702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,2,power_law_1.2,0.051999998092651364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,2,power_law_1.2,0.08464000225067139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,2,power_law_1.01,0.17636480331420898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,2,power_law_1.01,0.1801599979400635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,2,power_law_1.2,0.06382079720497132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,2,power_law_1.2,0.10529919862747192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,2,power_law_1.01,0.18971519470214843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,2,power_law_1.01,0.1972480058670044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,2,power_law_1.2,0.07564160227775574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,2,power_law_1.2,0.1289728045463562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,2,power_law_1.01,0.20842878818511962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,2,power_law_1.2,0.15601279735565185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,2,power_law_1.2,0.08833919763565064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,2,balanced,0.7561866442362467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,2,power_law_1.01,0.23513600826263428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,2,power_law_1.2,0.19441920518875122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,2,power_law_1.2,0.11712640523910522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,2,power_law_1.01,0.2700799942016602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,2,power_law_1.2,0.19955840110778808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,2,power_law_1.01,0.3256256103515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,2,power_law_1.2,0.11940480470657348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,2,power_law_1.01,0.36867198944091795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,2,power_law_1.2,0.21015679836273193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,2,power_law_1.2,0.12357759475708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,2,power_law_1.01,0.04848000109195709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,2,balanced,1.2702346642812092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,2,power_law_1.01,0.4565311908721924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,2,power_law_1.2,0.2119296073913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,2,power_law_1.01,0.5652927875518798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,2,power_law_1.2,0.21616001129150392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,2,power_law_1.2,0.12922879457473754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,2,power_law_1.01,0.05253120064735413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,2,power_law_1.01,0.7730559825897216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,2,power_law_1.2,0.22306559085845948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,2,balanced,0.935312032699585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,2,power_law_1.01,0.053395199775695804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,2,power_law_1.2,0.1315392017364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,2,power_law_1.01,0.9614463806152344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,2,power_law_1.2,0.2321727991104126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,2,power_law_1.01,0.06321280002593994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,2,power_law_1.01,1.3166463851928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,2,power_law_1.2,0.14627200365066528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,2,power_law_1.2,0.23369600772857665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,2,power_law_1.01,0.08147839903831482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,2,power_law_1.01,1.7162559509277344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,2,power_law_1.2,0.2476799964904785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,2,power_law_1.2,0.020895999670028687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,2,power_law_1.01,0.09599999785423279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,2,power_law_1.2,0.14854400157928466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,2,power_law_1.01,2.0914495468139647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,2,power_law_1.2,0.2535423994064331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,2,power_law_1.01,0.1182528018951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,2,power_law_1.2,0.16318080425262452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,2,power_law_1.01,3.1880640029907226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,2,power_law_1.2,0.02227199971675873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,2,power_law_1.01,0.11978240013122558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,2,power_law_1.2,0.2698303937911987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,2,power_law_1.2,0.174726402759552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,2,power_law_1.01,0.12303999662399293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,2,power_law_1.2,0.024864000082015992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,2,power_law_1.2,0.297216010093689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,2,power_law_1.01,5.943251037597657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,2,power_law_1.2,0.1828223943710327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,2,power_law_1.01,0.12549760341644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,2,balanced,1.6098826726277669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.2,0.32436480522155764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,2,balanced,1.3675467173258464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,2,power_law_1.2,0.02983039915561676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,2,power_law_1.2,0.18212480545043946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,2,power_law_1.01,0.13569920063018798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.2,0.38127360343933103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,2,power_law_1.2,0.039724799990653994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,2,power_law_1.2,0.21556479930877687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,2,power_law_1.01,0.13912960290908813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.2,0.4442431926727295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,2,power_law_1.01,0.1447487950325012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,2,power_law_1.2,0.04415999948978424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,2,power_law_1.2,0.24335360527038574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.2,0.5794623851776123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,2,power_law_1.01,0.16016639471054078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,2,power_law_1.2,0.3045120000839233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,2,power_law_1.2,0.05637120008468628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.2,0.6856639862060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,2,power_law_1.2,0.3640448093414307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,2,power_law_1.01,0.17478400468826294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.2,0.8855615615844726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,2,power_law_1.2,0.05976319909095764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,2,power_law_1.2,0.47530241012573243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.2,1.1758912086486817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,2,power_law_1.01,0.18081920146942138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,2,power_law_1.2,0.0611519992351532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,2,power_law_1.01,0.17686400413513184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,2,power_law_1.2,0.5808640003204346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.2,1.653433609008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,2,power_law_1.01,0.2040895938873291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,2,balanced,1.7788000106811523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,2,power_law_1.2,0.06214399933815003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,2,power_law_1.2,0.7937791824340821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.2,2.2048831939697267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,2,power_law_1.01,0.23915519714355468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,2,power_law_1.2,0.9874688148498535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,2,power_law_1.2,0.06539520025253295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,2,balanced,1.9666026433308919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,2,power_law_1.2,2.562905693054199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,2,power_law_1.01,0.29112958908081055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,2,power_law_1.2,0.0690559983253479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,2,power_law_1.2,1.4841728210449219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,2,power_law_1.01,0.3602688074111938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,2,power_law_1.2,0.07153279781341552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,2,power_law_1.2,4.1009471893310545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,2,power_law_1.01,0.4819968223571777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,2,power_law_1.2,1.9328256607055665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,2,power_law_1.2,0.07334399819374085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,2,power_law_1.01,0.5760575771331787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,2,power_law_1.2,0.07772799730300903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,2,power_law_1.2,2.4650623321533205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,2,power_law_1.2,0.08282880187034607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,2,power_law_1.2,8.486450958251954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,2,power_law_1.01,0.8230079650878906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,2,power_law_1.2,0.08985599875450134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,2,power_law_1.01,1.016864013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,2,power_law_1.2,3.715270233154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,2,power_law_1.2,0.10814720392227173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,2,power_law_1.01,1.4556415557861329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.2,0.11740800142288207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.2,0.1465407967567444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,2,power_law_1.2,7.379840087890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,2,power_law_1.01,1.8939712524414063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,2,balanced,2.184650739034017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.2,0.1709439992904663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,2,power_law_1.01,2.3081279754638673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.2,0.21475839614868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.2,0.25528318881988527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,2,power_law_1.01,3.770207977294922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.2,0.3251904010772705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.2,0.3825599908828735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,2,power_law_1.01,7.155033874511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.2,0.527462387084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,2,balanced,3.016010602315267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.2,0.669817590713501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,2,power_law_1.2,0.7841023921966552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,2,power_law_1.2,1.2545920372009278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,2,power_law_1.2,2.3607551574707033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,2,balanced,3.3770666122436523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,2,balanced,5.814629236857097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,2,balanced,6.638704299926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,2,power_law_1.2,0.05143679976463318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,2,power_law_1.2,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,2,power_law_1.2,0.05597440004348755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,2,power_law_1.2,0.06692479848861695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,2,power_law_1.2,0.08647680282592773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,2,power_law_1.2,0.1023360013961792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,2,power_law_1.2,0.1397503972053528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,2,power_law_1.2,0.14503040313720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,2,power_law_1.2,0.1509119987487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,2,power_law_1.2,0.15134079456329347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,2,power_law_1.2,0.163155198097229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,2,power_law_1.2,0.16803840398788453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,2,power_law_1.2,0.1787328004837036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,2,power_law_1.2,0.19221760034561158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,2,power_law_1.2,0.20227200984954835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,2,power_law_1.2,0.21330559253692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,2,power_law_1.2,0.22169599533081055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,2,power_law_1.2,0.2410111904144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,2,power_law_1.2,0.27778561115264894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,2,power_law_1.2,0.33870079517364504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,2,power_law_1.2,0.37114880084991453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,2,power_law_1.2,0.46856322288513186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,2,power_law_1.2,0.5899519920349121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,2,power_law_1.2,0.8056703567504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,2,power_law_1.2,1.015123176574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,2,power_law_1.2,1.391859245300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,2,power_law_1.2,1.79052791595459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,2,power_law_1.2,2.284614372253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,2,power_law_1.2,3.3827518463134765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,2,power_law_1.2,6.89117431640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,32,balanced,0.027493332823117573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,32,balanced,0.025072000920772552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,32,balanced,0.025589334468046825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,32,balanced,0.02756800005833308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,32,balanced,0.027248000105222065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,32,balanced,0.029578665892283123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,32,balanced,0.029578665892283123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,32,balanced,0.029343999922275543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,32,balanced,0.029178666571776073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,32,balanced,0.02956266701221466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,32,balanced,0.031717332700888314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,32,balanced,0.030213333666324615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,32,balanced,0.02956266701221466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,32,balanced,0.03390933324893316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,32,balanced,0.03711466739575068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,32,balanced,0.037861332297325134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,32,balanced,0.039936001102129616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,32,balanced,0.037952000896135964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,32,balanced,0.041290665666262306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,32,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,32,balanced,0.04412800073623657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,32,balanced,0.04799466828505198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,32,balanced,0.0499946673711141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,32,balanced,0.06305600206057231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,32,balanced,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,32,balanced,0.0906773308912913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,32,balanced,0.11195733149846394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,32,balanced,0.132533331712087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,32,balanced,0.168778657913208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,32,balanced,0.03183999905983607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,32,balanced,0.031248000760873158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,32,balanced,0.029450667401154835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,32,balanced,0.03151999910672506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,32,balanced,0.03197866678237915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,32,balanced,0.03365333378314972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,32,balanced,0.03379199902216593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,32,balanced,0.03369066615899404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,32,balanced,0.2870560089747111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,32,balanced,0.03369066615899404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,32,balanced,0.035605333745479584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,32,balanced,0.033701332906881966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,32,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,32,balanced,0.03487999985615412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,32,balanced,0.037717332442601524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,32,balanced,0.03999999910593033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,32,balanced,0.03968533376852671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,32,balanced,0.03982933362325033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,32,balanced,0.04166933397452036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,32,balanced,0.043866669138272606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,32,balanced,0.045834665497144066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,32,balanced,0.050053333242734276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,32,balanced,0.055402666330337524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,32,balanced,0.06439466774463654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,32,balanced,0.08914132912953694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,32,balanced,0.11545067032178243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,1,balanced,0.05791999896367391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,1,balanced,0.060218666990598045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,1,balanced,0.07016533116499583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,1,balanced,0.09703999757766724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,32,balanced,0.15335466464360556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,1,balanced,0.14285332957903543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,1,balanced,0.2212053338686625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,1,balanced,0.22265599171320596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,1,balanced,0.22458666563034058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,1,balanced,0.2271733283996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,1,balanced,0.23002133766810098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,1,balanced,0.2307093342145284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,1,balanced,0.23541333278020224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,1,balanced,0.23996265729268393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,1,balanced,0.24361066023508707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,1,balanced,0.2503733237584432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,1,balanced,0.2562133272488912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,1,balanced,0.2743626634279887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,32,balanced,0.18276800711949667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,1,balanced,0.3139999906222026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,1,balanced,0.3346879879633586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,1,balanced,0.4068426688512166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,1,balanced,0.48042134443918866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,1,balanced,0.670746644337972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,1,balanced,0.7794293562571207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,32,balanced,0.22295467058817545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,1,balanced,1.1571520169576008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,1,power_law_1.01,0.05172479748725891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,1,power_law_1.01,0.05498239994049072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,1,power_law_1.01,0.06468480229377746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,1,balanced,1.4434666633605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,1,power_law_1.01,0.08376960158348083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,1,power_law_1.01,0.11233919858932495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,1,power_law_1.01,0.13377280235290528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,1,power_law_1.01,0.16931840181350707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,1,power_law_1.01,0.1785599946975708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,1,power_law_1.01,0.18334720134735108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,1,power_law_1.01,0.1925696015357971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,1,power_law_1.01,0.1983296036720276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,1,power_law_1.01,0.20364160537719728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,1,balanced,2.0918025970458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,1,power_law_1.01,0.21338241100311278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,1,power_law_1.01,0.23230080604553222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,1,power_law_1.01,0.24369280338287352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,1,power_law_1.01,0.2623487949371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,1,power_law_1.2,0.05185279846191406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,1,power_law_1.01,0.25626239776611326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,1,balanced,0.05599466462930044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,32,balanced,0.3154933253924052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,1,power_law_1.01,0.30919039249420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,1,power_law_1.2,0.05660799741744995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,1,power_law_1.01,0.3730560064315796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,1,balanced,0.05993066728115082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,1,power_law_1.2,0.0655232012271881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,1,balanced,0.06413866579532623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,1,power_law_1.01,0.4649983882904053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,1,power_law_1.2,0.07925760149955749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,1,balanced,0.0865119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,1,power_law_1.01,0.5138815879821778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,1,power_law_1.2,0.09824640154838563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,1,balanced,0.12788800398508707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,1,balanced,2.6932319005330405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,1,power_law_1.2,0.1267135977745056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,1,balanced,0.18728532393773398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,1,power_law_1.01,0.6716544151306152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,1,power_law_1.2,0.16614400148391723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,1,balanced,0.19050133228302002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,1,balanced,0.19220266739527384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,1,power_law_1.2,0.17832959890365602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,1,power_law_1.01,0.8385087966918945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,1,balanced,0.19552532831827799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,1,power_law_1.2,0.18624639511108398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,1,balanced,0.19819732507069907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,1,power_law_1.01,1.1509440422058106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,1,balanced,0.20127467314402261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,1,power_law_1.2,0.19228800535202026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,1,balanced,0.20544532934824625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,1,power_law_1.2,0.20562560558319093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,1,power_law_1.01,1.4887295722961427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,1,balanced,0.2132213314374288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,1,power_law_1.2,0.2119488000869751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,1,balanced,0.21753066778182983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,1,power_law_1.01,2.111008071899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,1,power_law_1.2,0.22340478897094726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,1,balanced,0.22787733872731528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,1,power_law_1.01,0.0497408002614975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,1,power_law_1.2,0.23196160793304443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,1,balanced,0.23733866214752197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,1,power_law_1.01,2.7600576400756838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,1,power_law_1.2,0.23571200370788575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,1,balanced,3.3462985356648765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,1,balanced,0.25507734219233197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,1,power_law_1.01,0.05299839973449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,1,power_law_1.01,3.370265579223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,1,power_law_1.2,0.24759678840637206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,1,balanced,0.29710932572682697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,1,power_law_1.01,0.06119679808616638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,1,power_law_1.2,0.26730880737304685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,1,power_law_1.01,5.273369598388672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,1,power_law_1.01,0.0773311972618103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,1,balanced,0.33269333839416504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,1,power_law_1.2,0.3223488092422485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,1,power_law_1.01,0.10092799663543701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,1,power_law_1.2,0.3615871906280518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,1,power_law_1.01,10.222643280029297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,1,balanced,0.40962668259938556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,1,power_law_1.01,0.11930880546569825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,1,power_law_1.2,0.45349760055541993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,1,power_law_1.01,0.14815360307693481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,1,power_law_1.2,0.5174208164215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,1,balanced,0.4700426657994588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,1,power_law_1.01,0.1553984045982361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,1,power_law_1.2,0.6779327869415284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,1,power_law_1.01,0.16202880144119264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,1,balanced,0.7002453009287516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,1,power_law_1.2,0.8590911865234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,1,power_law_1.01,0.16424319744110108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,1,power_law_1.01,0.17749119997024537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,1,power_law_1.2,1.1480768203735352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,1,balanced,0.8072853088378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,1,power_law_1.01,0.1880511999130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,1,power_law_1.01,0.19941120147705077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,1,power_law_1.2,1.462713623046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,1,power_law_1.01,0.2160192012786865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,1,balanced,5.220058759053548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,1,power_law_1.2,2.0726015090942385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,1,power_law_1.01,0.23637120723724364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,1,balanced,1.2062079906463623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,32,balanced,0.5734133323033651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,1,power_law_1.2,2.7593984603881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,1,power_law_1.01,0.2519295930862427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,1,power_law_1.01,0.2448256015777588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,1,power_law_1.2,0.04984320104122162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,1,power_law_1.2,3.377158355712891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,1,power_law_1.01,0.3002943992614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,1,power_law_1.2,0.054150402545928955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,1,power_law_1.01,0.337446403503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,1,power_law_1.2,5.211103820800782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,1,power_law_1.2,0.06227200031280518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,1,power_law_1.01,0.43736958503723145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,1,balanced,1.5135199228922527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,1,power_law_1.2,0.07596799731254578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,1,power_law_1.01,0.5190080165863037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,1,power_law_1.2,10.052409362792968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,1,power_law_1.2,0.09296640157699584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,1,power_law_1.01,0.6972352027893066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,1,power_law_1.2,0.11232000589370728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,1,power_law_1.2,0.1476863980293274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,1,power_law_1.01,0.8751999855041503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,1,power_law_1.2,0.15326720476150513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,1,power_law_1.01,1.2142911911010743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,1,power_law_1.2,0.16354559659957885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,1,power_law_1.01,1.5503423690795899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,1,balanced,2.214570681254069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,1,power_law_1.2,0.16691839694976807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,1,power_law_1.01,2.2312576293945314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,1,power_law_1.2,0.17908480167388915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,32,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,1,power_law_1.2,0.1894144058227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,1,power_law_1.01,2.913408088684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,1,power_law_1.2,0.2038975954055786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,32,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,1,power_law_1.01,3.5835391998291017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,1,power_law_1.2,0.22266879081726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,32,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,32,balanced,0.035631999373435974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,1,power_law_1.2,0.24102399349212647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,1,power_law_1.01,5.62182388305664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,32,balanced,0.03530666728814443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,1,power_law_1.2,0.25736958980560304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,32,balanced,0.03467733412981033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,1,power_law_1.2,0.2518079996109009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,32,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,1,balanced,2.900341351826986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,32,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,1,power_law_1.2,0.30842878818511965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,1,power_law_1.01,10.962675476074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,32,balanced,0.037392000357309975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,1,power_law_1.2,0.3518847942352295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,32,balanced,0.035786665976047516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,1,power_law_1.2,0.4422912120819092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,32,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,1,power_law_1.2,0.5301760196685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,32,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,1,power_law_1.2,0.7013696193695068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,32,balanced,0.03782399992148081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,1,power_law_1.2,0.8901375770568848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,32,balanced,0.03793599953254064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,1,power_law_1.2,1.2199487686157227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,32,balanced,0.041749333341916404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,1,power_law_1.2,1.5554559707641602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,32,balanced,0.04192533095677694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,1,power_law_1.2,2.225574493408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,32,balanced,0.04385066529115041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,1,balanced,10.2346560160319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,32,balanced,0.04426133135954539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,1,power_law_1.2,2.9146495819091798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,1,balanced,3.5814825693766275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,32,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,1,power_law_1.2,3.5867584228515623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,32,balanced,0.04684266448020935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,1,power_law_1.2,5.621785736083984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,32,balanced,0.05031466484069824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,1,power_law_1.2,10.970246124267579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,32,balanced,0.05421333511670431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,32,balanced,0.058304001887639366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,32,balanced,0.07660266757011414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,1,balanced,5.607327779134114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,32,balanced,0.08711466193199158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,32,balanced,0.10979732871055603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,32,balanced,0.1422719955444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,32,balanced,0.16579733292261759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,1,balanced,10.98520024617513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,32,balanced,0.22763200600941977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,32,balanced,0.3979146480560303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,32,balanced,0.019343999524911244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,32,balanced,0.01912533367673556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,32,balanced,0.019167999426523846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,32,balanced,0.019109333554903667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,32,balanced,0.01941866676012675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,32,balanced,0.023082666099071503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,32,balanced,0.03346133232116699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,32,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,32,balanced,0.03162133445342382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,32,balanced,0.03161066770553589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,32,balanced,0.03156266609827677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,32,balanced,0.031685332457224526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,32,power_law_1.01,0.02512640058994293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,32,balanced,0.029711998999118805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,32,power_law_1.01,0.02359039932489395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,32,balanced,0.029509333272775013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,32,power_law_1.01,0.022892799973487855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,32,balanced,0.03550933301448822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,32,power_law_1.01,0.026092800498008727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,32,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,32,power_law_1.01,0.02632319927215576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,32,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,32,power_law_1.01,0.027296000719070436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,32,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,32,power_law_1.01,0.027852800488471986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,32,power_law_1.01,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,32,balanced,0.044122666120529175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,32,power_law_1.01,0.028947201371192933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,32,power_law_1.01,0.030297601222991945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,32,power_law_1.01,0.030796799063682555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,32,balanced,0.05109333495299021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,32,power_law_1.01,0.027993598580360414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,32,power_law_1.01,0.032262399792671204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,32,power_law_1.01,0.030630400776863097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,32,balanced,0.060032000144322716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,32,power_law_1.01,0.03446399867534637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,32,power_law_1.01,0.030707201361656188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,32,power_law_1.01,0.033542400598526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,32,power_law_1.01,0.03258880078792572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,32,balanced,0.07633600135644276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,32,power_law_1.01,0.034841600060462954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,32,power_law_1.01,0.0328000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,32,power_law_1.01,0.03661440014839172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,32,power_law_1.01,0.03351039886474609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,32,power_law_1.01,0.040038400888442995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,32,balanced,0.09658666451772054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,32,power_law_1.01,0.03381119966506958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,32,power_law_1.01,0.04085119962692261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,32,power_law_1.01,0.034246399998664856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,32,power_law_1.01,0.03514240086078644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,32,power_law_1.01,0.04257920086383819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,32,power_law_1.01,0.034483200311660765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,32,power_law_1.01,0.044588801264762876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,32,power_law_1.01,0.033478400111198424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,32,power_law_1.01,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,32,balanced,0.12661866346995035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,32,power_law_1.01,0.04912000000476837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,32,power_law_1.01,0.03175039887428284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,32,power_law_1.01,0.03726719915866852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,32,power_law_1.01,0.05403519868850708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,32,power_law_1.01,0.03380480110645294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,32,power_law_1.01,0.03733119964599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,32,power_law_1.01,0.06368640065193176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,32,power_law_1.01,0.017452800273895265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,32,power_law_1.01,0.03395200073719025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,32,power_law_1.01,0.0409855991601944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,32,power_law_1.01,0.0715071976184845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,32,balanced,0.16109866897265115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,32,power_law_1.01,0.017375999689102174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,32,power_law_1.01,0.0417279988527298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,32,power_law_1.01,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,32,power_law_1.01,0.08450559973716736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,32,power_law_1.01,0.01719679981470108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,32,power_law_1.01,0.04323840141296387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,32,power_law_1.01,0.03596160113811493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,32,power_law_1.01,0.11061760187149047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,32,power_law_1.01,0.018118399381637573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,32,power_law_1.2,0.02598400115966797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,32,power_law_1.01,0.04778240025043488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,32,power_law_1.01,0.035622400045394895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,32,power_law_1.01,0.13124480247497558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,32,power_law_1.01,0.019392000138759614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,32,power_law_1.2,0.024537600576877594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,32,power_law_1.01,0.035897600650787356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,32,power_law_1.01,0.05187199711799621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,32,balanced,0.2262186606725057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,32,power_law_1.01,0.17216639518737792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,32,power_law_1.01,0.02192640006542206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,32,power_law_1.2,0.02396160066127777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,32,power_law_1.01,0.03597440123558045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,32,power_law_1.01,0.06030719876289368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,32,power_law_1.01,0.2346816062927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,32,power_law_1.01,0.03304960131645203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,32,power_law_1.2,0.026438400149345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,32,power_law_1.01,0.036499199271202085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,32,power_law_1.01,0.06311039924621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,32,power_law_1.01,0.27365760803222655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,32,power_law_1.01,0.03219839930534363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,32,power_law_1.2,0.0271807998418808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,32,power_law_1.01,0.037299200892448425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,32,power_law_1.01,0.08309119939804077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,32,power_law_1.01,0.4203904151916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,32,power_law_1.01,0.031219199299812317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,32,power_law_1.2,0.028339201211929323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,32,power_law_1.01,0.10002559423446655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,32,power_law_1.01,0.03813120126724243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,32,balanced,0.2935573259989421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,32,power_law_1.01,0.03152639865875244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,32,power_law_1.01,0.6964352130889893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,32,power_law_1.2,0.02812800109386444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,32,power_law_1.01,0.1440832018852234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,32,power_law_1.01,0.03863680064678192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,32,power_law_1.01,0.03144960105419159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,32,power_law_1.2,0.03081600069999695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,32,power_law_1.01,0.1652351975440979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,32,power_law_1.01,0.042854401469230655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,32,power_law_1.01,0.03201920092105866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,32,power_law_1.2,0.03230080008506775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,32,power_law_1.01,0.2320255994796753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,32,power_law_1.01,0.04283519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,32,power_law_1.01,0.031699201464653014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,32,power_law_1.2,0.03287039995193482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,32,power_law_1.01,0.3359551906585693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,32,power_law_1.01,0.04424319863319397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,32,power_law_1.01,0.03159680068492889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,32,power_law_1.2,0.033478400111198424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,32,power_law_1.01,0.43319039344787597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,32,power_law_1.01,0.029708799719810487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,32,power_law_1.2,0.03467519879341126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,32,power_law_1.01,0.047660800814628604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,32,balanced,0.36052266756693524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,32,power_law_1.01,0.5615744113922119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,32,power_law_1.01,0.03048959970474243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,32,power_law_1.2,0.03578239977359772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,32,power_law_1.01,0.05071359872817993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,32,power_law_1.01,0.031692799925804135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,32,power_law_1.01,1.2935999870300292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,32,power_law_1.01,0.05738239884376526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,32,power_law_1.2,0.03840000033378601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,32,power_law_1.01,0.03335039913654327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,32,power_law_1.01,0.06410239934921265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,32,power_law_1.2,0.041740798950195314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,32,power_law_1.01,0.03619199991226196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,32,power_law_1.2,0.03660799860954285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,32,power_law_1.01,0.07271680235862732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,32,power_law_1.2,0.043584001064300534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,32,power_law_1.01,0.043244799971580504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,32,power_law_1.2,0.028966400027275085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,32,power_law_1.01,0.08918399810791015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,32,power_law_1.2,0.044064000248909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,32,power_law_1.01,0.049209600687026976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,32,power_law_1.2,0.02834559977054596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,32,power_law_1.01,0.12407679557800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,32,power_law_1.2,0.04936319887638092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,32,power_law_1.01,0.05797759890556335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,32,power_law_1.2,0.03031040132045746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,32,power_law_1.01,0.14896639585494995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,32,power_law_1.2,0.052985602617263795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,32,power_law_1.01,0.06808320283889771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,32,power_law_1.2,0.030854400992393494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,32,power_law_1.01,0.17764480113983155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,32,power_law_1.2,0.05791360139846802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,32,power_law_1.2,0.03225600123405457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,32,power_law_1.01,0.08773120045661927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,32,power_law_1.01,0.27104640007019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,32,balanced,0.5605119864145914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,32,power_law_1.2,0.07137280106544494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,32,power_law_1.2,0.03223040103912354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,32,power_law_1.01,0.11857279539108276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,32,power_law_1.01,0.2825023889541626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,32,power_law_1.2,0.08592000007629394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,32,power_law_1.2,0.03318400084972382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,32,power_law_1.01,0.16242560148239135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,32,power_law_1.01,0.4633664131164551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,32,power_law_1.2,0.10188800096511841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,32,power_law_1.2,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,32,power_law_1.01,0.2132863998413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,32,power_law_1.01,0.8805760383605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,32,power_law_1.2,0.13938560485839843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,32,power_law_1.2,0.033932799100875856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,32,power_law_1.01,0.26557440757751466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,32,power_law_1.2,0.03467519879341126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,32,power_law_1.2,0.1903807997703552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,32,power_law_1.01,0.40252161026000977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,32,power_law_1.2,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,32,power_law_1.2,0.26899840831756594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,32,power_law_1.2,0.0377920001745224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,32,power_law_1.01,0.734668779373169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,32,power_law_1.2,0.35792639255523684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,32,power_law_1.2,0.037574398517608645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,32,power_law_1.2,0.5162176132202149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,32,power_law_1.2,0.042559999227523806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,32,power_law_1.2,0.6705408096313477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,32,power_law_1.2,0.04302079975605011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,32,power_law_1.2,1.7845375061035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,32,power_law_1.2,0.045587199926376346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,32,power_law_1.2,0.04935680031776428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,32,power_law_1.2,0.05368319749832153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,32,power_law_1.2,0.06543999910354614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,32,power_law_1.2,0.0698303997516632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,32,power_law_1.2,0.0912447988986969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,32,power_law_1.2,0.10714880228042603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,32,power_law_1.2,0.03568640053272247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,32,power_law_1.2,0.1878592014312744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,32,balanced,1.1003786722819011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,32,power_law_1.2,0.031020799279212953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,32,power_law_1.2,0.21877760887145997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,32,power_law_1.2,0.03121280074119568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,32,power_law_1.2,0.3060800075531006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,32,power_law_1.2,0.033369600772857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,32,power_law_1.2,0.3774847984313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,32,power_law_1.2,0.03330560028553009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,32,power_law_1.2,0.6205311775207519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,32,power_law_1.2,0.03425279855728149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,32,power_law_1.2,0.8793087959289551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,32,power_law_1.2,0.03457280099391937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,32,power_law_1.2,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,32,power_law_1.2,2.28790397644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,32,power_law_1.2,0.035699200630187986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,32,power_law_1.2,0.03601920008659363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,32,power_law_1.2,0.036185601353645326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,32,power_law_1.2,0.03723520040512085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,32,power_law_1.2,0.0379584014415741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,32,power_law_1.2,0.03901439905166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,32,power_law_1.2,0.042028799653053284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,32,power_law_1.2,0.04334079921245575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,32,power_law_1.2,0.046988800168037415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,32,power_law_1.2,0.05198079943656921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,32,power_law_1.2,0.05230720043182373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,32,power_law_1.2,0.059513598680496216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,32,power_law_1.2,0.06792320013046264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,32,power_law_1.2,0.0818943977355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,32,power_law_1.2,0.10458240509033204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,32,power_law_1.2,0.1574399948120117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,32,power_law_1.2,0.1813248038291931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,32,power_law_1.2,0.28855679035186765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,32,power_law_1.2,0.3592576026916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,32,power_law_1.2,0.3724544048309326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,32,power_law_1.2,0.8844415664672851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,32,power_law_1.2,1.498591995239258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,32,power_law_1.2,0.01704320013523102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,32,power_law_1.2,0.016889600455760954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,32,power_law_1.2,0.017088000476360322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,32,power_law_1.2,0.017433600127696992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,32,power_law_1.2,0.019065600633621217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,32,power_law_1.2,0.022086399793624877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,32,power_law_1.2,0.03363839983940124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,32,power_law_1.2,0.032864001393318173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,32,power_law_1.2,0.03255679905414581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,32,power_law_1.2,0.03144960105419159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,32,power_law_1.2,0.03125759959220886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,32,power_law_1.2,0.031385600566864014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,32,power_law_1.2,0.030931198596954347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,32,power_law_1.2,0.029747200012207032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,32,power_law_1.2,0.02940160036087036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,32,power_law_1.2,0.0322816014289856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,32,power_law_1.2,0.03167999982833862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,32,power_law_1.2,0.033670398592948916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,32,power_law_1.2,0.03712640106678009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,32,power_law_1.2,0.04564479887485504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,32,power_law_1.2,0.0492031991481781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,32,power_law_1.2,0.06430720090866089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,32,power_law_1.2,0.0753600001335144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,32,power_law_1.2,0.10103679895401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,32,power_law_1.2,0.13890559673309327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,32,power_law_1.2,0.1837056040763855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,32,power_law_1.2,0.23530240058898927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,32,power_law_1.2,0.29640960693359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,32,power_law_1.2,0.491974401473999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,32,power_law_1.2,0.9430912017822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,16,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,16,balanced,0.035802667339642845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,16,balanced,0.019274666905403137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,16,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,16,balanced,0.03669866671164831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,16,balanced,0.017184000462293625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,16,balanced,0.03754133234421412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,16,balanced,0.01729600007335345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,16,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,16,balanced,0.0191040001809597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,16,balanced,0.03782933453718821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,16,balanced,0.019120000302791595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,16,balanced,0.03773866593837738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,16,balanced,0.01894933357834816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,16,balanced,0.03939199944337209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,16,balanced,0.02093333254257838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,16,balanced,0.038133333126703896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,16,balanced,0.021295999487241108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,16,balanced,0.03782399992148081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,16,balanced,0.02313599983851115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,16,balanced,0.03944533318281174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,16,balanced,0.022997332115968067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,16,balanced,0.03961066653331121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,16,balanced,0.024698667228221893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,16,balanced,0.03984000037113825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,16,balanced,0.025407999753952026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,16,balanced,0.043509334325790405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,16,balanced,0.025205334027608235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,16,balanced,0.025285333395004272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,16,balanced,0.043621331453323364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,16,power_law_1.01,0.03792639970779419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,16,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,16,balanced,0.027237333357334137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,16,balanced,0.045850664377212524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,16,balanced,0.037050666908423104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,16,power_law_1.01,0.03420799970626831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,16,balanced,0.025727999707063038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,16,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,16,balanced,0.03694933404525121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,16,power_law_1.01,0.032153600454330446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,16,balanced,0.027119999130566914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,16,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,16,balanced,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,16,power_law_1.01,0.03379839956760407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,16,balanced,0.02789866675933202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,16,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,16,power_law_1.01,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,16,balanced,0.05211733281612396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,16,balanced,0.02923733244339625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,16,balanced,0.04146133363246918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,16,power_law_1.01,0.035545599460601804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,16,balanced,0.05825066566467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,16,balanced,0.030826665461063385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,16,balanced,0.0498933345079422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,16,power_law_1.01,0.03610239923000336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,16,balanced,0.03145600110292435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,16,balanced,0.03133333226044973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,16,balanced,0.06381866832574208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,16,balanced,0.0551146666208903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,16,power_law_1.01,0.03663359880447388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,16,balanced,0.03331200033426285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,16,balanced,0.03124266614516576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,16,power_law_1.01,0.03640320003032684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,16,balanced,0.07052266597747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,16,balanced,0.0701386680205663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,16,balanced,0.03348266581694285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,16,balanced,0.029322666426499683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,16,power_law_1.01,0.03670400083065033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,16,balanced,0.0316746657093366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,16,balanced,0.031311998764673867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,16,balanced,0.08169066905975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,16,balanced,0.09303999940554301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,16,power_law_1.01,0.03776639997959137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,16,balanced,0.031189332405726116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,16,balanced,0.032405334214369454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,16,power_law_1.01,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,16,balanced,0.11203199625015259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,16,balanced,0.03178666780392329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,16,power_law_1.01,0.03875199854373932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,16,balanced,0.033786666889985405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,16,balanced,0.11731200416882832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,16,balanced,0.035530666510264076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,16,power_law_1.01,0.040703999996185306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,16,balanced,0.13808000087738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,16,balanced,0.03443199892838796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,16,balanced,0.03781333317359289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,16,power_law_1.01,0.043961599469184875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,16,balanced,0.033359999457995095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,16,balanced,0.03846933444341024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,16,balanced,0.15084266662597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,16,power_law_1.01,0.04577920138835907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,16,balanced,0.03456533451875051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,16,balanced,0.0376800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,16,balanced,0.19244800011316934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,16,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,16,power_law_1.01,0.048614400625228885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,16,balanced,0.035487999518712364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,16,balanced,0.03966933240493139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,16,power_law_1.01,0.028307199478149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,16,power_law_1.01,0.05057920217514038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,16,balanced,0.03566399961709976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,16,balanced,0.1876373291015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,16,balanced,0.041877334316571556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,16,power_law_1.01,0.025702399015426636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,16,power_law_1.01,0.05489280223846436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,16,balanced,0.03572266548871994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,16,balanced,0.24698134263356528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,16,power_law_1.01,0.02502399981021881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,16,balanced,0.043712000052134194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,16,power_law_1.01,0.06259199976921082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,16,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,16,power_law_1.01,0.03826560080051422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,16,power_law_1.01,0.0244159996509552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,16,power_law_1.01,0.07022719979286193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,16,balanced,0.04621866842110952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,16,balanced,0.04065600037574768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,16,balanced,0.22095467646916708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,16,power_law_1.01,0.0328575998544693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,16,power_law_1.01,0.026604801416397095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,16,power_law_1.01,0.08983039855957031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,16,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,16,balanced,0.054085334142049156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,16,balanced,0.30214399099349976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,16,power_law_1.01,0.030476799607276915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,16,power_law_1.01,0.02794240117073059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,16,balanced,0.05963733295599619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,16,power_law_1.01,0.10711040496826171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,16,balanced,0.04155199974775314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,16,power_law_1.01,0.03089280128479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,16,power_law_1.01,0.02849920094013214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,16,power_law_1.01,0.1344831943511963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,16,balanced,0.0436160018046697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,16,balanced,0.07418133318424225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,16,power_law_1.01,0.031251201033592226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,16,power_law_1.01,0.03216640055179596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,16,power_law_1.01,0.16266239881515504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,16,balanced,0.31598933537801105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,16,power_law_1.01,0.03341439962387085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,16,power_law_1.01,0.2317888021469116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,16,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,16,power_law_1.01,0.03436160087585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,16,balanced,0.09471999605496724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,16,power_law_1.01,0.03249920010566711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,16,power_law_1.01,0.3317055940628052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,16,balanced,0.46669332186381024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,16,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,16,power_law_1.01,0.034560000896453856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,16,power_law_1.01,0.03320960104465485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,16,power_law_1.01,0.39009919166564944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,16,balanced,0.05397333204746246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,16,power_law_1.01,0.03699840009212494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,16,power_law_1.01,0.033766400814056394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,16,balanced,0.11514666676521301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,16,power_law_1.01,0.5440832138061523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,16,power_law_1.01,0.04039680063724518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,16,balanced,0.06237866481145223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,16,power_law_1.01,0.03413119912147522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,16,power_law_1.01,0.0416703999042511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,16,power_law_1.01,1.2348544120788574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,16,power_law_1.01,0.034355199337005614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,16,balanced,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,16,power_law_1.01,0.03660799860954285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,16,power_law_1.01,0.035436800122261046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,16,balanced,0.1381760040918986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,16,power_law_1.01,0.04044800102710724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,16,balanced,0.10226133465766907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,16,power_law_1.01,0.036364799737930296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,16,power_law_1.01,0.04178560078144074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,16,power_law_1.01,0.03918719887733459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,16,power_law_1.01,0.0430976003408432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,16,balanced,0.5892906586329142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,16,balanced,0.13167466719945273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,16,power_law_1.01,0.04284160137176514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,16,power_law_1.01,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,16,power_law_1.01,0.043558400869369504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,16,balanced,0.16114667057991028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,16,power_law_1.01,0.050432002544403075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,16,power_law_1.01,0.044275200366973876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,16,balanced,0.9103626410166422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,16,power_law_1.01,0.05792639851570129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,16,power_law_1.01,0.04794879853725433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,16,balanced,0.1759200096130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,16,power_law_1.01,0.06170880198478699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,16,power_law_1.01,0.05236480236053467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,16,power_law_1.01,0.08028799891471863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,16,power_law_1.01,0.06062080264091492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,16,power_law_1.01,0.08251519799232483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,16,power_law_1.01,0.06801279783248901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,16,balanced,0.21970132986704508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,16,power_law_1.01,0.11530239582061767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,16,balanced,0.2238346735636393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,16,power_law_1.01,0.09006080031394958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,16,power_law_1.01,0.1392575979232788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,16,power_law_1.01,0.10062719583511352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,16,power_law_1.01,0.19633280038833617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,16,power_law_1.01,0.14888960123062134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,16,power_law_1.01,0.2663104057312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,16,power_law_1.01,0.17344640493392943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,16,power_law_1.01,0.3091327905654907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,16,balanced,0.26492265860239667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,16,power_law_1.01,0.25012478828430174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,16,power_law_1.01,0.4414463996887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,16,power_law_1.01,0.33315839767456057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,16,power_law_1.01,0.7842944145202637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,16,power_law_1.01,0.3872960090637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,16,power_law_1.01,0.6331136226654053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,16,power_law_1.01,1.3149375915527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,16,balanced,0.3835573196411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,16,balanced,0.4068106810251872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,16,balanced,0.7325812975565592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,16,power_law_1.01,0.01687680035829544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,16,power_law_1.01,0.016710400581359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,16,power_law_1.01,0.017292800545692443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,16,power_law_1.01,0.016761599481105803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,16,power_law_1.01,0.017907199263572694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,16,power_law_1.01,0.01865600049495697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,16,power_law_1.01,0.02195200026035309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,16,power_law_1.01,0.021376000344753267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,16,power_law_1.01,0.03317759931087494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,16,power_law_1.01,0.03342719972133636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,16,power_law_1.01,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,16,power_law_1.01,0.032204800844192506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,16,power_law_1.01,0.03201279938220978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,16,power_law_1.01,0.03131519854068756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,16,power_law_1.01,0.03201920092105866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,16,power_law_1.01,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,16,power_law_1.01,0.033062401413917544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,16,power_law_1.01,0.035180801153182985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,16,power_law_1.01,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,16,power_law_1.01,0.04134399890899658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,16,power_law_1.01,0.04407680034637451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,16,power_law_1.01,0.05630720257759094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,16,power_law_1.01,0.06418560147285461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,16,power_law_1.01,0.08611840009689331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,16,power_law_1.01,0.10945279598236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,16,power_law_1.01,0.15144959688186646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,16,power_law_1.01,0.18341120481491088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,16,power_law_1.01,0.2227328062057495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,16,power_law_1.01,0.3572479963302612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,16,power_law_1.01,0.6732096195220947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,16,power_law_1.2,0.027084800601005554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,16,power_law_1.2,0.03692159950733185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,16,power_law_1.2,0.025145599246025087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,16,power_law_1.2,0.028896000981330872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,16,power_law_1.2,0.02221439927816391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,16,power_law_1.2,0.028672000765800475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,16,power_law_1.2,0.0251008003950119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,16,power_law_1.2,0.03017599880695343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,16,power_law_1.2,0.02507520020008087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,16,power_law_1.2,0.03132160007953644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,16,power_law_1.2,0.03738240003585815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,16,power_law_1.2,0.028166401386260986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,16,power_law_1.2,0.03282560110092163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,16,power_law_1.2,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,16,power_law_1.2,0.028121599555015565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,16,power_law_1.2,0.03251200020313263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,16,power_law_1.2,0.03258880078792572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,16,power_law_1.2,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,16,power_law_1.2,0.03357439935207367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,16,power_law_1.2,0.03438720107078552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,16,power_law_1.2,0.034329599142074584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,16,power_law_1.2,0.034995201230049136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,16,power_law_1.2,0.03487359881401062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,16,power_law_1.2,0.03492479920387268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,16,power_law_1.2,0.03516159951686859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,16,power_law_1.2,0.03628160059452057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,16,power_law_1.2,0.03620480000972748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,16,power_law_1.2,0.035872000455856326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,16,power_law_1.2,0.036268800497055054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,16,power_law_1.2,0.036652800440788266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,16,power_law_1.2,0.036006399989128114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,16,power_law_1.2,0.03663359880447388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,16,power_law_1.2,0.03708159923553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,16,power_law_1.2,0.038099199533462524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,16,power_law_1.2,0.03665919899940491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,16,power_law_1.2,0.03800320029258728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,16,power_law_1.2,0.01719679981470108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,16,power_law_1.2,0.03842560052871704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,16,power_law_1.2,0.03830400109291077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,16,power_law_1.2,0.04085760116577149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,16,power_law_1.2,0.016543999314308167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,16,power_law_1.2,0.04413439929485321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,16,power_law_1.2,0.03760640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,16,power_law_1.2,0.042412799596786496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,16,power_law_1.2,0.016889600455760954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,16,power_law_1.2,0.04526079893112182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,16,power_law_1.2,0.03919360041618347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,16,power_law_1.2,0.01759999990463257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,16,power_law_1.2,0.04647040069103241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,16,power_law_1.2,0.044326400756835936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,16,power_law_1.2,0.04012799859046936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,16,power_law_1.2,0.01842560023069382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,16,power_law_1.2,0.050195199251174924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,16,power_law_1.2,0.04824320077896118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,16,power_law_1.2,0.041254401206970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,16,power_law_1.2,0.018828800320625304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,16,power_law_1.2,0.053401601314544675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,16,power_law_1.2,0.04525440037250519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,16,power_law_1.2,0.051648002862930295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,16,power_law_1.2,0.021593600511550903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,16,power_law_1.2,0.04755200147628784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,16,power_law_1.2,0.0663424015045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,16,power_law_1.2,0.060678398609161376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,16,power_law_1.2,0.021254399418830873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,16,power_law_1.2,0.048895999789237976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,16,power_law_1.2,0.07014399766921997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,16,power_law_1.2,0.06920959949493408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,16,power_law_1.2,0.03312639892101288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,16,power_law_1.2,0.05377280116081238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,16,power_law_1.2,0.09729920029640197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,16,power_law_1.2,0.08084480166435241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,16,power_law_1.2,0.03335039913654327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,16,power_law_1.2,0.05772799849510193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,16,power_law_1.2,0.11872639656066894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,16,power_law_1.2,0.09715840220451355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,16,power_law_1.2,0.03308799862861633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,16,power_law_1.2,0.06835839748382569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,16,power_law_1.2,0.1654207944869995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,16,power_law_1.2,0.1442944049835205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,16,power_law_1.2,0.03191039860248566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,16,power_law_1.2,0.07504000067710877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,16,power_law_1.2,0.2217087984085083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,16,power_law_1.2,0.18291840553283692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,16,power_law_1.2,0.03210879862308502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,16,power_law_1.2,0.1023103952407837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,16,power_law_1.2,0.32441599369049073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,16,power_law_1.2,0.2188352108001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,16,power_law_1.2,0.030508801341056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,16,power_law_1.2,0.13076479434967042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,16,power_law_1.2,0.5530240058898925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,16,power_law_1.2,0.35202560424804685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,16,power_law_1.2,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,16,power_law_1.2,0.17345919609069824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,16,power_law_1.2,0.6351871967315674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,16,power_law_1.2,0.449567985534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,16,power_law_1.2,0.03175039887428284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,16,power_law_1.2,0.20469119548797607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,16,power_law_1.2,1.0775808334350585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,16,power_law_1.2,0.6708352088928222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,16,power_law_1.2,0.032569599151611325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,16,power_law_1.2,0.30970239639282227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,16,power_law_1.2,2.2904895782470702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,16,power_law_1.2,1.2026047706604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,16,power_law_1.2,0.035622400045394895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,16,power_law_1.2,0.4326335906982422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,16,power_law_1.2,0.036774399876594546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,16,power_law_1.2,0.48387842178344725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,16,power_law_1.2,0.04197759926319122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,16,power_law_1.2,0.985478401184082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,16,power_law_1.2,0.047007998824119566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,16,power_law_1.2,2.2413183212280274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,16,power_law_1.2,0.05772160291671753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,16,power_law_1.2,0.07088000178337098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,16,power_law_1.2,0.09374079704284669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,16,power_law_1.2,0.11627520322799682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,16,power_law_1.2,0.1618175983428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,16,power_law_1.2,0.20865280628204347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,16,power_law_1.2,0.25289599895477294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,16,power_law_1.2,0.4052864074707031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,16,power_law_1.2,0.7317887783050537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,8,balanced,0.02693866689999898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,8,balanced,0.024149333437283833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,8,balanced,0.025221332907676697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,8,balanced,0.027104000250498455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,8,balanced,0.027376001079877216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,8,balanced,0.02754666656255722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,8,balanced,0.027295999228954315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,8,balanced,0.027376001079877216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,8,balanced,0.027349332968393963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,8,balanced,0.029487999776999157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,8,balanced,0.030085332691669464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,8,balanced,0.02961066613594691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,8,balanced,0.02917333443959554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,8,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,8,balanced,0.03740799923737844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,8,balanced,0.03781333317359289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,8,balanced,0.03956799954175949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,8,balanced,0.041696002086003624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,8,balanced,0.04809066653251648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,8,balanced,0.0315733328461647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,8,balanced,0.04806933303674062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,8,balanced,0.029994666576385498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,8,balanced,0.03126933425664902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,8,balanced,0.05340266724427541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,8,balanced,0.03198933353026708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,8,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,8,balanced,0.0602453351020813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,8,balanced,0.03540800015131632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,8,balanced,0.03566933423280716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,8,balanced,0.07437333464622498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,8,balanced,0.035760000348091125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,8,balanced,0.03753600021203359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,8,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,8,balanced,0.09814400474230449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,8,balanced,0.037087999284267426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,8,balanced,0.03775999943415324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,8,balanced,0.03938666731119156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,8,balanced,0.12105600039164226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,8,balanced,0.037818667789300285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,8,power_law_1.01,0.027584001421928406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,8,balanced,0.04261866708596548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,8,power_law_1.01,0.02396800071001053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,8,balanced,0.0436106671889623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,8,power_law_1.01,0.024083200097084045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,8,balanced,0.043968002001444496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,8,power_law_1.01,0.025203201174736022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,8,balanced,0.16477866967519125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,8,balanced,0.04562666515509287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,8,power_law_1.01,0.02651520073413849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,8,balanced,0.05189866820971171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,8,power_law_1.01,0.025939199328422546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,8,power_law_1.01,0.026572799682617186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,8,balanced,0.05787733197212219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,8,power_law_1.01,0.028204798698425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,8,balanced,0.06413333117961884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,8,power_law_1.01,0.031091201305389404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,8,power_law_1.01,0.031539198756217954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,8,balanced,0.19334399700164795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,8,balanced,0.07454933226108551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,8,power_law_1.01,0.03247359991073608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,8,power_law_1.01,0.03351680040359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,8,balanced,0.0920799970626831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,8,power_law_1.01,0.03432320058345795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,8,power_law_1.01,0.03585920035839081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,8,power_law_1.01,0.03949440121650696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,8,balanced,0.12648000319798788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,8,power_law_1.01,0.04049279987812042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,8,power_law_1.01,0.04501760005950928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,8,balanced,0.24016000827153525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,8,power_law_1.01,0.04915199875831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,8,balanced,0.1662613352139791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,8,power_law_1.01,0.054662400484085084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,8,power_law_1.01,0.06298879981040954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,8,power_law_1.01,0.06695039868354798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,8,power_law_1.01,0.08840960264205933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,8,power_law_1.01,0.10490239858627319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,8,balanced,0.23471466700236002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,8,power_law_1.01,0.13997440338134765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,8,power_law_1.01,0.15478399991989136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,8,power_law_1.01,0.25626239776611326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,8,power_law_1.01,0.27164158821105955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,8,balanced,0.3470826546351115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,8,power_law_1.01,0.370796799659729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,8,balanced,0.29039466381073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,8,power_law_1.01,0.5845312118530274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,8,power_law_1.01,1.055788803100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,8,balanced,0.3581013282140096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,8,power_law_1.01,0.034355199337005614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,8,power_law_1.01,0.030425599217414855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,8,power_law_1.01,0.02884480059146881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,8,power_law_1.01,0.03076480031013489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,8,power_law_1.01,0.030700799822807313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,8,power_law_1.01,0.03361920118331909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,8,power_law_1.01,0.033907198905944826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,8,power_law_1.01,0.03497599959373474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,8,power_law_1.01,0.03516159951686859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,8,balanced,0.5449706713358561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,8,power_law_1.01,0.03665919899940491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,8,balanced,0.6491146485010783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,8,power_law_1.01,0.036774399876594546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,8,power_law_1.01,0.03743999898433685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,8,power_law_1.01,0.03962239921092987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,8,power_law_1.01,0.03911679983139038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,8,power_law_1.01,0.04333440065383911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,8,power_law_1.01,0.04673919975757599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,8,power_law_1.01,0.04872959852218628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,8,power_law_1.01,0.053887999057769774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,8,power_law_1.01,0.06120960116386413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,8,power_law_1.01,0.06990079879760742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,8,power_law_1.01,0.07360640168190002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,8,power_law_1.01,0.09857919812202454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,8,power_law_1.01,0.12402559518814087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,8,power_law_1.01,0.1749567985534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,8,power_law_1.01,0.218720006942749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,8,power_law_1.01,0.33990399837493895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,8,balanced,1.0638559659322102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,8,power_law_1.01,0.41680002212524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,8,power_law_1.01,0.5130303859710693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,8,power_law_1.01,0.8163968086242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,8,power_law_1.01,1.5299903869628906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,8,power_law_1.2,0.026649600267410277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,8,power_law_1.2,0.02369920015335083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,8,power_law_1.2,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,8,power_law_1.2,0.022329600155353548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,8,power_law_1.2,0.02710399925708771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,8,power_law_1.2,0.025260800123214723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,8,power_law_1.2,0.027423998713493346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,8,power_law_1.2,0.02682879865169525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,8,power_law_1.2,0.03033599853515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,8,power_law_1.2,0.02686080038547516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,8,power_law_1.2,0.030604800581932066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,8,power_law_1.2,0.026969599723815917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,8,power_law_1.2,0.03412480056285858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,8,power_law_1.2,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,8,power_law_1.2,0.03394559919834137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,8,power_law_1.2,0.032262399792671204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,8,power_law_1.2,0.03480319976806641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,8,power_law_1.2,0.032627201080322264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,8,power_law_1.2,0.0348800003528595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,8,power_law_1.2,0.03322240114212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,8,power_law_1.2,0.035206401348114015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,8,power_law_1.2,0.03449600040912628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,8,power_law_1.2,0.03671039938926697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,8,power_law_1.2,0.03555839955806732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,8,power_law_1.2,0.03856000006198883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,8,power_law_1.2,0.03751679956912994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,8,power_law_1.2,0.03999359905719757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,8,power_law_1.2,0.03935360014438629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,8,power_law_1.2,0.04077439904212952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,8,power_law_1.2,0.04478079974651337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,8,power_law_1.2,0.04110080003738403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,8,power_law_1.2,0.04721280038356781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,8,power_law_1.2,0.04509440064430237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,8,power_law_1.2,0.04911360144615173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,8,power_law_1.2,0.050348800420761106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,8,power_law_1.2,0.053964799642562865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,8,power_law_1.2,0.056383997201919556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,8,power_law_1.2,0.06277120113372803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,8,power_law_1.2,0.07192320227622986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,8,power_law_1.2,0.06600959897041321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,8,power_law_1.2,0.08558080196380616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,8,power_law_1.2,0.07678719758987426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,8,power_law_1.2,0.10150400400161744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,8,power_law_1.2,0.09848960041999817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,8,power_law_1.2,0.133951997756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,8,power_law_1.2,0.11408640146255493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,8,power_law_1.2,0.1918272018432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,8,power_law_1.2,0.14137599468231202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,8,power_law_1.2,0.2631808042526245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,8,power_law_1.2,0.20667519569396972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,8,power_law_1.2,0.36307199001312257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,8,power_law_1.2,0.3144767999649048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,8,power_law_1.2,0.4664000034332275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,8,power_law_1.2,0.37049601078033445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,8,power_law_1.2,0.6669312000274659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,8,power_law_1.2,0.49271039962768554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,8,power_law_1.2,1.1832127571105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,8,power_law_1.2,0.800102424621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,8,power_law_1.2,2.288089561462402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,8,power_law_1.2,1.43505916595459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,8,balanced,0.035818666219711304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,8,balanced,0.03796799977620443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,8,balanced,0.03498666733503342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,8,balanced,0.032655999064445496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,8,balanced,0.03336533407370249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,8,balanced,0.033333333830038704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,8,balanced,0.03344533344109853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,8,balanced,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,8,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,8,balanced,0.03331200033426285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,8,balanced,0.03379733363787333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,8,balanced,0.03141866624355316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,8,balanced,0.03533866753180822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,8,balanced,0.031685332457224526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,8,balanced,0.033770665526390076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,8,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,8,balanced,0.033573334415753685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,8,balanced,0.03350399931271871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,8,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,8,balanced,0.03368533402681351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,8,balanced,0.035818666219711304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,8,balanced,0.03352533280849457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,8,balanced,0.04051200052102407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,8,balanced,0.03339733431736628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,8,balanced,0.039936001102129616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,8,balanced,0.03572266548871994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,8,balanced,0.041877334316571556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,8,balanced,0.034645333886146545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,8,balanced,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,8,balanced,0.035546667873859406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,8,balanced,0.04459733267625173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,8,balanced,0.03438399980465571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,8,balanced,0.05680533250172933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,8,balanced,0.03578133384386698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,8,balanced,0.060122668743133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,8,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,8,balanced,0.0827893316745758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,8,balanced,0.04083200047413508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,8,balanced,0.07999999821186066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,8,balanced,0.0415040006240209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,8,balanced,0.10973866780598958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,8,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,8,balanced,0.04776533444722494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,8,balanced,0.12265066305796306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,8,balanced,0.058117335041364036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,8,balanced,0.16265599926312765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,8,balanced,0.06018666426340739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,8,balanced,0.0906986693541209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,8,balanced,0.1933493415514628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,8,balanced,0.10719466209411621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,8,balanced,0.14921067158381143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,8,balanced,0.22775999704996744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,8,balanced,0.1790026624997457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,8,balanced,0.24286933739980063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,8,balanced,0.32924799124399823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,8,balanced,0.29556800921758014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,8,balanced,0.3569120168685913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,8,balanced,0.586949348449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,8,balanced,0.5448053280512491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,8,balanced,1.0274346669514973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,8,balanced,0.0358240008354187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,8,balanced,0.0377813329299291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,8,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,8,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,8,balanced,0.03621866554021835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,8,power_law_1.01,0.03501439988613129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,8,balanced,0.03584533433119456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,8,power_law_1.01,0.03462400138378143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,8,balanced,0.037530665596326195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,8,power_law_1.01,0.031411200761795044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,8,balanced,0.036117332677046456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,8,power_law_1.01,0.03165439963340759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,8,balanced,0.035760000348091125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,8,power_law_1.01,0.03274880051612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,8,balanced,0.03770666569471359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,8,power_law_1.01,0.02929919958114624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,8,power_law_1.01,0.032543998956680295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,8,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,8,balanced,0.019253333409627277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,8,power_law_1.01,0.028569599986076354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,8,power_law_1.01,0.033030399680137636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,8,balanced,0.039621333281199135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,8,balanced,0.019274666905403137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,8,power_law_1.01,0.028806400299072266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,8,power_law_1.01,0.035097599029541016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,8,balanced,0.037791999677817024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,8,balanced,0.019002666076024372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,8,power_law_1.01,0.03049600124359131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,8,power_law_1.01,0.03739520013332367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,8,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,8,balanced,0.019359999646743137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,8,power_law_1.01,0.030585598945617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,8,power_law_1.01,0.03917439877986908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,8,balanced,0.019509332875410717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,8,balanced,0.044026667873064675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,8,power_law_1.01,0.03149439990520477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,8,power_law_1.01,0.03651840090751648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,8,balanced,0.02117866774400075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,8,balanced,0.042538667718569435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,8,power_law_1.01,0.03281280100345611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,8,power_law_1.01,0.04134399890899658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,8,balanced,0.02128000060717265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,8,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,8,power_law_1.01,0.033523198962211606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,8,balanced,0.022757334013779957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,8,power_law_1.01,0.0421887993812561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,8,power_law_1.01,0.03509120047092438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,8,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,8,balanced,0.023541333774725597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,8,power_law_1.01,0.03591040074825287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,8,power_law_1.01,0.040633600950241086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,8,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,8,balanced,0.023408000667889912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,8,power_law_1.01,0.03505280017852783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,8,power_law_1.01,0.056883198022842404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,8,balanced,0.05603733162085215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,8,balanced,0.03143466760714849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,8,power_law_1.01,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,8,power_law_1.01,0.03754880130290985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,8,balanced,0.036917333801587425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,8,power_law_1.01,0.05682560205459595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,8,balanced,0.056234667698542275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,8,power_law_1.01,0.04012160003185272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,8,balanced,0.03510933369398117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,8,power_law_1.01,0.06396160125732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,8,balanced,0.07266133526961009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,8,power_law_1.01,0.04118399918079376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,8,balanced,0.03166399896144867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,8,power_law_1.01,0.07567359805107117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,8,power_law_1.01,0.04811519980430603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,8,balanced,0.03758399933576584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,8,power_law_1.01,0.09388160109519958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,8,balanced,0.08328000207742055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,8,balanced,0.033717334270477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,8,power_law_1.01,0.050969600677490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,8,power_law_1.01,0.10376319885253907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,8,balanced,0.03572800010442734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,8,power_law_1.01,0.052857601642608644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,8,balanced,0.10755200187365214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,8,power_law_1.01,0.1505087971687317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,8,balanced,0.04781333108743032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,8,power_law_1.01,0.06170240044593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,8,power_law_1.01,0.16493439674377441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,8,power_law_1.01,0.06903039813041686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,8,balanced,0.12585066755612692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,8,balanced,0.050437331199645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,8,power_law_1.01,0.2155519962310791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,8,power_law_1.01,0.08919039964675904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,8,power_law_1.01,0.2831167936325073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,8,balanced,0.06469333171844482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,8,power_law_1.01,0.10142079591751099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,8,power_law_1.01,0.41661438941955564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,8,balanced,0.1656053364276886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,8,balanced,0.07652799785137177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,8,power_law_1.01,0.13860479593276978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,8,power_law_1.01,0.49233279228210447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,8,balanced,0.09967999656995137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,8,power_law_1.01,0.17235840559005738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,8,power_law_1.01,0.6692800045013427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,8,balanced,0.20288532972335815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,8,power_law_1.01,0.22080640792846679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,8,balanced,0.12275733550389607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,8,power_law_1.01,0.907539176940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,8,power_law_1.01,0.3038784027099609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,8,power_law_1.01,1.8460096359252929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,8,power_law_1.01,0.4228544235229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,8,balanced,0.16644799709320068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,8,balanced,0.24124266703923544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,8,power_law_1.01,0.5225728034973145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,8,balanced,0.20850133895874023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,8,power_law_1.01,0.6557375907897949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,8,power_law_1.01,1.00513916015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,8,balanced,0.2949120004971822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,8,balanced,0.3555999994277954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,8,power_law_1.01,1.9097536087036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,8,power_law_1.01,0.01834239959716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,8,power_law_1.01,0.03165439963340759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,8,balanced,0.38493335247039795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,8,power_law_1.01,0.018585599958896637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,8,power_law_1.01,0.033478400111198424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,8,power_law_1.01,0.01794559955596924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,8,power_law_1.01,0.03203200101852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,8,balanced,0.47167468070983887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,8,power_law_1.01,0.018438400328159334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,8,power_law_1.01,0.03367680013179779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,8,power_law_1.01,0.01976960003376007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,8,balanced,0.6502240101496378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,8,power_law_1.01,0.033606401085853575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,8,power_law_1.01,0.020761600136756896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,8,balanced,0.7362346649169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,8,power_law_1.01,0.0345984011888504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,8,power_law_1.01,0.03758719861507416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,8,power_law_1.01,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,8,power_law_1.01,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,8,power_law_1.01,0.03547520041465759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,8,power_law_1.01,0.030124801397323608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,8,power_law_1.01,0.036268800497055054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,8,power_law_1.01,0.029843199253082275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,8,power_law_1.01,0.03672960102558136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,8,power_law_1.01,0.029702401161193846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,8,power_law_1.01,0.03625600039958954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,8,balanced,1.4435200691223145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,8,power_law_1.01,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,8,power_law_1.01,0.037894400954246524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,8,power_law_1.01,0.03386879861354828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,8,power_law_1.01,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,8,power_law_1.01,0.030847999453544616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,8,power_law_1.01,0.040006399154663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,8,power_law_1.01,0.03639039993286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,8,power_law_1.01,0.04659200012683869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,8,power_law_1.01,0.03269760012626648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,8,power_law_1.01,0.048767998814582825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,8,power_law_1.01,0.03455359935760498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,8,power_law_1.01,0.049779200553894044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,8,power_law_1.01,0.04622080028057098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,8,power_law_1.01,0.058911997079849246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.01,0.04816640019416809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.01,0.06662399768829345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.01,0.06264320015907288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.01,0.07338879704475403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.01,0.07561600208282471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.01,0.08062080144882203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.01,0.09768319725990296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.01,0.11208319664001465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.01,0.12026879787445069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.01,0.14529279470443726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.01,0.16442240476608277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.01,0.18504320383071898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.01,0.23070080280303956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.01,0.20635519027709961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.01,0.3214143991470337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.01,0.29361920356750487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.01,0.4272640228271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.01,0.3811455965042114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,8,power_law_1.01,0.4692863941192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,8,power_law_1.01,0.5262720108032226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,8,power_law_1.01,0.7340223789215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,8,power_law_1.01,0.7968575954437256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,8,power_law_1.01,1.439891242980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,8,power_law_1.01,1.2802559852600097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,8,8,power_law_1.2,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,8,8,power_law_1.2,0.03543039858341217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,8,8,power_law_1.2,0.03128960132598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,8,8,power_law_1.2,0.031142398715019226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,8,8,power_law_1.2,0.031788799166679385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,8,8,power_law_1.2,0.03224959969520569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,8,8,power_law_1.2,0.029606398940086365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,8,8,power_law_1.2,0.03319680094718933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,8,8,power_law_1.2,0.029785600304603577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,8,8,power_law_1.2,0.03467519879341126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,8,8,power_law_1.2,0.02911359965801239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,8,8,power_law_1.2,0.03565439879894257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,8,8,power_law_1.2,0.03155199885368347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,8,8,power_law_1.2,0.03790079951286316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,8,8,power_law_1.2,0.03141759932041168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,8,8,power_law_1.2,0.03640320003032684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,8,8,power_law_1.2,0.03187200129032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,8,8,power_law_1.2,0.042342400550842284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,8,8,power_law_1.2,0.0328000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,8,8,power_law_1.2,0.04261760115623474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,8,8,power_law_1.2,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,8,8,power_law_1.2,0.041331198811531064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,8,8,power_law_1.2,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,8,8,power_law_1.2,0.047603198885917665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,8,8,power_law_1.2,0.03790720105171204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,8,8,power_law_1.2,0.053376001119613645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,8,8,power_law_1.2,0.035872000455856326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,8,8,power_law_1.2,0.05480319857597351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,8,8,power_law_1.2,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,8,8,power_law_1.2,0.06901119947433472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,8,8,power_law_1.2,0.04157440066337585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,8,8,power_law_1.2,0.07951359748840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,8,8,power_law_1.2,0.04209280014038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,8,8,power_law_1.2,0.09861760139465332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,8,8,power_law_1.2,0.0497408002614975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,8,8,power_law_1.2,0.11605119705200195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,8,8,power_law_1.2,0.05088000297546387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,8,8,power_law_1.2,0.14539519548416138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,8,8,power_law_1.2,0.05429760217666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,8,8,power_law_1.2,0.1731775999069214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,8,8,power_law_1.2,0.06087679862976074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,8,8,power_law_1.2,0.22670719623565674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,8,8,power_law_1.2,0.06915839910507202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,8,8,power_law_1.2,0.274399995803833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,8,8,power_law_1.2,0.08754559755325317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,8,8,power_law_1.2,0.4203199863433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,8,8,power_law_1.2,0.1021183967590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,8,8,power_law_1.2,0.5163839817047119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,8,8,power_law_1.2,0.14238719940185546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,8,8,power_law_1.2,0.6320064067840576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,8,8,power_law_1.2,0.16535040140151977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,8,8,power_law_1.2,0.9313471794128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,8,8,power_law_1.2,0.23299839496612548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,8,8,power_law_1.2,1.824070358276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,8,8,power_law_1.2,0.2984127998352051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,8,8,power_law_1.2,0.4343616008758545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,8,8,power_law_1.2,0.5708032131195069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,8,8,power_law_1.2,0.6937664031982422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,8,8,power_law_1.2,1.010489559173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,8,8,power_law_1.2,2.0007488250732424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,8,8,power_law_1.2,0.033542400598526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,8,8,power_law_1.2,0.0341376006603241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,8,8,power_law_1.2,0.03333759903907776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,8,8,power_law_1.2,0.03415040075778961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,8,8,power_law_1.2,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,8,8,power_law_1.2,0.03491199910640717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,8,8,power_law_1.2,0.035622400045394895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,8,8,power_law_1.2,0.018195199966430663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,8,8,power_law_1.2,0.03614720106124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,8,8,power_law_1.2,0.018169599771499633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,8,8,power_law_1.2,0.03681280016899109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,8,8,power_law_1.2,0.018182399868965148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,8,8,power_law_1.2,0.03665919899940491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,8,8,power_law_1.2,0.01797119975090027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,8,8,power_law_1.2,0.03656960129737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,8,8,power_law_1.2,0.018291200697422027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,8,8,power_law_1.2,0.038047999143600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,8,8,power_law_1.2,0.020351999998092653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,8,8,power_law_1.2,0.03856000006198883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,8,8,power_law_1.2,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,8,8,power_law_1.2,0.04069119989871979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,8,8,power_law_1.2,0.03834879994392395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,8,8,power_law_1.2,0.04748159945011139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,8,8,power_law_1.2,0.031116798520088196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,8,8,power_law_1.2,0.04747520089149475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,8,8,power_law_1.2,0.02967039942741394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,8,8,power_law_1.2,0.05172479748725891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,8,8,power_law_1.2,0.030060800909996032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,8,8,power_law_1.2,0.057920002937316896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,8,8,power_law_1.2,0.03540480136871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.2,0.06503679752349853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,8,8,power_law_1.2,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.2,0.07494400143623352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,8,8,power_law_1.2,0.03139199912548065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.2,0.08736640214920044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,8,8,power_law_1.2,0.03601920008659363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.2,0.11197439432144166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,8,8,power_law_1.2,0.03288959860801697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.2,0.14615679979324342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,8,8,power_law_1.2,0.03420799970626831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.2,0.1855232000350952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,8,8,power_law_1.2,0.04564479887485504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.2,0.2267712116241455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.2,0.047276800870895384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.2,0.3277440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.2,0.06224640011787415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.2,0.43685760498046877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.2,0.07582719922065735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,8,8,power_law_1.2,0.5444352149963378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.2,0.09697920083999634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,8,8,power_law_1.2,0.7838592052459716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.2,0.12017920017242431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,8,8,power_law_1.2,1.5277055740356444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.2,0.1640768051147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.2,0.20698239803314208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.2,0.2940736055374146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.2,0.3848896026611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,8,8,power_law_1.2,0.4694784164428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,8,8,power_law_1.2,0.7353343963623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,8,8,power_law_1.2,1.4408896446228028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,4,balanced,0.03571200122435888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,4,balanced,0.03329599897066752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,4,balanced,0.03162666658560435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,4,balanced,0.01907733331123988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,4,balanced,0.03329599897066752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,4,balanced,0.0191040001809597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,4,balanced,0.03436266630887985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,4,balanced,0.03585066646337509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,4,balanced,0.01897066707412402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,4,balanced,0.03345066557327906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,4,balanced,0.03753600021203359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,4,balanced,0.018944000204404194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,4,balanced,0.03428266694148382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,4,balanced,0.03612266729275385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,4,balanced,0.01903466631968816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,4,balanced,0.033258666594823204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,4,balanced,0.0354720006386439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,4,balanced,0.019093333433071773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,4,balanced,0.033610666791598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,4,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,4,balanced,0.023039999107519787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,4,balanced,0.035258665680885315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,4,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,4,balanced,0.01940800001223882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,4,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,4,balanced,0.03299733251333237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,4,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,4,balanced,0.02313599983851115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,4,balanced,0.03364266703526179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,4,balanced,0.03137599925200144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,4,balanced,0.03752533346414566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,4,balanced,0.02348266790310542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,4,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,4,balanced,0.033557333052158356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,4,balanced,0.037647999823093414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,4,balanced,0.03426666557788849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,4,balanced,0.0233599990606308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,4,balanced,0.033386667569478355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,4,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,4,balanced,0.023205332458019257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,4,balanced,0.04351999859015147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,4,balanced,0.03151999910672506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,4,balanced,0.03775999943415324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,4,balanced,0.041840001940727234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,4,balanced,0.03346666693687439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,4,balanced,0.025077333052953083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,4,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,4,balanced,0.04154133299986521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,4,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,4,balanced,0.0314026673634847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,4,balanced,0.0378506655494372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,4,balanced,0.04576533536116282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,4,balanced,0.0335359995563825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,4,balanced,0.0339626669883728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,4,balanced,0.04810666541258494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,4,balanced,0.035760000348091125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,4,balanced,0.039493332306543984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,4,balanced,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,4,balanced,0.06022400160630544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,4,balanced,0.0352906659245491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,4,balanced,0.04570133487383524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,4,balanced,0.033674667278925575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,4,balanced,0.06424533327420552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,4,balanced,0.0334346666932106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,4,balanced,0.043696001172065735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,4,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,4,balanced,0.08337600032488506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,4,balanced,0.03366400053103765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,4,balanced,0.04573333263397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,4,balanced,0.04171733558177948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,4,balanced,0.04887466629346212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,4,balanced,0.09343467156092326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,4,balanced,0.035504000882307686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,4,power_law_1.01,0.03293440043926239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,4,balanced,0.05519466598828634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,4,balanced,0.052149335543314614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,4,balanced,0.036533333361148834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,4,balanced,0.13056000073750815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,4,power_law_1.01,0.031040000915527343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,4,balanced,0.06030400097370148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,4,balanced,0.05845333139101664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,4,balanced,0.04376000165939331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,4,power_law_1.01,0.029785600304603577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,4,balanced,0.1451413333415985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,4,power_law_1.01,0.02948479950428009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,4,balanced,0.08105066418647766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,4,balanced,0.06411199768384297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,4,balanced,0.043194666504859924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,4,power_law_1.01,0.03144960105419159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,4,balanced,0.1941439906756083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,4,power_law_1.01,0.03126400113105774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,4,balanced,0.09797867139180501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,4,balanced,0.07875200112660725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,4,balanced,0.0444106658299764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,4,power_law_1.01,0.030316799879074097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,4,power_law_1.01,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,4,balanced,0.2367466688156128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,4,balanced,0.1279039978981018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,4,balanced,0.04801600178082784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,4,balanced,0.10098133484522502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,4,power_law_1.01,0.03132160007953644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,4,balanced,0.2813173333803813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,4,power_law_1.01,0.03272959887981415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,4,balanced,0.05176533261934916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,4,balanced,0.16108799974123636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,4,power_law_1.01,0.03146879971027374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,4,balanced,0.13369066516558328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,4,power_law_1.01,0.032576000690460204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,4,balanced,0.22483199834823608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,4,balanced,0.060309335589408875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,4,balanced,0.4089599847793579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,4,power_law_1.01,0.03209599852561951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,4,balanced,0.15125333269437155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,4,power_law_1.01,0.03288320004940033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,4,balanced,0.06980266670385997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,4,balanced,0.2877440055211385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,4,power_law_1.01,0.03301759958267212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,4,balanced,0.20616000890731812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,4,power_law_1.01,0.034668800234794614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,4,balanced,0.099781334400177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,4,balanced,0.7420159975687662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,4,power_law_1.01,0.034764799475669864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,4,power_law_1.01,0.03375360071659088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,4,balanced,0.3521226644515991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,4,power_law_1.01,0.035436800122261046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,4,balanced,0.25969600677490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,4,balanced,0.12271466851234436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,4,power_law_1.01,0.03463680148124695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,4,power_law_1.01,0.037887999415397645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,4,balanced,0.17148800690968832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,4,balanced,0.5438773234685262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,4,power_law_1.01,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,4,balanced,0.30991466840108234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,4,power_law_1.01,0.037836799025535585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,4,balanced,0.20331199963887533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,4,power_law_1.01,0.035673600435256955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,4,balanced,0.46507732073465985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,4,power_law_1.01,0.04337919950485229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,4,balanced,1.0571786562601726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,4,power_law_1.01,0.03713279962539673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,4,balanced,0.2798080046971639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,4,power_law_1.01,0.04922879934310913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,4,power_law_1.01,0.03735679984092712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,4,balanced,0.349018653233846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,4,power_law_1.01,0.0492031991481781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,4,power_law_1.01,0.041228801012039185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,4,balanced,0.8785813649495443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,4,power_law_1.01,0.05184000134468079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,4,balanced,0.4274933338165283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,4,power_law_1.01,0.0488319993019104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,4,power_law_1.01,0.0635968029499054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,4,power_law_1.01,0.05140479803085327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,4,power_law_1.01,0.07178239822387696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,4,power_law_1.01,0.05333120226860046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,4,balanced,0.651690681775411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,4,power_law_1.01,0.08346239924430847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,4,power_law_1.01,0.05736960172653198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,4,power_law_1.01,0.0921343982219696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,4,power_law_1.01,0.0723904013633728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,4,power_law_1.01,0.12803200483322144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,4,power_law_1.01,0.08754559755325317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,4,power_law_1.01,0.1529728055000305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,4,balanced,1.2375306288401287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,4,power_law_1.01,0.10051840543746948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,4,power_law_1.01,0.19306880235671997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,4,power_law_1.01,0.13742079734802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,4,power_law_1.01,0.2635135889053345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,4,power_law_1.01,0.1698815941810608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,4,power_law_1.01,0.2723583936691284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,4,power_law_1.01,0.38046081066131593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,4,power_law_1.01,0.33692800998687744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,4,power_law_1.01,0.4229568004608154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,4,power_law_1.01,0.4372096061706543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,4,power_law_1.01,0.5003327846527099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,4,power_law_1.01,0.5780223846435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,4,power_law_1.01,0.8200192451477051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,4,power_law_1.01,0.8307519912719726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,4,power_law_1.01,1.9650304794311524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,4,power_law_1.01,1.08789119720459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,4,power_law_1.01,2.2593088150024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,4,power_law_1.01,0.03978239893913269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,4,power_law_1.01,0.039603200554847715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,4,power_law_1.01,0.03873920142650604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,4,power_law_1.01,0.03868159949779511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,4,power_law_1.01,0.017555199563503265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,4,power_law_1.01,0.040038400888442995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,4,power_law_1.01,0.017664000391960144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,4,power_law_1.01,0.040908798575401306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,4,power_law_1.01,0.017190399765968322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,4,power_law_1.01,0.040191999077796935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,4,power_law_1.01,0.017574399709701538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,4,power_law_1.01,0.04068480134010315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,4,power_law_1.01,0.017420800030231477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,4,power_law_1.01,0.04154239892959595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,4,power_law_1.01,0.018438400328159334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,4,power_law_1.01,0.04066559970378876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,4,power_law_1.01,0.021433599293231964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,4,power_law_1.01,0.040966400504112245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,4,power_law_1.01,0.018432000279426576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,4,power_law_1.01,0.04298880100250244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,4,power_law_1.01,0.021408000588417055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,4,power_law_1.01,0.04254719913005829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,4,power_law_1.01,0.02222079932689667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,4,power_law_1.01,0.04455040097236633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,4,power_law_1.01,0.022303999960422517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,4,power_law_1.01,0.05077120065689087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,4,power_law_1.01,0.02282879948616028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,4,power_law_1.01,0.0502016007900238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,4,power_law_1.01,0.023552000522613525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,4,power_law_1.01,0.05356159806251526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,4,power_law_1.01,0.03108479976654053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,4,power_law_1.01,0.061791998147964475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,4,power_law_1.01,0.03353599905967712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.01,0.06815360188484192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,4,power_law_1.01,0.03260799944400787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.01,0.08293120265007019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,4,power_law_1.01,0.03281280100345611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.01,0.09639679789543151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,4,power_law_1.01,0.03720960021018982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.01,0.1248960018157959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.01,0.04043520092964172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.01,0.13834880590438842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.01,0.05326719880104065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.01,0.2091264009475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.01,0.05920000076293945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.01,0.2628351926803589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.01,0.07904639840126038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.01,0.41886081695556643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.01,0.09559040069580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.01,0.5347008228302002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.01,0.1250048041343689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,4,power_law_1.01,0.7321599960327149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.01,0.15916160345077515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,4,power_law_1.01,1.2447744369506837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.01,0.22176640033721923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,4,power_law_1.01,2.133907127380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.01,0.28601601123809817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,4,power_law_1.01,0.350163197517395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,4,power_law_1.01,0.5424448013305664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,4,power_law_1.01,1.056281566619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,4,power_law_1.2,0.0326335996389389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,4,power_law_1.2,0.030399999022483824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,4,power_law_1.2,0.02968960106372833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,4,power_law_1.2,0.03127039968967438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,4,power_law_1.2,0.03222399950027466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,4,power_law_1.2,0.03287039995193482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,4,power_law_1.2,0.03277440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,4,power_law_1.2,0.03407999873161316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,4,power_law_1.2,0.03578239977359772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,4,power_law_1.2,0.02959359884262085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,4,power_law_1.2,0.03454720079898834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,4,power_law_1.2,0.029414400458335876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,4,power_law_1.2,0.034841600060462954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,4,power_law_1.2,0.029631999135017396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,4,power_law_1.2,0.03621760010719299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,4,power_law_1.2,0.030508801341056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,4,power_law_1.2,0.03739520013332367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,4,power_law_1.2,0.038335999846458434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,4,power_law_1.2,0.031052801012992858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,4,power_law_1.2,0.038335999846458434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,4,power_law_1.2,0.043647998571395875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,4,power_law_1.2,0.03201279938220978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,4,power_law_1.2,0.03821440041065216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,4,power_law_1.2,0.050886398553848265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,4,power_law_1.2,0.031999999284744264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,4,power_law_1.2,0.03969280123710632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,4,power_law_1.2,0.05069440007209778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,4,power_law_1.2,0.03359360098838806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,4,power_law_1.2,0.040633600950241086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,4,power_law_1.2,0.05234559774398804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,4,power_law_1.2,0.03424000144004822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,4,power_law_1.2,0.04145280122756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,4,power_law_1.2,0.06240000128746033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,4,power_law_1.2,0.035097599029541016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,4,power_law_1.2,0.040601599216461184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,4,power_law_1.2,0.017228800058364867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,4,power_law_1.2,0.0762943983078003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,4,power_law_1.2,0.035776001214981076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,4,power_law_1.2,0.04041599929332733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,4,power_law_1.2,0.01744000017642975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,4,power_law_1.2,0.037088000774383546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,4,power_law_1.2,0.08989440202713013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,4,power_law_1.2,0.04128639996051788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,4,power_law_1.2,0.01704320013523102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,4,power_law_1.2,0.03935360014438629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,4,power_law_1.2,0.09813119769096375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,4,power_law_1.2,0.040556800365447995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,4,power_law_1.2,0.018067200481891633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,4,power_law_1.2,0.04081279933452606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,4,power_law_1.2,0.139027202129364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,4,power_law_1.2,0.0177279993891716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,4,power_law_1.2,0.04133760035037994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,4,power_law_1.2,0.04897280037403107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,4,power_law_1.2,0.16673280000686647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,4,power_law_1.2,0.018240000307559966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,4,power_law_1.2,0.04256640076637268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,4,power_law_1.2,0.051737600564956666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,4,power_law_1.2,0.2005631923675537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,4,power_law_1.2,0.021273599565029146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,4,power_law_1.2,0.04535680115222931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,4,power_law_1.2,0.051596802473068235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,4,power_law_1.2,0.2958656072616577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,4,power_law_1.2,0.018323199450969697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,4,power_law_1.2,0.04438399970531463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,4,power_law_1.2,0.06213759779930115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,4,power_law_1.2,0.4626880168914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,4,power_law_1.2,0.0500927984714508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,4,power_law_1.2,0.021164800226688384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,4,power_law_1.2,0.06370559930801392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,4,power_law_1.2,0.5281216144561768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,4,power_law_1.2,0.05229439735412598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,4,power_law_1.2,0.022361600399017335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,4,power_law_1.2,0.08643199801445008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,4,power_law_1.2,0.7056704044342041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,4,power_law_1.2,0.05275520086288452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,4,power_law_1.2,0.02184319943189621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,4,power_law_1.2,0.0983295977115631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,4,power_law_1.2,0.8799103736877442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,4,power_law_1.2,0.02308479994535446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,4,power_law_1.2,0.06096000075340271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,4,power_law_1.2,0.1359935998916626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,4,power_law_1.2,1.379321575164795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,4,power_law_1.2,0.02308479994535446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.2,0.0732479989528656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,4,power_law_1.2,0.18965760469436646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.2,0.08511360287666321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,4,power_law_1.2,0.03023360073566437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,4,power_law_1.2,0.24658560752868652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.2,0.11391359567642212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,4,power_law_1.2,0.03213439881801605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.2,0.12666879892349242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,4,power_law_1.2,0.31090559959411623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,4,power_law_1.2,0.03219200074672699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.2,0.15037440061569213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,4,power_law_1.2,0.49271678924560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,4,power_law_1.2,0.03292160034179688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.2,0.2593408107757568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,4,power_law_1.2,0.5567103862762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,4,power_law_1.2,0.038822400569915774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.2,0.30707199573516847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,4,power_law_1.2,0.7234687805175781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.2,0.040275201201438904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.2,0.4401984214782715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,4,power_law_1.2,1.2687104225158692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.2,0.05389440059661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.2,0.539353609085083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,4,power_law_1.2,2.7275583267211916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,4,power_law_1.2,0.6897471904754638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.2,0.059628802537918094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,4,power_law_1.2,1.0990336418151856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.2,0.07955200076103211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,4,power_law_1.2,2.432934379577637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.2,0.0943231999874115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.2,0.1252671957015991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.2,0.15756160020828247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.2,0.22101120948791503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.2,0.2844032049179077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,4,power_law_1.2,0.3490880012512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,4,power_law_1.2,0.5414207935333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,4,power_law_1.2,1.0553855895996094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,2,power_law_1.01,0.029183998703956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,2,power_law_1.01,0.028620800375938414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,2,power_law_1.01,0.028934401273727418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,2,power_law_1.01,0.03227519989013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,2,power_law_1.01,0.028601598739624024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,2,power_law_1.01,0.03198080062866211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,2,power_law_1.01,0.028703999519348145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,2,power_law_1.01,0.032678401470184325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,2,power_law_1.01,0.028435200452804565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,2,power_law_1.01,0.033267199993133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,2,power_law_1.01,0.031052801012992858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,2,power_law_1.01,0.03463039994239807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,2,power_law_1.01,0.0315200001001358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,2,power_law_1.01,0.03591040074825287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,2,power_law_1.01,0.03184640109539032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,2,power_law_1.01,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,2,power_law_1.01,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,2,power_law_1.01,0.035123199224472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,2,power_law_1.01,0.03443840146064758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,2,power_law_1.01,0.03760640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,2,power_law_1.01,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,2,power_law_1.01,0.03916159868240356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,2,power_law_1.01,0.034822401404380796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,2,power_law_1.01,0.03906559944152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,2,power_law_1.01,0.03513599932193756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,2,power_law_1.01,0.044531199336051944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,2,power_law_1.01,0.03660799860954285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,2,balanced,0.031370667119820915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,2,power_law_1.01,0.04797439873218536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,2,power_law_1.01,0.037836799025535585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,2,balanced,0.03140799949566523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,2,power_law_1.01,0.05048959851264954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,2,power_law_1.01,0.040294399857521056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,2,balanced,0.03364799916744232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,2,power_law_1.01,0.0599232017993927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,2,power_law_1.01,0.04597119987010956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,2,balanced,0.03386666625738144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,2,power_law_1.01,0.07103360295295716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,2,power_law_1.01,0.04878720045089722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,2,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,2,balanced,0.031386665999889374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,2,power_law_1.01,0.09004160165786743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,2,power_law_1.01,0.05023999810218811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,2,balanced,0.03339199970165888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,2,balanced,0.032485333581765495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,2,power_law_1.01,0.09674879908561707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,2,power_law_1.01,0.05878400206565857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,2,balanced,0.03382399926582972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,2,power_law_1.01,0.12554880380630493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,2,balanced,0.03330666571855545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,2,power_law_1.01,0.06604160070419311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,2,power_law_1.01,0.16339839696884156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,2,balanced,0.03374933451414108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,2,balanced,0.03356266766786575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,2,power_law_1.01,0.07944959998130799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,2,power_law_1.01,0.24712960720062255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,2,balanced,0.033770665526390076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,2,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,2,power_law_1.01,0.10163840055465698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,2,power_law_1.01,0.2205183982849121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,2,balanced,0.035717333356539406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,2,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,2,power_law_1.01,0.14351999759674072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,2,power_law_1.01,0.3573440074920654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,2,balanced,0.035429333647092186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,2,balanced,0.03416533271471659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,2,power_law_1.01,0.16008960008621215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,2,power_law_1.01,0.4496640205383301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,2,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,2,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,2,power_law_1.01,0.24763519763946534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,2,power_law_1.01,0.5901440143585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,2,balanced,0.0358240008354187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,2,balanced,0.035775999228159584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,2,power_law_1.01,0.28173439502716063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,2,power_law_1.01,0.857203197479248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,2,balanced,0.03579200059175491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,2,balanced,0.03547733277082443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,2,power_law_1.01,0.41971840858459475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,2,power_law_1.01,1.672831916809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,2,balanced,0.044266665975252785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,2,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,2,power_law_1.01,0.6197375774383544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,2,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,2,balanced,0.03834133346875509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,2,power_law_1.01,0.7266240119934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,2,balanced,0.04393066465854645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,2,balanced,0.03739733248949051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,2,power_law_1.01,1.1192831993103027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,2,balanced,0.05212800204753876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,2,balanced,0.03692266593376795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,2,power_law_1.01,2.3338623046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,2,balanced,0.053823997577031456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,2,balanced,0.044768000642458596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,2,balanced,0.07221333185831706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,2,balanced,0.04568533102671305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,2,balanced,0.07442133128643036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,2,balanced,0.04710400104522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,2,balanced,0.10703466335932414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,2,balanced,0.0521066685517629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,2,balanced,0.12145599722862244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,2,balanced,0.059392000238100685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,2,balanced,0.16547200083732605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,2,balanced,0.07354666789372762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,2,balanced,0.17880533138910928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,2,balanced,0.08166400094827016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,2,balanced,0.24940800666809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,2,balanced,0.12170666456222534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,2,balanced,0.14261333147684732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,2,balanced,0.32758933305740356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,2,balanced,0.19735999902089438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,2,balanced,0.3930720090866089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,2,balanced,0.2400373419125875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,2,balanced,0.34201598167419434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,2,balanced,0.5875786542892456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,2,balanced,0.45629334449768066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,2,balanced,0.5612746477127075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,2,balanced,1.103834629058838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,2,balanced,0.8520320256551107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,2,balanced,1.653765360514323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,2,power_law_1.2,0.02881920039653778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,2,power_law_1.2,0.02871040105819702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,2,power_law_1.2,0.027846398949623107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,2,power_law_1.2,0.02849920094013214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,2,power_law_1.2,0.028601598739624024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,2,power_law_1.2,0.028454399108886717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,2,power_law_1.2,0.03144319951534271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,2,power_law_1.2,0.030502399802207945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,2,power_law_1.2,0.031788799166679385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,2,power_law_1.2,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,2,power_law_1.2,0.032646399736404416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,2,power_law_1.2,0.032332798838615416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,2,power_law_1.2,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,2,power_law_1.2,0.03361920118331909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,2,power_law_1.2,0.03387520015239716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,2,power_law_1.2,0.033657601475715636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,2,power_law_1.2,0.03562879860401154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,2,power_law_1.2,0.03387520015239716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,2,power_law_1.2,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,2,power_law_1.2,0.03511039912700653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,2,power_law_1.2,0.036006399989128114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,2,power_law_1.2,0.03511680066585541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,2,power_law_1.2,0.038047999143600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,2,power_law_1.2,0.03664000034332275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,2,power_law_1.2,0.03746559917926788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,2,power_law_1.2,0.03811840116977692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,2,power_law_1.2,0.038675200939178464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,2,power_law_1.2,0.040627199411392215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,2,power_law_1.2,0.042745599150657655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,2,power_law_1.2,0.045638400316238406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,2,power_law_1.2,0.04952960014343262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,2,power_law_1.2,0.049753600358963014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,2,power_law_1.2,0.05006080269813538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,2,power_law_1.2,0.048819199204444885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,2,power_law_1.2,0.060499197244644164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,2,power_law_1.2,0.0585919976234436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,2,power_law_1.2,0.06869120001792908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,2,power_law_1.2,0.06812800168991089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,2,power_law_1.2,0.0875711977481842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,2,power_law_1.2,0.0882752001285553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,2,power_law_1.2,0.09750400185585022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,2,power_law_1.2,0.11423360109329224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,2,power_law_1.2,0.11720319986343383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,2,power_law_1.2,0.1595136046409607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,2,power_law_1.2,0.16500480175018312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,2,power_law_1.2,0.1774783968925476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,2,power_law_1.2,0.23009281158447265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,2,power_law_1.2,0.2521023988723755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,2,power_law_1.2,0.24384000301361083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,2,power_law_1.2,0.3307136058807373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,2,power_law_1.2,0.3646784067153931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,2,power_law_1.2,0.4300992012023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,2,power_law_1.2,0.43077759742736815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,2,power_law_1.2,0.6146175861358643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,2,power_law_1.2,0.6215167999267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,2,power_law_1.2,0.6987071990966797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,2,power_law_1.2,0.8934207916259765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,2,power_law_1.2,1.2110207557678223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,2,power_law_1.2,1.8006719589233398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,2,power_law_1.2,2.4868160247802735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,8,power_law_1.01,0.043398401141166686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,8,power_law_1.01,0.04261119961738587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,8,power_law_1.01,0.03418880105018616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,8,power_law_1.01,0.03425279855728149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,8,power_law_1.01,0.03175680041313171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,8,balanced,0.03557866563399633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,8,power_law_1.01,0.03451519906520843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,8,power_law_1.01,0.031711998581886294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,8,balanced,0.03379733363787333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,8,power_law_1.01,0.03518719971179962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,8,power_law_1.01,0.03113600015640259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,8,balanced,0.04378133515516917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,8,balanced,0.03369600077470144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,8,power_law_1.01,0.037452799081802365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,8,power_law_1.01,0.03274880051612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,8,balanced,0.03522133330504099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,8,balanced,0.04684799909591675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,8,power_law_1.01,0.038438400626182555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,8,power_law_1.01,0.03344640135765076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,8,balanced,0.03522133330504099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,8,balanced,0.03739733248949051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,8,power_law_1.01,0.04450559914112091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,8,power_law_1.01,0.03482879996299744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,8,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,8,balanced,0.03583466758330663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,8,power_law_1.01,0.04529919922351837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,8,balanced,0.03583466758330663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,8,balanced,0.03576533248027166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,8,power_law_1.01,0.03670400083065033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,8,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,8,power_law_1.01,0.04619520008563995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,8,balanced,0.03573333223660787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,8,balanced,0.020821332931518555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,8,balanced,0.037605332831541695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,8,power_law_1.01,0.03468160033226013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,8,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,8,balanced,0.03777066618204117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,8,power_law_1.01,0.048044800758361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,8,balanced,0.037605332831541695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,8,balanced,0.021498667697111767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,8,balanced,0.03549333413441976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,8,power_law_1.01,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,8,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,8,power_law_1.01,0.05260159969329834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,8,balanced,0.03563733398914337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,8,balanced,0.021312000850836437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,8,balanced,0.03742400060097376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,8,power_law_1.01,0.03736959993839264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,8,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,8,power_law_1.01,0.05665919780731201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,8,power_law_1.01,0.04002560079097748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,8,balanced,0.037615999579429626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,8,balanced,0.03781333317359289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,8,balanced,0.021231998999913532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,8,balanced,0.0358240008354187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,8,power_law_1.01,0.06573439836502075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,8,power_law_1.01,0.04320639967918396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,8,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,8,balanced,0.03790933390458425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,8,balanced,0.021370666722456615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,8,balanced,0.03762666632731756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,8,power_law_1.01,0.06865280270576476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,8,power_law_1.01,0.04549759924411774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,8,balanced,0.039664000272750854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,8,balanced,0.03961600114901861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,8,balanced,0.021615999440352123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,8,balanced,0.037808001041412354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,8,power_law_1.01,0.07486079931259156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,8,power_law_1.01,0.045484799146652224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,8,balanced,0.0436106671889623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,8,balanced,0.03775466730197271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,8,balanced,0.022448000808556873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,8,balanced,0.037903999288876854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,8,power_law_1.01,0.08672639727592468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,8,power_law_1.01,0.05681920051574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,8,balanced,0.04543466866016388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,8,balanced,0.03809066613515218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,8,power_law_1.01,0.0590399980545044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,8,balanced,0.02531733363866806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,8,balanced,0.03779733429352442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,8,power_law_1.01,0.1034559965133667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,8,balanced,0.04739200075467428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,8,balanced,0.0395359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,8,power_law_1.01,0.062483197450637816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,8,balanced,0.027386667827765148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,8,balanced,0.04153066625197729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,8,power_law_1.01,0.07733759880065919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,8,power_law_1.01,0.13525760173797607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,8,balanced,0.047797332207361855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,8,balanced,0.05221333106358846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,8,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,8,balanced,0.02595199892918269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,8,power_law_1.01,0.08700159788131714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,8,power_law_1.01,0.14533120393753052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,8,balanced,0.04872000217437744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,8,balanced,0.057616000374158226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,8,balanced,0.039642666776975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,8,balanced,0.035631999373435974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,8,power_law_1.01,0.11261440515518188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,8,power_law_1.01,0.20293118953704833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,8,balanced,0.05268799761931101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,8,balanced,0.07301333546638489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,8,balanced,0.041850666205088295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,8,balanced,0.03994133323431015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,8,power_law_1.01,0.13823360204696655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,8,power_law_1.01,0.2259648084640503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,8,balanced,0.060645331939061485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,8,balanced,0.09006399909655254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,8,balanced,0.03988266736268997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,8,balanced,0.055946667989095054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,8,balanced,0.04030400017897288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,8,power_law_1.01,0.18330880403518676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,8,power_law_1.01,0.3381632089614868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,8,balanced,0.04613333443800608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,8,balanced,0.11769599715868632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,8,balanced,0.0377866675456365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,8,balanced,0.07421866556008656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,8,power_law_1.01,0.23116800785064698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,8,power_law_1.01,0.38075520992279055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,8,balanced,0.045754666129748024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,8,balanced,0.14202133814493814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,8,balanced,0.04456533491611481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,8,balanced,0.07642666498819987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,8,power_law_1.01,0.31074559688568115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,8,power_law_1.01,0.5223936080932617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,8,balanced,0.0467199981212616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,8,balanced,0.1872160037358602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,8,balanced,0.039861333866914116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,8,balanced,0.09610666831334432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,8,power_law_1.01,0.414521598815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,8,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,8,power_law_1.01,0.7342527866363525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,8,balanced,0.04985600213209788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,8,balanced,0.23505600293477377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,8,balanced,0.10711466272672017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,8,power_law_1.01,0.5802495956420899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,8,balanced,0.056032001972198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,8,power_law_1.01,0.8692928314208984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,8,balanced,0.05585066477457682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,8,balanced,0.3232799967130025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,8,balanced,0.06065066655476888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,8,balanced,0.13538133104642233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,8,power_law_1.01,0.7477695941925049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,8,balanced,0.4169066747029622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,8,balanced,0.060549333691596985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,8,power_law_1.01,0.9310272216796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,8,power_law_1.01,1.4342656135559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,8,balanced,0.07168533404668172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,8,balanced,0.16450666387875876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,8,balanced,0.08285866677761078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,8,balanced,0.5054719845453898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,8,power_law_1.01,1.4791232109069825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,8,power_law_1.01,2.9190143585205077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,8,balanced,0.09486400087674458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,8,balanced,0.21604265769322714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,8,balanced,0.1034346620241801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,8,power_law_1.01,2.8563711166381838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,8,balanced,0.103301336367925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,8,balanced,0.7619199752807617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,8,balanced,0.27768532435099286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,8,balanced,0.13781332969665527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,8,balanced,0.1309760014216105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,8,balanced,0.32947733004887897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,8,balanced,0.16718933979670206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,8,balanced,0.15583999951680502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,8,balanced,1.462821324666341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,8,balanced,0.2385866641998291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,8,balanced,0.20813866456349692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,8,balanced,0.4663306474685669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,8,balanced,0.3025066653887431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,8,balanced,0.26266666253407794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,8,balanced,0.44006399313608807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,8,balanced,0.3147520025571187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,8,balanced,0.8659786383310953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,8,balanced,0.5761919816335043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,8,balanced,0.46988264719645184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,8,balanced,0.7102773189544678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,8,balanced,1.113871971766154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,8,balanced,0.8796586990356445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,8,balanced,2.194821357727051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,8,power_law_1.01,0.019788800179958342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,8,power_law_1.01,0.03656960129737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,8,power_law_1.01,0.02019840031862259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,8,power_law_1.01,0.036083200573921205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,8,power_law_1.01,0.020006400346755982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,8,power_law_1.01,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,8,power_law_1.01,0.01953279972076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,8,power_law_1.01,0.03589119911193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,8,power_law_1.01,0.020268799364566804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,8,power_law_1.01,0.036550399661064145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,8,power_law_1.01,0.02192640006542206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,8,power_law_1.01,0.037145599722862244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,8,power_law_1.01,0.045798400044441225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,8,power_law_1.01,0.03764480054378509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,8,power_law_1.01,0.045721599459648134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,8,power_law_1.01,0.03806720077991486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,8,power_law_1.01,0.03671680092811584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,8,power_law_1.01,0.03871360123157501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,8,power_law_1.01,0.03357439935207367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,8,power_law_1.01,0.03985919952392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,8,power_law_1.01,0.034508800506591795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,8,power_law_1.01,0.03991039991378784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,8,power_law_1.01,0.04063999950885773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,8,power_law_1.01,0.04152320027351379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,8,power_law_1.01,0.03992320001125336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,8,power_law_1.01,0.04302720129489899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,8,power_law_1.01,0.036032000184059144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,8,power_law_1.01,0.045363199710845944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,8,power_law_1.01,0.04297600090503693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,8,power_law_1.01,0.05242239832878113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,8,power_law_1.01,0.03804160058498383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,8,power_law_1.01,0.05335680246353149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,8,power_law_1.01,0.0419840008020401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,8,power_law_1.01,0.0591871976852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,8,power_law_1.01,0.0542464017868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,8,power_law_1.01,0.06996480226516724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.01,0.0588096022605896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.01,0.0688704013824463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.01,0.08165760040283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.01,0.08623999953269959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.01,0.10085760354995728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.01,0.11275520324707031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.01,0.13595520257949828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.01,0.1499392032623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.01,0.1644863963127136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.01,0.16542719602584838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.01,0.23618559837341307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.01,0.2339008092880249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.01,0.300217604637146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.01,0.3092992067337036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.01,0.4367551803588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.01,0.42554240226745604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.01,0.5728320121765137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.01,0.5491903781890869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,8,power_law_1.01,0.7058239936828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,8,power_law_1.01,0.6890944004058838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,8,power_law_1.01,1.1101247787475585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,8,power_law_1.01,1.0389887809753418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,8,power_law_1.01,2.1872703552246096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,8,power_law_1.01,1.9921279907226563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,8,8,power_law_1.2,0.03159680068492889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,8,8,power_law_1.2,0.04348160028457641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,8,8,power_law_1.2,0.03447040021419525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,8,8,power_law_1.2,0.019993600249290467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,8,8,power_law_1.2,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,8,8,power_law_1.2,0.04294399917125702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,8,8,power_law_1.2,0.03583360016345978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,8,8,power_law_1.2,0.01964160054922104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,8,8,power_law_1.2,0.030668801069259642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,8,8,power_law_1.2,0.03457280099391937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,8,8,power_law_1.2,0.034560000896453856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,8,8,power_law_1.2,0.019596800208091736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,8,8,power_law_1.2,0.03573119938373566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,8,8,power_law_1.2,0.03260799944400787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,8,8,power_law_1.2,0.019046400487422944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,8,8,power_law_1.2,0.03431040048599243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,8,8,power_law_1.2,0.036620798707008365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,8,8,power_law_1.2,0.033548799157142636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,8,8,power_law_1.2,0.01998720020055771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,8,8,power_law_1.2,0.03421440124511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,8,8,power_law_1.2,0.03696640133857727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,8,8,power_law_1.2,0.0345984011888504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,8,8,power_law_1.2,0.022303999960422517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,8,8,power_law_1.2,0.034918400645256045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,8,8,power_law_1.2,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,8,8,power_law_1.2,0.036313599348068236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,8,8,power_law_1.2,0.046316799521446225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,8,8,power_law_1.2,0.03669120073318481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,8,8,power_law_1.2,0.03818880021572113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,8,8,power_law_1.2,0.0348800003528595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,8,8,power_law_1.2,0.046086400747299194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,8,8,power_law_1.2,0.0384768009185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,8,8,power_law_1.2,0.037913599610328676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,8,8,power_law_1.2,0.03622399866580963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,8,8,power_law_1.2,0.035488000512123107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,8,8,power_law_1.2,0.04178560078144074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,8,8,power_law_1.2,0.03825919926166534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,8,8,power_law_1.2,0.039577600359916684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,8,8,power_law_1.2,0.03402239978313446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,8,8,power_law_1.2,0.03924480080604553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,8,8,power_law_1.2,0.044921600818634035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,8,8,power_law_1.2,0.04141440093517303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,8,8,power_law_1.2,0.04009599983692169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,8,8,power_law_1.2,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,8,8,power_law_1.2,0.046316799521446225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,8,8,power_law_1.2,0.041484799981117246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,8,8,power_law_1.2,0.0429504007101059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,8,8,power_law_1.2,0.04157440066337585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,8,8,power_law_1.2,0.05144960284233093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,8,8,power_law_1.2,0.045132800936698914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,8,8,power_law_1.2,0.04354560077190399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,8,8,power_law_1.2,0.04080640077590943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,8,8,power_law_1.2,0.053376001119613645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,8,8,power_law_1.2,0.04575999975204468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,8,8,power_law_1.2,0.053881597518920896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,8,8,power_law_1.2,0.03670400083065033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,8,8,power_law_1.2,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,8,8,power_law_1.2,0.05262719988822937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,8,8,power_law_1.2,0.05534719824790955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,8,8,power_law_1.2,0.04195840060710907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,8,8,power_law_1.2,0.06690559983253479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,4,balanced,0.021216000119845074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,8,8,power_law_1.2,0.05860480070114136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,8,8,power_law_1.2,0.06147199869155884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,8,8,power_law_1.2,0.038924801349639895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,8,8,power_law_1.2,0.07134720087051391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,4,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,8,8,power_law_1.2,0.042444801330566405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,4,balanced,0.021087999145189922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,8,8,power_law_1.2,0.06279039978981019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,8,8,power_law_1.2,0.07063680291175842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,8,8,power_law_1.2,0.07563520073890687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,4,balanced,0.03773866593837738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,8,8,power_law_1.2,0.05470079779624939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,4,balanced,0.019920000185569126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,8,8,power_law_1.2,0.07978879809379577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.2,0.07333760261535645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,4,balanced,0.02109866589307785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,8,8,power_law_1.2,0.09580159783363343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,4,balanced,0.03736533224582672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.2,0.05947520136833191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,8,8,power_law_1.2,0.08748160004615783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.2,0.09282559752464295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,4,balanced,0.021087999145189922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,8,8,power_law_1.2,0.10010240077972413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,4,balanced,0.03638399889071783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.2,0.0825984001159668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,8,8,power_law_1.2,0.11300480365753174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.2,0.11768319606781005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,4,balanced,0.021087999145189922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,8,8,power_law_1.2,0.14092799425125122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,4,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.2,0.10116480588912964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,8,8,power_law_1.2,0.1379647970199585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.2,0.14703359603881835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,4,balanced,0.021045332153638203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,8,8,power_law_1.2,0.1515455961227417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,4,balanced,0.03775999943415324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.2,0.13749760389328003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,8,8,power_law_1.2,0.1870527982711792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.2,0.1807487964630127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,4,balanced,0.023120000958442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,8,8,power_law_1.2,0.19938559532165528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,4,balanced,0.039477333426475525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.2,0.16482559442520142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,8,8,power_law_1.2,0.23742079734802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.2,0.2462143898010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,4,balanced,0.02342933416366577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,8,8,power_law_1.2,0.22943360805511476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,4,balanced,0.037802666425704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.2,0.23732481002807618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,8,8,power_law_1.2,0.3317440032958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,4,balanced,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.2,0.297216010093689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,4,balanced,0.02347733328739802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,8,8,power_law_1.2,0.35207679271698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,4,balanced,0.03773866593837738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.2,0.30046720504760743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,8,8,power_law_1.2,0.41390719413757326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,4,balanced,0.0460746685663859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,4,balanced,0.032032000521818794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.2,0.4458752155303955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,4,balanced,0.02716800073782603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,4,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,8,8,power_law_1.2,0.4036543846130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.2,0.4365695953369141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,8,8,power_law_1.2,0.5998655796051026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,4,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,4,balanced,0.03346133232116699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.2,0.546566390991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,4,balanced,0.027232001225153606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,4,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,8,8,power_law_1.2,0.5971007823944092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.2,0.5733952045440673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,8,8,power_law_1.2,0.7723264217376709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,4,balanced,0.03575466573238373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,4,balanced,0.03410666684309641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,8,8,power_law_1.2,0.7038656234741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,4,balanced,0.027232001225153606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,4,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,8,8,power_law_1.2,0.7569664001464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,8,8,power_law_1.2,0.7064767837524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,8,8,power_law_1.2,0.9548095703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,4,balanced,0.03736000011364619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,4,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,8,8,power_law_1.2,1.0682687759399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,4,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,4,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,8,8,power_law_1.2,0.9597760200500488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,8,8,power_law_1.2,1.1117568016052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,8,8,power_law_1.2,1.4788352012634278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,4,balanced,0.03621333340803782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,4,balanced,0.0354720006386439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,8,8,power_law_1.2,1.8752639770507813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,4,balanced,0.03957333415746689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,4,balanced,0.04114133367935816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,8,8,power_law_1.2,1.4334848403930665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,8,8,power_law_1.2,2.1889728546142577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,4,balanced,0.049685334165891014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,8,8,power_law_1.2,2.8376575469970704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,4,balanced,0.03731200098991394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,4,balanced,0.035018667578697205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,4,balanced,0.03789866715669632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,8,8,power_law_1.2,2.663148880004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,4,balanced,0.048021331429481506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,4,balanced,0.03620799879233042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,4,balanced,0.03700266778469086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,4,balanced,0.04163199911514918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,4,balanced,0.048810665806134544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,4,balanced,0.035690667728583016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,4,balanced,0.037589333951473236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,4,balanced,0.04430399835109711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,4,balanced,0.05388799806435903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,4,balanced,0.037791999677817024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,4,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,4,balanced,0.04762666424115499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,4,balanced,0.05819199979305267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,4,balanced,0.0377813329299291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,4,balanced,0.03770666569471359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,4,balanced,0.06815466781457265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,4,balanced,0.06437866886456807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,4,balanced,0.0378560001651446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,4,balanced,0.03919466584920883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,4,balanced,0.07685333490371704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,4,balanced,0.07727466523647308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,4,balanced,0.037690666814645134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,4,balanced,0.039893334110577904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,4,balanced,0.09898133079210918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,4,balanced,0.10213866829872131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,4,balanced,0.039919999738534294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,4,balanced,0.039818666875362396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,4,balanced,0.12108266353607178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,4,balanced,0.12769066294034323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,4,balanced,0.039642666776975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,4,balanced,0.048170665899912514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,4,balanced,0.15915733575820923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,4,balanced,0.1739199956258138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,4,balanced,0.049642667174339294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,4,balanced,0.04993600149949392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,4,balanced,0.18706132968266806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,4,balanced,0.21176000436147055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,4,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,4,balanced,0.04878933231035868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,4,balanced,0.25626667340596515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,4,balanced,0.05824000140031179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,4,balanced,0.31040000915527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,4,balanced,0.04980266590913137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,4,balanced,0.3262186646461487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,4,balanced,0.06413333117961884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,4,balanced,0.06028800209363302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,4,balanced,0.3965119918187459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,4,balanced,0.08665600419044495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,4,balanced,0.0668639987707138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,4,balanced,0.3940639893213908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,4,balanced,0.09822932879130046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,4,balanced,0.0765119989713033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,4,balanced,0.4926453431447347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,4,balanced,0.1299626628557841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,4,balanced,0.08126933375994365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,4,balanced,0.5979413191477457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,4,balanced,0.16237866878509521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,4,balanced,0.7621653079986572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,4,balanced,0.1074666678905487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,4,balanced,0.22288533051808676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,4,balanced,0.12613866726557413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,4,balanced,0.26823999484380084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,4,balanced,1.1378986835479736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,4,balanced,0.158869336048762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,4,balanced,1.4858187039693196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,4,balanced,0.372538685798645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,4,balanced,0.18926399946212769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,4,balanced,0.47951467831929523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,4,balanced,0.2538986603418986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,4,balanced,0.5838826497395834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,4,balanced,0.3223946690559387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,4,balanced,0.3863786856333415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,4,balanced,0.893882671991984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,4,balanced,0.6137386560440063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,4,balanced,1.7244960467020671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,4,balanced,1.1198453108469646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,4,power_law_1.01,0.04206080138683319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,4,power_law_1.01,0.018803200125694274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,4,power_law_1.01,0.04296320080757141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,4,power_law_1.01,0.04591360092163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,4,power_law_1.01,0.03203200101852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,4,power_law_1.01,0.019091199338436126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,4,power_law_1.01,0.04328320026397705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,4,power_law_1.01,0.03434880077838898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,4,power_law_1.01,0.031916800141334536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,4,power_law_1.2,0.0321152001619339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,4,power_law_1.01,0.018988800048828126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,4,power_law_1.01,0.04648320078849792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,4,power_law_1.2,0.050335997343063356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,4,power_law_1.01,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,4,power_law_1.01,0.030623999238014222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,4,power_law_1.2,0.0310591995716095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,4,power_law_1.01,0.01870719939470291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,4,power_law_1.01,0.04595839977264404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,4,power_law_1.2,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,4,power_law_1.01,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,4,power_law_1.01,0.03322240114212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,4,power_law_1.2,0.04259839951992035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,4,power_law_1.2,0.03086720108985901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,4,power_law_1.01,0.019257600605487823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,4,power_law_1.01,0.047116801142692566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,4,power_law_1.2,0.033030399680137636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,4,power_law_1.01,0.03480960130691528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,4,power_law_1.01,0.03335680067539215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,4,power_law_1.2,0.044531199336051944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,4,power_law_1.2,0.03296639919281006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,4,power_law_1.01,0.02025599926710129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,4,power_law_1.2,0.018681600689888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,4,power_law_1.01,0.04575999975204468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,4,power_law_1.2,0.03488639891147614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,4,power_law_1.01,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,4,power_law_1.01,0.03449600040912628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,4,power_law_1.2,0.04357759952545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,4,power_law_1.2,0.03333120048046112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,4,power_law_1.01,0.036211198568344115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,4,power_law_1.01,0.020902399718761445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,4,power_law_1.2,0.018457600474357606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,4,power_law_1.01,0.04535680115222931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,4,power_law_1.2,0.03495680093765259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,4,power_law_1.01,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,4,power_law_1.2,0.04630399942398071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,4,power_law_1.2,0.03431040048599243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,4,power_law_1.01,0.03640320003032684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,4,power_law_1.01,0.02134400010108948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,4,power_law_1.2,0.019097599387168884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,4,power_law_1.01,0.045772799849510194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,4,power_law_1.2,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,4,power_law_1.01,0.03711999952793121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,4,power_law_1.2,0.046988800168037415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,4,power_law_1.2,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,4,power_law_1.01,0.03728640079498291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,4,power_law_1.01,0.02223999947309494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,4,power_law_1.2,0.018764799833297728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,4,power_law_1.01,0.045151999592781066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,4,power_law_1.2,0.03607679903507233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,4,power_law_1.01,0.03708159923553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,4,power_law_1.2,0.04580479860305786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,4,power_law_1.2,0.035820800065994265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,4,power_law_1.01,0.03850879967212677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,4,power_law_1.01,0.022707200050354003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,4,power_law_1.2,0.018783999979496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,4,power_law_1.01,0.04615040123462677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,4,power_law_1.2,0.03694080114364624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,4,power_law_1.01,0.04051840007305145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,4,power_law_1.2,0.04539520144462585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,4,power_law_1.2,0.03678080141544342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,4,power_law_1.01,0.038624000549316403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,4,power_law_1.01,0.025625601410865784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,4,power_law_1.2,0.019705599546432494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,4,power_law_1.01,0.048044800758361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,4,power_law_1.2,0.040966400504112245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,4,power_law_1.01,0.04229759871959686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,4,power_law_1.2,0.045849600434303285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,4,power_law_1.2,0.03932160139083862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,4,power_law_1.01,0.039212799072265624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,4,power_law_1.01,0.025113600492477416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,4,power_law_1.2,0.020608000457286835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,4,power_law_1.01,0.04948480129241943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,4,power_law_1.2,0.04077439904212952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,4,power_law_1.01,0.04542079865932465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,4,power_law_1.2,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,4,power_law_1.2,0.039155200123786926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,4,power_law_1.01,0.042438399791717527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,4,power_law_1.01,0.02620159983634949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,4,power_law_1.2,0.021465599536895752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,4,power_law_1.01,0.050374400615692136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,4,power_law_1.2,0.043084800243377686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,4,power_law_1.01,0.04575360119342804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,4,power_law_1.2,0.04097920060157776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,4,power_law_1.2,0.04559360146522522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,4,power_law_1.01,0.0430976003408432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,4,power_law_1.01,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,4,power_law_1.01,0.0531711995601654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,4,power_law_1.2,0.022259199619293214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,4,power_law_1.01,0.054745602607727054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,4,power_law_1.2,0.04676479995250702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,4,power_law_1.01,0.053881597518920896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,4,power_law_1.2,0.04221439957618713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,4,power_law_1.2,0.04576640129089356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,4,power_law_1.01,0.03706879913806915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,4,power_law_1.01,0.05589759945869446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,4,power_law_1.2,0.02240640074014664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,4,power_law_1.01,0.055795198678970336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,4,power_law_1.2,0.04759039878845215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,4,power_law_1.01,0.06042879819869995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,4,power_law_1.2,0.0477183997631073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,4,power_law_1.2,0.048588800430297854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,4,power_law_1.01,0.03716480135917664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,4,power_law_1.01,0.06195840239524841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,4,power_law_1.2,0.025222399830818178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,4,power_law_1.01,0.0641152024269104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,4,power_law_1.2,0.052102398872375486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,4,power_law_1.01,0.06214399933815003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,4,power_law_1.2,0.0542527973651886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,4,power_law_1.2,0.04927360117435455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,4,power_law_1.01,0.04043520092964172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,4,power_law_1.2,0.05864959955215454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,4,power_law_1.01,0.07464960217475891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,4,power_law_1.2,0.025356799364089966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,4,power_law_1.01,0.07274240255355835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,4,power_law_1.2,0.026015999913215637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,4,power_law_1.2,0.05982080101966858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,4,power_law_1.01,0.06414080262184144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,4,power_law_1.2,0.04864639937877655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,4,power_law_1.01,0.04354560077190399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,4,power_law_1.2,0.060031998157501223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,4,power_law_1.01,0.09039999842643738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.01,0.04580479860305786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.01,0.0856768012046814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,4,power_law_1.2,0.03600000143051148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,4,power_law_1.2,0.06398720145225525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,4,power_law_1.01,0.07509120106697083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,4,power_law_1.2,0.05607680082321167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,4,power_law_1.2,0.06741120219230652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,4,power_law_1.01,0.10786559581756591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.01,0.06612480282783509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.01,0.10173439979553223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,4,power_law_1.2,0.03711999952793121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,4,power_law_1.2,0.06895359754562377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,4,power_law_1.01,0.09221119880676269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,4,power_law_1.2,0.05817599892616272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,4,power_law_1.2,0.08761600255966187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,4,power_law_1.01,0.13246079683303832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.01,0.07467520236968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.01,0.1230463981628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,4,power_law_1.2,0.037171199917793274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,4,power_law_1.2,0.08187519907951354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,4,power_law_1.01,0.10820480585098266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,4,power_law_1.2,0.05932160019874573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,4,power_law_1.2,0.12283519506454468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,4,power_law_1.01,0.18168959617614747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.01,0.09931520223617554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,4,power_law_1.2,0.040966400504112245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.01,0.15166079998016357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,4,power_law_1.2,0.09393919706344604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,4,power_law_1.01,0.1218559980392456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,4,power_law_1.2,0.07640960216522216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,4,power_law_1.2,0.1349184036254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.2,0.08294399976730346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,4,power_law_1.01,0.2317375898361206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.01,0.124236798286438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,4,power_law_1.2,0.04317440092563629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.01,0.20730879306793212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,4,power_law_1.2,0.10768640041351318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,4,power_law_1.01,0.17389440536499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,4,power_law_1.2,0.19319679737091064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.2,0.10012799501419067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,4,power_law_1.01,0.3285183906555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.01,0.17192959785461426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.2,0.04699519872665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.01,0.2731519937515259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,4,power_law_1.2,0.12187520265579224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,4,power_law_1.01,0.20407679080963134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,4,power_law_1.2,0.260153603553772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.2,0.12358399629592895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,4,power_law_1.01,0.45281281471252444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.01,0.21057920455932616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.2,0.0661184012889862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.01,0.37575039863586424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,4,power_law_1.2,0.18977919816970826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,4,power_law_1.01,0.2877248048782349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,4,power_law_1.2,0.36883840560913084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.2,0.15394560098648072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,4,power_law_1.01,0.6486527919769287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.01,0.30624639987945557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.2,0.07467520236968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.01,0.4440256118774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.2,0.09966719746589661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,4,power_law_1.2,0.23315200805664063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,4,power_law_1.01,0.3580607891082764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,4,power_law_1.2,0.4218751907348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.2,0.23375999927520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,4,power_law_1.01,0.8066816329956055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.01,0.39334399700164796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.01,0.5302591800689698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.2,0.1238592028617859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,4,power_law_1.2,0.3691391944885254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,4,power_law_1.01,0.5255231857299805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,4,power_law_1.2,0.6943103790283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.2,0.2738687992095947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,4,power_law_1.01,1.0445055961608887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,4,power_law_1.01,0.48737277984619143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,4,power_law_1.01,0.8770688056945801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.2,0.17136000394821166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,4,power_law_1.2,0.3927743911743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,4,power_law_1.01,0.7910463809967041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,4,power_law_1.2,0.9568703651428223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.2,0.32984960079193115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,4,power_law_1.01,1.5303423881530762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,4,power_law_1.01,0.759116792678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,4,power_law_1.01,1.3847871780395509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.2,0.20929279327392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,4,power_law_1.2,0.5486911773681641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.2,0.3105792045593262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,4,power_law_1.01,0.722105598449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,4,power_law_1.2,1.0307071685791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.2,0.550867223739624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,4,power_law_1.01,2.9281408309936525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,4,power_law_1.01,1.4851967811584472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.2,0.639302396774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,4,power_law_1.2,1.395359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,4,power_law_1.01,2.3159679412841796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,4,power_law_1.2,0.7693439960479737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.2,0.3919487953186035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,4,power_law_1.01,1.5108863830566406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,4,power_law_1.2,0.8952896118164062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,4,power_law_1.2,3.4563838958740236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,4,power_law_1.2,0.876921558380127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,4,power_law_1.2,0.4879487991333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,4,power_law_1.01,3.032582473754883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,4,power_law_1.2,1.3122624397277831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,4,power_law_1.2,1.5081472396850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,4,power_law_1.2,0.7586239814758301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,4,power_law_1.2,2.922073554992676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,4,power_law_1.2,3.4220417022705076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,4,power_law_1.2,1.4839232444763184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,2,balanced,0.018965333700180054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,2,power_law_1.01,0.018956799805164338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,2,balanced,0.018917333334684372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,2,power_law_1.01,0.01820160001516342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,2,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,2,balanced,0.018981333822011948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,2,power_law_1.01,0.03669120073318481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,2,power_law_1.01,0.018080000579357148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,2,balanced,0.037733333806196846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,2,balanced,0.019248000035683315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,2,power_law_1.01,0.03705599904060364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,2,power_law_1.01,0.018412800133228303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,2,balanced,0.039994666973749794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,2,power_law_1.01,0.036697599291801455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,2,balanced,0.01915733392039935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,2,power_law_1.01,0.01849599927663803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,2,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,2,power_law_1.01,0.03777279853820801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,2,balanced,0.01937066639463107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,2,power_law_1.01,0.0191103994846344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,2,balanced,0.03984000037113825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,2,power_law_1.01,0.03834240138530731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,2,balanced,0.02094399929046631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,2,power_law_1.01,0.01968639940023422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,2,balanced,0.03982399900754293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,2,power_law_1.01,0.03960959911346436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,2,balanced,0.021541332205136616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,2,power_law_1.01,0.020710399746894835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,2,balanced,0.04014399896065394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,2,power_law_1.01,0.039110401272773744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,2,balanced,0.023152001202106476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,2,power_law_1.01,0.02077440023422241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,2,balanced,0.04195199906826019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,2,power_law_1.01,0.04016000032424927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,2,balanced,0.023434666295846302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,2,power_law_1.01,0.021478399634361267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,2,balanced,0.042037333051363625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,2,power_law_1.01,0.04025599956512451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,2,balanced,0.02476799984773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,2,power_law_1.01,0.02279680073261261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,2,balanced,0.041936000188191734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,2,power_law_1.01,0.04080640077590943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,2,balanced,0.025199999411900837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,2,power_law_1.01,0.023180800676345825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,2,balanced,0.04178666571776072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,2,power_law_1.01,0.04192639887332916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,2,balanced,0.025290665527184803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,2,power_law_1.01,0.024160000681877136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,2,power_law_1.01,0.04476799964904785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,2,balanced,0.043968002001444496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,2,balanced,0.027530667682488758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,2,power_law_1.01,0.026719999313354493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,2,power_law_1.01,0.04478079974651337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,2,balanced,0.04388799766699473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,2,balanced,0.029359998802344005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,2,power_law_1.01,0.02686080038547516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,2,power_law_1.01,0.04521600008010864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,2,balanced,0.04590400060017904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,2,balanced,0.029504001140594482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,2,power_law_1.01,0.029228800535202028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,2,power_law_1.01,0.031353598833084105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,2,power_law_1.01,0.05011199712753296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,2,power_law_1.01,0.03364480137825012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,2,balanced,0.03355200091997782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,2,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,2,balanced,0.041519999504089355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,2,balanced,0.05028266708056132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,2,power_law_1.01,0.04019840061664581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,2,power_law_1.01,0.03189760148525238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,2,power_law_1.01,0.05438079833984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,2,power_law_1.01,0.03418239951133728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,2,balanced,0.03444266567627589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,2,power_law_1.01,0.03182719945907593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,2,power_law_1.01,0.056403201818466184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,2,balanced,0.033887999753157295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,2,balanced,0.04586133360862732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,2,balanced,0.05028266708056132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,2,power_law_1.01,0.04318720102310181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,2,power_law_1.01,0.03415040075778961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,2,balanced,0.03376533339420954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,2,power_law_1.01,0.03284479975700379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,2,balanced,0.035829332967599235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,2,power_law_1.01,0.06839039921760559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,2,balanced,0.035904000202814736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,2,balanced,0.04756799836953481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,2,balanced,0.053120002150535583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.01,0.04557439982891083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,2,power_law_1.01,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,2,power_law_1.01,0.034272000193595886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,2,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.01,0.07304319739341736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,2,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,2,balanced,0.054661333560943604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,2,balanced,0.06039466460545858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.01,0.05237759947776795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,2,power_law_1.01,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,2,power_law_1.01,0.03489919900894165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,2,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.01,0.09949439764022827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,2,balanced,0.03549866626660029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,2,balanced,0.06346666812896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,2,balanced,0.03576533248027166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,2,balanced,0.06537599861621857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.01,0.061286401748657224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,2,power_law_1.01,0.036211198568344115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,2,power_law_1.01,0.035846400260925296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.01,0.11076480150222778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,2,balanced,0.03612266729275385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,2,balanced,0.08731200297673543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,2,balanced,0.037946666280428566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,2,balanced,0.07576533158620198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.01,0.08480640053749085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,2,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,2,power_law_1.01,0.036595198512077334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,2,power_law_1.01,0.036339199542999266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.01,0.14237439632415771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,2,balanced,0.037733333806196846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,2,balanced,0.10200533270835876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,2,balanced,0.09337600072224934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.01,0.0990015983581543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,2,balanced,0.03754133234421412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,2,power_law_1.01,0.037747201323509214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.01,0.1341055989265442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,2,power_law_1.01,0.03685120046138764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.01,0.18366719484329225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,2,balanced,0.03787733366092046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,2,balanced,0.1373973290125529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,2,balanced,0.11770666639010112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,2,balanced,0.039861333866914116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,2,power_law_1.01,0.037555199861526486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.01,0.17093759775161743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,2,power_law_1.01,0.038489601016044615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.01,0.24595201015472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,2,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,2,balanced,0.17362133661905924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,2,balanced,0.03993066648642222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,2,balanced,0.14296533664067587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,2,power_law_1.01,0.03973119854927063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.01,0.24262399673461915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,2,power_law_1.01,0.038675200939178464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.01,0.28437759876251223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,2,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,2,balanced,0.039861333866914116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,2,balanced,0.24632533391316733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,2,balanced,0.19183999300003052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,2,balanced,0.041802664597829185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,2,power_law_1.01,0.042854401469230655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.01,0.30449280738830564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,2,power_law_1.01,0.039468801021575926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.01,0.3975359916687012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,2,balanced,0.03969600051641464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,2,balanced,0.23587733507156372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,2,balanced,0.3079040050506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,2,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,2,power_law_1.01,0.044198399782180785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,2,power_law_1.01,0.3778559923171997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,2,power_law_1.01,0.04030719995498657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.01,0.5985983848571778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,2,balanced,0.03994133323431015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,2,balanced,0.052149335543314614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,2,balanced,0.33007999261220294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,2,power_law_1.01,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,2,balanced,0.3799360195795695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,2,power_law_1.01,0.043424001336097716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,2,power_law_1.01,0.5850304126739502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,2,power_law_1.01,0.6551424026489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,2,balanced,0.040565334260463715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,2,balanced,0.0521919975678126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,2,power_law_1.01,0.04898560047149658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,2,power_law_1.01,1.014521598815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,2,balanced,0.4291520118713379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,2,balanced,0.062122667829195656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,2,power_law_1.01,0.052697598934173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,2,power_law_1.01,1.1414976119995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,2,balanced,0.587285319964091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,2,balanced,0.04177600145339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,2,power_law_1.01,0.057171201705932616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,2,power_law_1.01,2.1121343612670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,2,balanced,0.07435733576615651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,2,power_law_1.01,0.05386880040168762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,2,balanced,0.5262613296508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,2,balanced,0.052784000833829246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,2,power_law_1.01,0.061990398168563846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,2,power_law_1.01,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,2,balanced,0.09658666451772054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,2,balanced,0.05588266750176748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,2,balanced,1.1443573633829753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,2,power_law_1.01,0.06558079719543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,2,balanced,0.8096799850463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,2,power_law_1.01,0.07698559761047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,2,balanced,0.054197331269582115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,2,balanced,0.11362133423487346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,2,power_law_1.01,0.08258559703826904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,2,power_law_1.01,0.08647680282592773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,2,balanced,0.06675200164318085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,2,balanced,0.14945066968599954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,2,power_law_1.01,0.08368639945983887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,2,balanced,1.5692532857259114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,2,power_law_1.01,0.10922240018844605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,2,balanced,0.07494399944941203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,2,balanced,0.18116267522176108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,2,power_law_1.01,0.10250240564346313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,2,power_law_1.01,0.1317055940628052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,2,balanced,0.08628267049789429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,2,balanced,0.25459200143814087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,2,power_law_1.01,0.1281599998474121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,2,power_law_1.01,0.1643903970718384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,2,balanced,0.09912533561388652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,2,balanced,0.3144373297691345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,2,power_law_1.01,0.16952320337295532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,2,power_law_1.01,0.22417919635772704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,2,balanced,0.13167466719945273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,2,balanced,0.45451200008392334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,2,power_law_1.01,0.1941248059272766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,2,power_law_1.01,0.3177216053009033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,2,balanced,0.14667733510335287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,2,power_law_1.01,0.2517184019088745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,2,power_law_1.01,0.40159358978271487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,2,balanced,0.5909706751505533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,2,power_law_1.01,0.36631040573120116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,2,balanced,0.19508800903956094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,2,power_law_1.01,0.5309760093688964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,2,power_law_1.01,0.4615039825439453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,2,power_law_1.01,0.7983424186706543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,2,balanced,0.24567999442418417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,2,power_law_1.01,0.5905920028686523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,2,balanced,0.7290453116099039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,2,power_law_1.01,0.9115008354187012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,2,power_law_1.01,0.7227263927459717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,2,balanced,0.33655468622843426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,2,power_law_1.01,1.406886386871338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,2,power_law_1.01,1.2485759735107422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,2,power_law_1.01,2.9280256271362304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,2,balanced,0.4339199860890706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,2,power_law_1.01,2.369740867614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,2,balanced,1.121824026107788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,2,balanced,0.5272426605224609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,2,balanced,0.7848426500956217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,2,balanced,2.2013920148213706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,2,balanced,1.5172746976216633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,2,power_law_1.2,0.03025279939174652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,2,power_law_1.2,0.03343999981880188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,2,power_law_1.2,0.030924800038337707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,2,power_law_1.2,0.03038719892501831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,2,power_law_1.2,0.03251200020313263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,2,power_law_1.2,0.03383040130138397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,2,power_law_1.2,0.032678401470184325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,2,power_law_1.2,0.03953279852867127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,2,power_law_1.2,0.033792001008987424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,2,power_law_1.2,0.034745600819587705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,2,power_law_1.2,0.04181120097637177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,2,power_law_1.2,0.03547520041465759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,2,power_law_1.2,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,2,power_law_1.2,0.04220159947872162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,2,power_law_1.2,0.036575999855995175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,2,power_law_1.2,0.03610239923000336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,2,power_law_1.2,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,2,power_law_1.2,0.03723520040512085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,2,power_law_1.2,0.036800000071525577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,2,power_law_1.2,0.04700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,2,power_law_1.2,0.03821440041065216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,2,power_law_1.2,0.0377023994922638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,2,power_law_1.2,0.04694400131702423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,2,power_law_1.2,0.038726401329040525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,2,power_law_1.2,0.03886080086231232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,2,power_law_1.2,0.04723199903964996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,2,power_law_1.2,0.0398719996213913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,2,power_law_1.2,0.040729600191116336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,2,power_law_1.2,0.046412798762321475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,2,power_law_1.2,0.018265600502490997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,2,power_law_1.2,0.04123519957065582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,2,power_law_1.2,0.04247680008411407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,2,power_law_1.2,0.046988800168037415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,2,power_law_1.2,0.018515199422836304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,2,power_law_1.2,0.04113920032978058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,2,power_law_1.2,0.04485760033130646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,2,power_law_1.2,0.047065600752830505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,2,power_law_1.2,0.018688000738620758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,2,power_law_1.2,0.044435200095176694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,2,power_law_1.2,0.04661119878292084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,2,power_law_1.2,0.04699519872665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,2,power_law_1.2,0.018432000279426576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,2,power_law_1.2,0.050783997774124144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,2,power_law_1.2,0.04795520007610321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,2,power_law_1.2,0.04862079918384552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,2,power_law_1.2,0.055225598812103274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,2,power_law_1.2,0.018220800161361694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,2,power_law_1.2,0.05862399935722351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,2,power_law_1.2,0.049983999133110045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,2,power_law_1.2,0.05626239776611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,2,power_law_1.2,0.01916159987449646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,2,power_law_1.2,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,2,power_law_1.2,0.050547200441360476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,2,power_law_1.2,0.07745280265808105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,2,power_law_1.2,0.020652799308300017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,2,power_law_1.2,0.06287360191345215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,2,power_law_1.2,0.05607039928436279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,2,power_law_1.2,0.08327040076255798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,2,power_law_1.2,0.020479999482631683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,2,power_law_1.2,0.07816960215568543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,2,power_law_1.2,0.05753600001335144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,2,power_law_1.2,0.12056959867477417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,2,power_law_1.2,0.021414400637149812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,2,power_law_1.2,0.08843520283699036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,2,power_law_1.2,0.061452800035476686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,2,power_law_1.2,0.13231359720230101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,2,power_law_1.2,0.021087999641895293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,2,power_law_1.2,0.10305279493331909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,2,power_law_1.2,0.07162240147590637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,2,power_law_1.2,0.1679487943649292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,2,power_law_1.2,0.023097600042819976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,2,power_law_1.2,0.13248000144958497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.2,0.08147199749946595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,2,power_law_1.2,0.23559041023254396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,2,power_law_1.2,0.023379200696945192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,2,power_law_1.2,0.1642367959022522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.2,0.10276479721069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,2,power_law_1.2,0.3244096040725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,2,power_law_1.2,0.0236735999584198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,2,power_law_1.2,0.20142080783843994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.2,0.12383359670639038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,2,power_law_1.2,0.3675008058547974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,2,power_law_1.2,0.02645759880542755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,2,power_law_1.2,0.28387839794158937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.2,0.16289279460906983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,2,power_law_1.2,0.6311808109283448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,2,power_law_1.2,0.02728320062160492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,2,power_law_1.2,0.3105664014816284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.2,0.17345919609069824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,2,power_law_1.2,0.8071295738220214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,2,power_law_1.2,0.02895359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,2,power_law_1.2,0.4669951915740967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.2,0.27134079933166505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,2,power_law_1.2,0.9678272247314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,2,power_law_1.2,0.040300801396369934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,2,power_law_1.2,0.6869823932647705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,64,balanced,0.07834133505821228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,64,balanced,0.06229333579540253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.2,0.3448319911956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,2,power_law_1.2,1.5371583938598632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,2,power_law_1.2,0.04270080029964447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,2,power_law_1.2,0.7634175777435303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,64,balanced,0.06193066636721293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,64,balanced,0.07893333335717519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,64,balanced,0.06387199958165486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.2,0.4746240139007568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,2,power_law_1.2,3.1732288360595704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,64,balanced,0.03588266670703888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.2,0.0455487996339798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,2,power_law_1.2,1.3941311836242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,64,balanced,0.07852800190448761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,64,balanced,0.06285866598288219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,64,balanced,0.06246933341026306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.2,0.677235221862793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,64,balanced,0.02959999938805898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.2,0.053235197067260744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,2,power_law_1.2,2.3089664459228514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,64,balanced,0.08078399797280629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,64,balanced,0.06272533535957336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,64,balanced,0.06530666848023732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,2,power_law_1.2,0.9219776153564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,64,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,64,balanced,0.07780799766381581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,64,balanced,0.029264000554879505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.2,0.06155520081520081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,64,balanced,0.09527466694513957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,2,power_law_1.2,1.2131008148193358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,64,balanced,0.07436800003051758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,64,balanced,0.09329600135485332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,64,balanced,0.03161066770553589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.2,0.08542079925537109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,64,balanced,0.1301653285821279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,2,power_law_1.2,2.7845312118530274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,64,balanced,0.09482133388519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,64,balanced,0.09179733196894328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,64,balanced,0.0344106654326121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.2,0.09943040013313294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,64,balanced,0.12038933237393697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,64,balanced,0.09816533327102661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,64,balanced,0.08935466408729553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,64,balanced,0.04205866654713949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.2,0.13568639755249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,64,balanced,0.12736533085505167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,64,balanced,0.10124799609184265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,64,balanced,0.09558399518330891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,64,balanced,0.06755200028419495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.2,0.1720255970954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,64,balanced,0.09756799538930257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,64,balanced,0.14201600352923074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,64,balanced,0.0965119997660319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,64,balanced,0.06790933509667714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,64,balanced,0.09490666786829631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.2,0.2439487934112549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,64,balanced,0.14247999588648477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,64,balanced,0.09513066212336223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,64,balanced,0.07368533313274384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,64,balanced,0.09506133198738098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.2,0.30661759376525877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,64,balanced,0.1283146639664968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,64,balanced,0.0985599954922994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,64,balanced,0.0729973316192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,64,balanced,0.09491200248400371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,2,power_law_1.2,0.3773632049560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,64,balanced,0.12559466560681662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,64,balanced,0.09547199805577596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,64,balanced,0.07122133175532024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,64,balanced,0.09993599851926167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,64,balanced,0.13429866234461466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,2,power_law_1.2,0.5850111961364746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,64,balanced,0.10293866197268169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,64,balanced,0.06495466828346252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,64,balanced,0.09545600414276123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,64,balanced,0.1346773306528727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,2,power_law_1.2,1.1435839653015136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,64,balanced,0.10520000259081523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,64,balanced,0.06115733087062836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,64,balanced,0.10000000397364299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,64,balanced,0.1422879993915558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,64,balanced,0.10214933753013611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,64,balanced,0.05332266787687937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,64,balanced,0.10301333665847778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,64,balanced,0.14012799660364786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,64,balanced,0.11056533455848694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,64,balanced,0.09172800183296204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,64,balanced,0.10782933235168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,64,balanced,0.14177067081133524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,64,balanced,0.11967466274897258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,64,balanced,0.11531200011571248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,64,balanced,0.0879200001557668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,64,balanced,0.14006400108337402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,64,balanced,0.12622400124867758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,64,balanced,0.11658666531244914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,64,balanced,0.0728000005086263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,64,balanced,0.14803199966748556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,64,balanced,0.1439253290494283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,64,balanced,0.12942933042844137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,64,balanced,0.09504000345865886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,64,balanced,0.150325338045756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,64,balanced,0.1611840029557546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,64,balanced,0.1379039982954661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,64,balanced,0.11530666550000508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,64,balanced,0.1590666671593984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,64,balanced,0.19850667317708334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,64,balanced,0.16230400403340658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,64,balanced,0.15681599577267966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,64,balanced,0.18220800161361694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,64,balanced,0.23645333449045816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,64,balanced,0.1876266598701477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,64,balanced,0.18794665733973184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,64,balanced,0.2037386695543925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,64,balanced,0.321615993976593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,64,balanced,0.27033599217732746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,64,balanced,0.25808533032735187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,64,balanced,0.27646400531133014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,64,balanced,0.3474133412043254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,64,balanced,0.38573865095774335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,64,balanced,0.3025066653887431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,64,balanced,0.31489066282908124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,64,balanced,0.5156213442484537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,64,balanced,0.5431093374888102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,64,balanced,0.41473066806793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,64,balanced,0.688917318979899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,64,balanced,0.4246453444163005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,64,balanced,0.6926666895548502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,64,balanced,1.020527998606364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,64,balanced,0.5252000093460083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,64,balanced,0.5356693267822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,64,balanced,0.8474346796671549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,64,balanced,1.348090648651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,64,balanced,0.6419893503189087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,64,balanced,0.6492960055669149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,64,balanced,1.6797706286112468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,64,balanced,1.305408000946045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,64,balanced,0.9837226867675781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,64,balanced,0.9724960327148438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,64,balanced,2.6707146962483725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,64,balanced,2.571392059326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,64,balanced,1.9285279909769695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,64,balanced,1.8914027214050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,64,balanced,5.3048906326293945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,64,power_law_1.01,0.12549760341644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,64,power_law_1.01,0.09329919815063477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,64,power_law_1.01,0.08451200127601624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,64,power_law_1.2,0.12388479709625244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,64,power_law_1.2,0.08455039858818054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,64,power_law_1.01,0.028697600960731505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,64,power_law_1.2,0.09313279986381531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,64,power_law_1.01,0.12256640195846558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,64,power_law_1.2,0.10454399585723877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,64,power_law_1.01,0.10093439817428589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,64,power_law_1.01,0.08584960103034973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,64,power_law_1.2,0.12403199672698975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,64,power_law_1.2,0.08524799942970276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,64,power_law_1.2,0.029139199852943422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,64,power_law_1.01,0.028089600801467895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,64,power_law_1.01,0.10428800582885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,64,power_law_1.2,0.07185919880867005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,64,power_law_1.01,0.07373440265655518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,64,power_law_1.01,0.08084480166435241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,64,power_law_1.2,0.08885759711265565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,64,power_law_1.2,0.07123839855194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,64,power_law_1.01,0.08303999900817871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,64,power_law_1.2,0.028332799673080444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,64,power_law_1.01,0.026547199487686156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,64,power_law_1.01,0.09392639994621277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,64,power_law_1.2,0.08169599771499633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,64,power_law_1.01,0.07907840013504028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,64,power_law_1.2,0.07959679961204529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,64,power_law_1.2,0.10351999998092651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,64,power_law_1.01,0.07713279724121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,64,power_law_1.2,0.024383999407291412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,64,power_law_1.01,0.02800639867782593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,64,power_law_1.01,0.11292799711227416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,64,power_law_1.01,0.08286719918251037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,64,power_law_1.2,0.07724159955978394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,64,power_law_1.2,0.08206719756126404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,64,power_law_1.2,0.10158720016479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,64,power_law_1.01,0.08899199962615967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,64,power_law_1.2,0.02985599935054779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,64,power_law_1.01,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,64,power_law_1.01,0.1173632025718689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,64,power_law_1.01,0.08671360015869141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,64,power_law_1.2,0.08286719918251037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,64,power_law_1.2,0.08231040239334106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,64,power_law_1.2,0.11532800197601319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,64,power_law_1.01,0.08796799778938294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,64,power_law_1.2,0.034143999218940735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,64,power_law_1.01,0.04523519873619079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,64,power_law_1.01,0.12505600452423096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,64,power_law_1.01,0.09185919761657715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,64,power_law_1.01,0.06432639956474304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,64,power_law_1.2,0.09154559969902039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,64,power_law_1.2,0.09413120150566101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,64,power_law_1.2,0.125164794921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,64,power_law_1.01,0.0957759976387024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,64,power_law_1.2,0.043628799915313723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,64,power_law_1.01,0.12785919904708862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,64,power_law_1.01,0.09194239974021912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,64,power_law_1.01,0.06526079773902893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,64,power_law_1.2,0.0924351990222931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,64,power_law_1.2,0.0934656023979187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,64,power_law_1.2,0.12418559789657593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,64,power_law_1.01,0.09066879749298096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,64,power_law_1.2,0.06647040247917176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,64,power_law_1.01,0.12562559843063353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,64,power_law_1.01,0.09040639996528625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,64,power_law_1.01,0.07677440047264099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,64,power_law_1.2,0.08906239867210389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,64,power_law_1.2,0.09120000004768372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,64,power_law_1.2,0.12478079795837402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,64,power_law_1.01,0.09315840005874634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,64,power_law_1.2,0.06664959788322448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,64,power_law_1.01,0.12759679555892944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,64,power_law_1.01,0.09641600251197815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,64,power_law_1.01,0.07700480222702026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,64,power_law_1.2,0.0889855980873108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,64,power_law_1.2,0.09155840277671815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,64,power_law_1.2,0.12259839773178101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,64,power_law_1.01,0.09096959829330445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,64,power_law_1.2,0.07441920042037964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,64,power_law_1.01,0.13004159927368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,64,power_law_1.01,0.0936191976070404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,64,power_law_1.01,0.07676799893379212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,64,power_law_1.2,0.09047039747238159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,64,power_law_1.2,0.09415040016174317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,64,power_law_1.2,0.11665920019149781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,64,power_law_1.01,0.09319679737091065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,64,power_law_1.2,0.07448959946632386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,64,power_law_1.01,0.1273792028427124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,64,power_law_1.01,0.09648000001907349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,64,power_law_1.01,0.07486720085144043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,64,power_law_1.2,0.09371520280838012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,64,power_law_1.2,0.09134079813957215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,64,power_law_1.2,0.122707200050354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,64,power_law_1.01,0.09530879855155945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,64,power_law_1.2,0.07397119998931885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,64,power_law_1.01,0.12798080444335938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,64,power_law_1.01,0.09366400241851806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,64,power_law_1.01,0.07337599992752075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,64,power_law_1.2,0.09041280150413514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,64,power_law_1.2,0.09464960098266602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,64,power_law_1.2,0.12828160524368287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,64,power_law_1.01,0.09589120149612426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,64,power_law_1.2,0.07103360295295716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,64,power_law_1.01,0.12789759635925294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,64,power_law_1.01,0.09674879908561707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,64,power_law_1.01,0.06851199865341187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,64,power_law_1.2,0.09455360174179077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,64,power_law_1.2,0.09718400239944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,64,power_law_1.2,0.12550400495529174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,64,power_law_1.01,0.09883520007133484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,64,power_law_1.2,0.06988160014152527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,64,power_law_1.01,0.1251904010772705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,64,power_law_1.01,0.10208640098571778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,64,power_law_1.01,0.06520959734916687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,64,power_law_1.2,0.1003775954246521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,64,power_law_1.2,0.09778559803962708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,64,power_law_1.2,0.1318400025367737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,64,power_law_1.01,0.10001920461654663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,64,power_law_1.2,0.10347520112991333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,64,power_law_1.2,0.0689791977405548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,64,power_law_1.01,0.1309823989868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,64,power_law_1.2,0.11102080345153809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,64,power_law_1.01,0.10449919700622559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,64,power_law_1.01,0.06154239773750305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,64,power_law_1.2,0.10217599868774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,64,power_law_1.2,0.1289471983909607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,64,power_law_1.01,0.10040960311889649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,64,power_law_1.2,0.06405119895935059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,64,power_law_1.01,0.1298367977142334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,64,power_law_1.2,0.13507839441299438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,64,power_law_1.01,0.10554239749908448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,64,power_law_1.01,0.07367039918899536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,64,power_law_1.2,0.1078976035118103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,64,power_law_1.2,0.1350399971008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,64,power_law_1.01,0.11044479608535766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,64,power_law_1.2,0.06652799844741822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,64,power_law_1.01,0.16061439514160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,64,power_law_1.2,0.1429311990737915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,64,power_law_1.01,0.12579840421676636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,64,power_law_1.01,0.07637760043144226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,64,power_law_1.2,0.1200063943862915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,64,power_law_1.2,0.16487679481506348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.01,0.12640000581741334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,64,power_law_1.2,0.07642880082130432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,64,power_law_1.01,0.16572799682617187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,64,power_law_1.2,0.1789695978164673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,64,power_law_1.01,0.135481595993042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.01,0.08320000171661376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.2,0.1414720058441162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,64,power_law_1.2,0.17683839797973633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.01,0.1487936019897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,64,power_law_1.2,0.07587839961051941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,64,power_law_1.01,0.19570560455322267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,64,power_law_1.2,0.2105151891708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,64,power_law_1.01,0.16531840562820435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.01,0.1017151951789856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.2,0.15767040252685546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,64,power_law_1.2,0.2073215961456299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.01,0.17134720087051392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.2,0.08632959723472595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,64,power_law_1.01,0.2131200075149536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,64,power_law_1.2,0.27457280158996583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,64,power_law_1.01,0.19895679950714112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.01,0.11585279703140258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.2,0.19278719425201415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,64,power_law_1.2,0.25357439517974856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.01,0.21390080451965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.2,0.1086400032043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,64,power_law_1.01,0.2805759906768799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,64,power_law_1.2,0.32969601154327394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,64,power_law_1.01,0.25026559829711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.01,0.15219839811325073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.2,0.23247361183166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,64,power_law_1.2,0.33792641162872317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.01,0.2582848072052002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.2,0.12536319494247436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.2,0.305184006690979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,64,power_law_1.01,0.3533951997756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,64,power_law_1.2,0.4394815921783447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,64,power_law_1.01,0.30716800689697266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.01,0.19407360553741454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,64,power_law_1.2,0.385100793838501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.01,0.3486207962036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.2,0.15882879495620728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.2,0.3894272089004517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,64,power_law_1.01,0.4368447780609131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,64,power_law_1.2,0.6291135787963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,64,power_law_1.01,0.4195072174072266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.01,0.26438400745391843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,64,power_law_1.2,0.5780608177185058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.01,0.42106242179870607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.2,0.22471680641174316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.2,0.49649920463562014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,64,power_law_1.01,0.580083179473877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,64,power_law_1.2,0.9525631904602051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,64,power_law_1.01,0.5461376190185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.01,0.3712127923965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,64,power_law_1.2,0.7231232166290283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.01,0.6189888000488282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.01,0.5512832164764404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.2,0.31872000694274905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.2,0.719814395904541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,64,power_law_1.01,0.7829376220703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,64,power_law_1.2,1.2553536415100097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,64,power_law_1.01,0.791980791091919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,64,power_law_1.2,0.9282431602478027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.01,0.7972864151000977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.01,0.7257215976715088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.2,0.41999359130859376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.2,0.9999679565429688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,64,power_law_1.01,0.9939583778381348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,64,power_law_1.2,1.4041536331176758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,64,power_law_1.01,1.01779203414917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,64,power_law_1.2,1.508364772796631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,64,power_law_1.01,1.0134400367736816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,64,power_law_1.01,0.8800064086914062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.2,0.5935296058654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,64,power_law_1.2,1.1554752349853517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,64,power_law_1.01,1.2299263954162598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,64,power_law_1.2,2.4635711669921876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,64,power_law_1.01,1.2856767654418946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,64,power_law_1.2,1.6053056716918945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,64,power_law_1.2,4.738054275512695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,64,power_law_1.01,1.4582528114318847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,64,power_law_1.01,1.487724781036377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.2,0.8657024383544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,64,power_law_1.2,2.100704002380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,64,power_law_1.01,1.8562559127807616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,64,power_law_1.01,2.001875114440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,64,power_law_1.2,3.1572992324829103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,64,power_law_1.01,3.0521215438842773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,64,power_law_1.01,2.768332862854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,64,power_law_1.2,0.9674752235412598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,64,power_law_1.2,3.965177536010742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,64,power_law_1.01,3.3887870788574217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,64,power_law_1.01,3.737356948852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,64,power_law_1.2,6.021926498413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,64,power_law_1.2,1.7505727767944337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,64,power_law_1.2,3.4585662841796876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,32,power_law_1.01,0.06702719926834107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,32,balanced,0.02533866713444392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,32,balanced,0.051967998345692955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,32,power_law_1.01,0.03589119911193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,32,power_law_1.01,0.06152960062026978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,32,balanced,0.025333332518736523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,32,balanced,0.052943999568621315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,32,power_law_1.01,0.03590399920940399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,32,power_law_1.01,0.05375360250473023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,32,balanced,0.025349333882331848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,32,balanced,0.05208000044027964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,32,power_law_1.01,0.02481279969215393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,32,power_law_1.01,0.05509120225906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,32,balanced,0.02516266703605652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,32,power_law_1.01,0.0234047994017601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,32,balanced,0.054133335749308266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,32,power_law_1.01,0.055961602926254274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,32,balanced,0.033301333586374916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,32,power_law_1.01,0.029631999135017396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,32,balanced,0.056234667698542275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,32,power_law_1.01,0.05820159912109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,32,balanced,0.03942933430274328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,32,power_law_1.01,0.029126399755477907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,32,balanced,0.06381866832574208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,32,power_law_1.01,0.061350399255752565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,32,balanced,0.04353600243727366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,32,power_law_1.01,0.03381119966506958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,32,balanced,0.0644160012404124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,32,power_law_1.01,0.06224640011787415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,32,balanced,0.04561600089073181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,32,power_law_1.01,0.0357120007276535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,32,balanced,0.06635733445485432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,32,power_law_1.01,0.06394240260124207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,32,balanced,0.04966400067011515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,32,power_law_1.01,0.03454720079898834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,32,balanced,0.06598933537801106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,32,power_law_1.01,0.06354560256004334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,32,balanced,0.048709332942962646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,32,power_law_1.01,0.036620798707008365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,32,balanced,0.06534400085608165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,32,power_law_1.01,0.06277120113372803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,32,balanced,0.0545066644748052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,32,power_law_1.01,0.0431551992893219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,32,balanced,0.06850666801134746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,32,power_law_1.01,0.0651968002319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,32,balanced,0.057578667998313904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,32,power_law_1.01,0.047788798809051514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,32,balanced,0.06957866748174031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,32,power_law_1.01,0.0662335991859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,32,balanced,0.05579733351866404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,32,power_law_1.01,0.0505407989025116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,32,balanced,0.07260266443093617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,32,power_law_1.01,0.06988800168037415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,32,balanced,0.0922879974047343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,32,power_law_1.01,0.06663680076599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,32,balanced,0.07460266848405202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,32,power_law_1.01,0.07711359858512878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,32,balanced,0.09479467074076335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,32,balanced,0.08007466793060303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,32,power_law_1.01,0.07312639951705932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,32,power_law_1.01,0.0804095983505249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,32,balanced,0.09511466821034749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,32,balanced,0.08015466729799907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,32,power_law_1.01,0.08121600151062011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,32,power_law_1.01,0.089055997133255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,32,balanced,0.08468799789746602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,32,balanced,0.12651733557383218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,32,power_law_1.01,0.09357439875602722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,32,power_law_1.01,0.10010240077972413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,32,balanced,0.09325866897900899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,32,balanced,0.16098666191101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,32,power_law_1.01,0.10024319887161255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.01,0.10857599973678589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.01,0.07631999850273133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,32,balanced,0.09750399986902873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,32,balanced,0.1745013395945231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,32,balanced,0.047882666190465294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,32,power_law_1.01,0.0713919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.01,0.1271615982055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.01,0.08841599822044373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,32,balanced,0.11356266339619954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,32,balanced,0.060405333836873375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,32,balanced,0.2467306653658549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,32,balanced,0.045978665351867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,32,power_law_1.01,0.07631360292434693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,32,balanced,0.15470932920773825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.01,0.1410752058029175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.01,0.10102399587631225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,32,power_law_1.01,0.10280959606170655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,32,balanced,0.1271519958972931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,32,balanced,0.04910400013128916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,32,balanced,0.047728002071380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,32,power_law_1.01,0.055641597509384154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.01,0.1746559977531433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,32,balanced,0.20561067263285318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.01,0.1313088059425354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,32,power_law_1.01,0.09613440036773682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,32,balanced,0.14826132853825888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,32,balanced,0.04978133241335551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,32,balanced,0.048058668772379555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,32,power_law_1.01,0.05632640123367309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.01,0.20821120738983154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.01,0.15790079832077025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,32,balanced,0.2627466718355815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,32,power_law_1.01,0.0803264021873474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,32,balanced,0.05016533533732096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,32,balanced,0.04990933338801066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,32,power_law_1.01,0.06852480173110961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,32,balanced,0.07231466472148895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,32,balanced,0.1765600045522054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,32,power_law_1.01,0.055846399068832396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.01,0.2917759895324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.01,0.22800641059875487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,32,balanced,0.3606719970703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,32,balanced,0.07566399872303009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,32,power_law_1.01,0.07927680015563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,32,balanced,0.09878399968147278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,32,power_law_1.01,0.05909759998321533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,32,balanced,0.2241493264834086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.01,0.3425600051879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.01,0.2814143896102905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,32,balanced,0.07578133543332417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,32,power_law_1.01,0.08177919983863831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,32,balanced,0.10026666522026062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,32,balanced,0.4574879805246989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,32,power_law_1.01,0.07057279944419861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.01,0.5624832153320313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.01,0.4152063846588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,32,balanced,0.07784000039100647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,32,power_law_1.01,0.09175040125846863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,32,balanced,0.2713013291358948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,32,balanced,0.0993333359559377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,32,power_law_1.01,0.06952319741249084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,32,balanced,0.7038773695627848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.01,0.6488448143005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.01,0.5838272094726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,32,balanced,0.07875733574231465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,32,power_law_1.01,0.0920960009098053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,32,balanced,0.10132799545923869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,32,power_law_1.01,0.07271040081977845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,32,balanced,0.37324265638987225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,32,power_law_1.01,0.8386879920959472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,32,power_law_1.01,0.6661632061004639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,32,balanced,0.07893866797288258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,32,power_law_1.01,0.09504640102386475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,32,balanced,0.9172373612721761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,32,balanced,0.10228266318639119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,32,power_law_1.01,0.07437440156936645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,32,power_law_1.01,1.2455807685852052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,32,power_law_1.01,0.9964672088623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,32,balanced,0.07875200112660725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,32,power_law_1.01,0.0966592013835907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,32,balanced,0.4718506733576457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,32,balanced,0.10199466347694397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,32,power_law_1.01,0.08054400086402894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,32,power_law_1.01,2.52476806640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,32,balanced,1.1417173544565837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,32,power_law_1.01,1.9919872283935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,32,balanced,0.08256533245245616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,32,power_law_1.01,0.09594879746437072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,32,balanced,0.08368000388145447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,32,balanced,0.10308266679445903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,32,power_law_1.01,0.07968639731407165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,32,balanced,0.577018658320109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,32,power_law_1.01,0.0945088028907776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,32,balanced,0.08358400066693623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,32,balanced,0.10325866937637329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,32,power_law_1.01,0.07998080253601074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,32,power_law_1.01,0.09921919703483581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,32,balanced,1.7949546178181965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,32,balanced,0.09099200367927551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,32,balanced,0.10499733686447144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,32,power_law_1.01,0.08412799835205079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,32,power_law_1.01,0.09831039905548096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,32,balanced,0.09103467067082723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,32,balanced,0.11158933242162068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,32,balanced,0.8659786383310953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,32,power_law_1.01,0.09096959829330445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,32,power_law_1.01,0.10821759700775146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,32,balanced,0.10980266332626343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,32,balanced,0.09614933530489604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,32,power_law_1.01,0.09763200283050537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,32,power_law_1.01,0.10558079481124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,32,balanced,0.11310933033625285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,32,balanced,0.10824533303578694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,32,power_law_1.01,0.10315519571304321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,32,power_law_1.01,0.11639679670333862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,32,balanced,0.11985599994659424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,32,balanced,3.563173294067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,32,balanced,0.11353066563606262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,32,power_law_1.01,0.11654399633407593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,32,power_law_1.01,0.13281919956207275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,32,balanced,0.12220266461372375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,32,power_law_1.01,0.13339519500732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,32,balanced,0.13136000434557596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,32,power_law_1.01,0.12202240228652954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,32,balanced,0.13390933473904928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,32,power_law_1.01,0.16619520187377929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,32,balanced,0.14908799529075623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,32,power_law_1.01,0.14874240159988403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,32,balanced,1.6866827011108398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,32,power_law_1.01,0.17455359697341918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,32,balanced,0.1437386671702067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,32,balanced,0.1856586734453837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,32,power_law_1.01,0.16525440216064452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,32,power_law_1.01,0.2230207920074463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,32,balanced,0.16547733545303345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,32,balanced,0.22209600607554117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,32,power_law_1.01,0.2189568042755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,32,power_law_1.01,0.24261119365692138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,32,balanced,0.19339734315872192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,32,power_law_1.01,0.25840001106262206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,32,power_law_1.01,0.3314687967300415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,32,balanced,0.2906773289044698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,32,power_law_1.01,0.35729920864105225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,32,power_law_1.01,0.4132415771484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,32,balanced,0.22898666063944498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,32,power_law_1.01,0.4763455867767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,32,power_law_1.01,0.563705587387085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,32,power_law_1.01,0.6793727874755859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,32,balanced,0.3606826861699422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,32,power_law_1.01,0.6794239997863769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,32,power_law_1.01,0.8449919700622559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,32,balanced,0.2723466753959656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,32,power_law_1.01,0.824505615234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,32,power_law_1.01,1.1024383544921874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,32,power_law_1.01,1.2617535591125488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,32,power_law_1.01,1.57128963470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,32,balanced,0.5023200114568075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,32,power_law_1.01,2.5530752182006835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,32,balanced,0.3657279809315999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,32,power_law_1.01,3.2350078582763673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,32,balanced,0.6466720104217529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,32,balanced,0.4573173522949219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,32,balanced,0.7889760335286459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,32,balanced,0.5517760117848715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,32,balanced,1.1976266702016194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,32,balanced,0.7907040119171143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,32,balanced,2.3328906695048013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,32,balanced,1.5056373278299968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,32,power_law_1.2,0.06803839802742004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,32,power_law_1.2,0.11011840105056762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,32,power_law_1.2,0.07255039811134338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,32,power_law_1.2,0.09208959937095643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,32,power_law_1.2,0.047891199588775635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,32,power_law_1.2,0.055692797899246214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,32,power_law_1.2,0.05246080160140991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,32,power_law_1.2,0.06568959951400757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,32,power_law_1.2,0.058336001634597776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,32,power_law_1.2,0.08140799999237061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,32,power_law_1.2,0.061990398168563846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,32,power_law_1.2,0.08113279938697815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,32,power_law_1.2,0.06759679913520814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,32,power_law_1.2,0.08956159949302674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,32,power_law_1.2,0.07230719923973083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,32,power_law_1.2,0.09653120040893555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,32,power_law_1.2,0.07032960057258605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,32,power_law_1.2,0.09735680222511292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,32,power_law_1.2,0.07070720195770264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,32,power_law_1.2,0.09345279932022095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,32,power_law_1.2,0.07776640057563781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,32,power_law_1.2,0.09826560020446777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,32,power_law_1.2,0.07857279777526856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,32,power_law_1.2,0.09418879747390747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,32,power_law_1.2,0.08336640000343323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,32,power_law_1.2,0.10231679677963257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,32,power_law_1.2,0.08617600202560424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,32,power_law_1.2,0.11164159774780273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,32,power_law_1.2,0.09148160219192505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,32,power_law_1.2,0.1165120005607605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,32,power_law_1.2,0.09716479778289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,32,power_law_1.2,0.1182144045829773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,32,power_law_1.2,0.10317440032958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,32,power_law_1.2,0.11984000205993653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,32,power_law_1.2,0.11759999990463257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,32,power_law_1.2,0.1338688015937805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,32,power_law_1.2,0.1264256000518799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,32,power_law_1.2,0.14764800071716308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,32,power_law_1.2,0.16278400421142578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,32,power_law_1.2,0.06242560148239136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,32,power_law_1.2,0.18823039531707764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,32,power_law_1.2,0.18017280101776123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,32,power_law_1.2,0.05728639960289002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,32,power_law_1.2,0.18649599552154542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,32,power_law_1.2,0.2332672119140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,32,power_law_1.2,0.0515392005443573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,32,power_law_1.2,0.2745471954345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,32,power_law_1.2,0.30665600299835205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,32,power_law_1.2,0.053958398103713986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,32,power_law_1.2,0.3019007921218872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,32,power_law_1.2,0.027238398790359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,32,power_law_1.2,0.4149888038635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,32,power_law_1.2,0.05674880146980286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,32,power_law_1.2,0.3890496015548706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,32,power_law_1.2,0.027577599883079527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,32,power_law_1.2,0.5204864025115967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,32,power_law_1.2,0.05802239775657654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,32,power_law_1.2,0.5202367782592774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,32,power_law_1.2,0.024480000138282776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,32,power_law_1.2,0.06058880090713501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,32,power_law_1.2,0.8161279678344726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,32,power_law_1.2,0.7868671894073487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,32,power_law_1.2,0.0243136003613472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,32,power_law_1.2,0.059564799070358276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,32,power_law_1.2,0.926585578918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,32,power_law_1.2,0.9325568199157714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,32,power_law_1.2,0.02675839960575104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,32,power_law_1.2,0.06269440054893494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,32,power_law_1.2,1.3104576110839843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,32,power_law_1.2,1.1553279876708984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,32,power_law_1.2,0.028748801350593566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,32,power_law_1.2,0.06257920265197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,32,power_law_1.2,2.121811294555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,32,power_law_1.2,1.8002815246582031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,32,power_law_1.2,0.04490239918231964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,32,power_law_1.2,0.06308479905128479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,32,power_law_1.2,3.553472137451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,32,power_law_1.2,0.044486400485038755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,32,power_law_1.2,3.2988609313964843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,32,power_law_1.2,0.06542720198631287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,32,power_law_1.2,0.06242560148239136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,32,power_law_1.2,0.06746879816055298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,32,power_law_1.2,0.06469119787216186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,32,power_law_1.2,0.07219840288162231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,32,power_law_1.2,0.06382719874382019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,32,power_law_1.2,0.07991039752960205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,32,power_law_1.2,0.07253119945526124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,32,power_law_1.2,0.08380159735679626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,32,power_law_1.2,0.07340800166130065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,32,power_law_1.2,0.09039999842643738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,32,power_law_1.2,0.07601280212402343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,32,power_law_1.2,0.0997759997844696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,32,power_law_1.2,0.07018880248069763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.2,0.11271040439605713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,32,power_law_1.2,0.07251840233802795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.2,0.1460736036300659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,32,power_law_1.2,0.07143679857254029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.2,0.15015039443969727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,32,power_law_1.2,0.07431039810180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.2,0.18895360231399536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.2,0.08138239979743958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.2,0.2463167905807495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.2,0.09642879962921143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.2,0.3187391996383667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.2,0.10981119871139526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.2,0.4264383792877197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.2,0.1454848051071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.2,0.5542655944824219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.2,0.16939519643783568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.2,0.8063039779663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.2,0.24338560104370116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,32,power_law_1.2,1.071769618988037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.2,0.3083199977874756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,32,power_law_1.2,1.7130239486694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.2,0.46036481857299805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,32,power_law_1.2,3.268844985961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.2,0.599891185760498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,32,power_law_1.2,0.7150271892547607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,32,power_law_1.2,1.1851967811584472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,32,power_law_1.2,2.477676773071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,16,balanced,0.023381332556406658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,16,balanced,0.023221333821614582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,16,balanced,0.023178666830062866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,16,balanced,0.023797333240509033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,16,balanced,0.027317332724730175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,16,balanced,0.04081599911053976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,16,balanced,0.04527466495831808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,16,balanced,0.053871999184290566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,16,balanced,0.04563199977080027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,16,balanced,0.053957333167394005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,16,balanced,0.04764799773693085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,16,balanced,0.05221333106358846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,16,balanced,0.048954665660858154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,16,balanced,0.05606399973233541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,16,balanced,0.05272000034650167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,16,balanced,0.05798399945100149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,16,balanced,0.053904001911481224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,16,balanced,0.06941333413124084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,16,balanced,0.055589333176612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,16,balanced,0.07155199845631917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,16,balanced,0.060693333546320595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,16,balanced,0.07464000085989635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,16,balanced,0.06233066817124685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,16,balanced,0.07442133128643036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,16,balanced,0.06262399752934773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,16,balanced,0.07266666491826375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,16,balanced,0.0992693305015564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,16,balanced,0.07648000121116638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,16,balanced,0.10844799876213074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,16,balanced,0.07673599819342296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,16,balanced,0.14478400349617004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,16,balanced,0.07849599917729695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,16,balanced,0.1889973282814026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,16,balanced,0.08106666803359985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,16,balanced,0.14427733421325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,16,balanced,0.08732799688975017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,16,balanced,0.08703999718030293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,16,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,16,balanced,0.17987734079360962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,16,balanced,0.045791998505592346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,16,balanced,0.09206933776537578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,16,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,16,balanced,0.04557866851488749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,16,balanced,0.219866673151652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,16,balanced,0.10169066985448201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,16,balanced,0.04776533444722494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,16,balanced,0.04669333497683207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,16,balanced,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,16,balanced,0.10712533195813496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,16,balanced,0.2842026750246684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,16,balanced,0.05183466772238413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,16,balanced,0.05022400120894114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,16,balanced,0.12353600064913432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,16,balanced,0.07580266892910004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,16,balanced,0.07240533332029979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,16,balanced,0.3534719944000244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,16,balanced,0.07851733267307281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,16,balanced,0.13872533043225607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,16,balanced,0.09787733356157939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,16,balanced,0.0784800002972285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,16,balanced,0.09893866380055745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,16,balanced,0.1664906640847524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,16,balanced,0.08055999875068665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,16,balanced,0.09943466385205586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,16,balanced,0.19537067413330078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,16,balanced,0.537066658337911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,16,balanced,0.08251733581225078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,16,balanced,0.10065066814422607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,16,balanced,0.25781333446502686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,16,balanced,0.08116266628106435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,16,balanced,0.10232533017794292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,16,balanced,0.08294933537642162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,16,balanced,0.09983999530474345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,16,balanced,0.0848479966322581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,16,balanced,0.1029919981956482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,16,balanced,0.3163520097732544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,16,balanced,0.6942400137583414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,16,balanced,0.08690133690834045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,16,balanced,0.1030346651871999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,16,balanced,0.09244267145792644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,16,balanced,0.10410666465759277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,16,balanced,0.094842662413915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,16,balanced,0.43518932660420734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,16,balanced,0.10910399754842122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,16,balanced,0.10098133484522502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,16,balanced,0.8820853233337402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,16,balanced,0.11151466766993205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,16,balanced,0.11328533291816711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,16,balanced,0.11187199751536052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,16,balanced,0.1207360029220581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,16,balanced,0.12004799644152324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,16,balanced,0.5541599988937378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,16,balanced,0.14215999841690063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,16,balanced,0.12597333391507468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,16,balanced,0.15991999705632529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,16,balanced,0.13853866855303446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,16,balanced,0.2023680011431376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,16,balanced,0.15568533539772034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,16,balanced,1.3661012649536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,16,balanced,0.24064000447591147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,16,balanced,0.6756532986958822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,16,balanced,0.17970667282740274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,16,balanced,0.31802666187286377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,16,balanced,0.2068586746851603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,16,balanced,0.40087465445200604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,16,balanced,0.2561653256416321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,16,balanced,1.03492267926534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,16,balanced,0.5614933172861735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,16,balanced,0.3153333266576131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,16,balanced,2.6988585789998374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,16,balanced,0.4261813163757324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,16,balanced,0.721285343170166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,16,balanced,0.5251413186391195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,16,balanced,0.8841546376546224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,16,balanced,2.013866742451986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,16,balanced,0.6375679969787598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,16,balanced,1.3487520217895508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,16,balanced,0.9376800060272217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,16,power_law_1.01,0.06616320013999939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,16,power_law_1.01,0.052723199129104614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,16,power_law_1.01,0.09390720129013061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,16,power_law_1.01,0.0716863989830017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,16,power_law_1.01,0.06494719982147217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,16,power_law_1.01,0.04542720019817352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,16,power_law_1.01,0.06499199867248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,16,balanced,2.631882667541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,16,power_law_1.01,0.05079039931297302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,16,power_law_1.01,0.07233279943466187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,16,power_law_1.01,0.054451197385787964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,16,balanced,1.78275728225708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,16,power_law_1.01,0.07674239873886109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,16,power_law_1.01,0.060192000865936277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,16,power_law_1.01,0.08635519742965699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,16,power_law_1.01,0.07036160230636597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,16,power_law_1.01,0.08900480270385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,16,power_law_1.01,0.072326397895813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,16,power_law_1.01,0.09292160272598267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,16,power_law_1.01,0.07344639897346497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,16,power_law_1.01,0.09340159893035889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,16,power_law_1.01,0.0738048017024994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,16,power_law_1.01,0.09597439765930176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,16,power_law_1.01,0.077920001745224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,16,power_law_1.01,0.09890559911727906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,16,power_law_1.01,0.08064000010490417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,16,power_law_1.01,0.1001855969429016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,16,power_law_1.01,0.08057600259780884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,16,power_law_1.01,0.09845759868621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,16,power_law_1.01,0.08543360233306885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,16,power_law_1.01,0.10602879524230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,16,power_law_1.01,0.0957759976387024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,16,power_law_1.01,0.11594879627227783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,16,power_law_1.01,0.09838719964027405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,16,power_law_1.01,0.11534719467163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,16,power_law_1.01,0.1034559965133667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,16,power_law_1.01,0.13000320196151732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,16,power_law_1.01,0.12135679721832275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,16,power_law_1.01,0.13699840307235717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,16,power_law_1.01,0.08220160007476807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,16,power_law_1.01,0.12687360048294066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,16,power_law_1.01,0.16009600162506105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,16,power_law_1.01,0.10557440519332886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,16,power_law_1.01,0.1554751992225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,16,power_law_1.01,0.17505279779434205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,16,power_law_1.01,0.07826560139656066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,16,power_law_1.01,0.17935359477996826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,16,power_law_1.01,0.22519679069519044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,16,power_law_1.01,0.06928640007972717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,16,power_law_1.01,0.24298241138458251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,16,power_law_1.01,0.2702143907546997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,16,power_law_1.01,0.02374400049448013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,16,power_law_1.01,0.07724159955978394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,16,power_law_1.01,0.2837951898574829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,16,power_law_1.01,0.3614464044570923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,16,power_law_1.01,0.030220800638198854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,16,power_law_1.01,0.09080320000648498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,16,power_law_1.01,0.393068790435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,16,power_law_1.01,0.41385598182678224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,16,power_law_1.01,0.022188800573349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,16,power_law_1.01,0.09475200176239014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,16,power_law_1.01,0.4753471851348877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,16,power_law_1.01,0.6033728122711182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,16,power_law_1.01,0.024031999707221984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,16,power_law_1.01,0.09678080081939697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,16,power_law_1.01,0.6919104099273682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,16,power_law_1.01,0.7668992042541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,16,power_law_1.01,0.025260800123214723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,16,power_law_1.01,0.09473919868469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,16,power_law_1.01,0.9121664047241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,16,power_law_1.01,0.9020159721374512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,16,power_law_1.01,0.02815360128879547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,16,power_law_1.01,0.09945600032806397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,16,power_law_1.01,1.1255935668945312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,16,power_law_1.01,1.3593664169311523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,16,power_law_1.01,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,16,power_law_1.01,0.09992960095405579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,16,power_law_1.01,1.679136085510254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,16,power_law_1.01,0.033267199993133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,16,power_law_1.01,2.8271360397338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,16,power_law_1.01,0.09835519790649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,16,power_law_1.01,0.03377920091152191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,16,power_law_1.01,3.2539520263671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,16,power_law_1.01,0.09694719910621644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,16,power_law_1.01,0.040217599272727965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,16,power_law_1.01,0.10157439708709717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,16,power_law_1.01,0.03899520039558411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,16,power_law_1.01,0.10606720447540283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,16,power_law_1.01,0.03824639916419983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,16,power_law_1.01,0.1055232048034668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,16,power_law_1.01,0.04524160027503967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,16,power_law_1.01,0.1109760046005249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,16,power_law_1.01,0.05309439897537231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,16,power_law_1.01,0.12205439805984497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,16,power_law_1.01,0.05676159858703613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.01,0.13097599744796753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,16,power_law_1.01,0.05994880199432373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.01,0.1523519992828369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,16,power_law_1.01,0.08895999789237977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.01,0.17110400199890136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,16,power_law_1.01,0.10043519735336304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.01,0.22152960300445557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.01,0.10633599758148193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.01,0.2644416093826294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.01,0.12498559951782226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.01,0.35810561180114747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.01,0.10742399692535401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.01,0.44316802024841306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.01,0.12250239849090576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.01,0.6141312122344971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.01,0.149510395526886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.01,0.819155216217041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.01,0.20471038818359374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,16,power_law_1.01,0.9840255737304687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.01,0.2519615888595581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,16,power_law_1.01,1.5537407875061036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.01,0.3529599905014038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,16,power_law_1.01,3.1276031494140626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.01,0.4809216022491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,16,power_law_1.01,0.5829504013061524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,16,power_law_1.01,0.9024576187133789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,16,power_law_1.01,1.8879295349121095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,16,power_law_1.2,0.06617599725723267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,16,power_law_1.2,0.09095039963722229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,16,power_law_1.2,0.051737600564956666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,16,power_law_1.2,0.061919999122619626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,16,power_law_1.2,0.06817920207977295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,16,power_law_1.2,0.07052159905433655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,16,power_law_1.2,0.08581119775772095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,16,power_law_1.2,0.054681599140167236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,16,power_law_1.2,0.09126399755477906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,16,power_law_1.2,0.07497599720954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,16,power_law_1.2,0.08967679738998413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,16,power_law_1.2,0.047219198942184445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,16,power_law_1.2,0.09516159892082214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,16,power_law_1.2,0.052275198698043826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,16,power_law_1.2,0.09587839841842652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,16,power_law_1.2,0.05553920269012451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,16,power_law_1.2,0.09619839787483216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,16,power_law_1.2,0.06092159748077393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,16,power_law_1.2,0.0995519995689392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,16,power_law_1.2,0.07184000015258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,16,power_law_1.2,0.1002303957939148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,16,power_law_1.2,0.07233279943466187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,16,power_law_1.2,0.11383039951324463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,16,power_law_1.2,0.07496960163116455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,16,power_law_1.2,0.11662720441818238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,16,power_law_1.2,0.0752128005027771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,16,power_law_1.2,0.11463680267333984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,16,power_law_1.2,0.07797120213508606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,16,power_law_1.2,0.13251839876174926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,16,power_law_1.2,0.08143360018730164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,16,power_law_1.2,0.14035840034484864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,16,power_law_1.2,0.08357120156288148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,16,power_law_1.2,0.17286399602890015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,16,power_law_1.2,0.08858240246772767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,16,power_law_1.2,0.19204479455947876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,16,power_law_1.2,0.09724159836769104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,16,power_law_1.2,0.24245119094848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,16,power_law_1.2,0.10204800367355346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,16,power_law_1.2,0.28393599987030027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,16,power_law_1.2,0.1063040018081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,16,power_law_1.2,0.0806335985660553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,16,power_law_1.2,0.0229312002658844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,16,power_law_1.2,0.3960832118988037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,16,power_law_1.2,0.12476160526275634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,16,power_law_1.2,0.1017408013343811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,16,power_law_1.2,0.029868799448013305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,16,power_law_1.2,0.49018239974975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,16,power_law_1.2,0.12799999713897706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,16,power_law_1.2,0.058412802219390866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,16,power_law_1.2,0.022361600399017335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,16,power_law_1.2,0.7187456130981446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,16,power_law_1.2,0.16106239557266236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,16,power_law_1.2,0.0657920002937317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,16,power_law_1.2,0.19137920141220094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,16,power_law_1.2,0.02308479994535446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,16,power_law_1.2,0.8360896110534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,16,power_law_1.2,0.07314559817314148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,16,power_law_1.2,0.23674240112304687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,16,power_law_1.2,0.024435199797153473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,16,power_law_1.2,1.0719167709350585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,16,power_law_1.2,0.08316159844398499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,16,power_law_1.2,0.3054656028747559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,16,power_law_1.2,0.027775999903678895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,16,power_law_1.2,1.7784576416015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,16,power_law_1.2,0.09151359796524047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,16,power_law_1.2,0.4269375801086426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,16,power_law_1.2,0.03729279935359955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,16,power_law_1.2,0.09790719747543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,16,power_law_1.2,3.278054428100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,16,power_law_1.2,0.5368127822875977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,16,power_law_1.2,0.03544319868087768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,16,power_law_1.2,0.09516800045967103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,16,power_law_1.2,0.8154623985290528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,16,power_law_1.2,0.03715200126171112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,16,power_law_1.2,0.09523199796676636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,16,power_law_1.2,1.0162303924560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,16,power_law_1.2,0.04043520092964172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,16,power_law_1.2,0.09444479942321778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,16,power_law_1.2,1.3342399597167969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,16,power_law_1.2,0.03746559917926788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,16,power_law_1.2,0.10131839513778687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,16,power_law_1.2,2.0099456787109373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,16,power_law_1.2,0.041715198755264284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,16,power_law_1.2,0.0960640013217926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,16,power_law_1.2,0.04574080109596253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,16,power_law_1.2,4.101728057861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,16,power_law_1.2,0.10016640424728393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,16,power_law_1.2,0.05082880258560181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,16,power_law_1.2,0.1061568021774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,16,power_law_1.2,0.0586624026298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,16,power_law_1.2,0.10657919645309448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,16,power_law_1.2,0.06177279949188232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,16,power_law_1.2,0.11136640310287475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,16,power_law_1.2,0.09063680171966552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,16,power_law_1.2,0.12697600126266478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,16,power_law_1.2,0.10069760084152221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.2,0.1363711953163147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.2,0.10717439651489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.2,0.1636031985282898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.2,0.13169280290603638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.2,0.18317439556121826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.2,0.10537600517272949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.2,0.23651199340820311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.2,0.13596160411834718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.2,0.28706560134887693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.2,0.1523200035095215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.2,0.3789504051208496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.2,0.22092161178588868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.2,0.5076288223266602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.2,0.27151999473571775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.2,0.6604608058929443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.2,0.3808511972427368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.2,0.9520000457763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.2,0.5192639827728271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,16,power_law_1.2,1.1423168182373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,16,power_law_1.2,0.6346303939819335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,16,power_law_1.2,1.7208831787109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,16,power_law_1.2,0.9828351974487305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,16,power_law_1.2,3.755091094970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,16,power_law_1.2,2.043270492553711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,8,balanced,0.021546666820844013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,8,balanced,0.022954667607943218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,8,balanced,0.025370667378107708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,8,balanced,0.025402667621771496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,8,balanced,0.029103999336560566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,8,balanced,0.04142933338880539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,8,balanced,0.052000001072883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,8,balanced,0.04407466451327006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,8,balanced,0.053823997577031456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,8,balanced,0.04407466451327006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,8,balanced,0.05736533304055532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,8,balanced,0.04597333570321401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,8,balanced,0.05622933308283488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,8,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,8,balanced,0.06543999910354614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,8,balanced,0.0498933345079422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,8,balanced,0.08158400158087413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,8,balanced,0.05209066470464071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,8,balanced,0.08058133224646251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,8,balanced,0.055445333321889244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,8,balanced,0.08505599697430928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,8,balanced,0.0589279979467392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,8,balanced,0.08452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,8,balanced,0.06423466900984447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,8,balanced,0.08590933680534363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,8,balanced,0.04785066843032837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,8,balanced,0.06930133203665416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,8,balanced,0.08718933661778767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,8,balanced,0.07393066585063934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,8,balanced,0.09078933795293172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,8,balanced,0.047728002071380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,8,balanced,0.08793600400288899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,8,balanced,0.050383999943733215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,8,balanced,0.09281599521636963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,8,balanced,0.11934933066368103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,8,balanced,0.05376000205675761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,8,balanced,0.09303999940554301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,8,balanced,0.04437866806983948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,8,balanced,0.15920533736546835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,8,balanced,0.07445333401362102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,8,balanced,0.099589337905248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,8,balanced,0.04577066500981649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,8,balanced,0.14909332990646362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,8,balanced,0.10572800040245056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,8,balanced,0.10358933607737224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,8,balanced,0.04775466521581014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,8,balanced,0.1067680021127065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,8,balanced,0.17988266547520956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,8,balanced,0.10528000195821126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,8,balanced,0.05049066742261251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,8,balanced,0.10777599612871806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,8,balanced,0.11762666702270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,8,balanced,0.20325867335001627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,8,balanced,0.05819733440876007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,8,balanced,0.12846400340398154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,8,balanced,0.10851200421651204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,8,balanced,0.08191999793052673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,8,balanced,0.257152001063029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,8,balanced,0.1460693379243215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,8,balanced,0.1091306706269582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,8,balanced,0.08178666730721791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,8,balanced,0.16220266620318094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,8,balanced,0.3166026671727498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,8,balanced,0.10902933279673259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,8,balanced,0.08527466654777527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,8,balanced,0.19688532749811807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,8,balanced,0.10980799794197083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,8,balanced,0.0848640004793803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,8,balanced,0.4790613253911336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,8,balanced,0.23435733715693155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,8,balanced,0.1111306647459666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,8,balanced,0.08649599552154541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,8,balanced,0.11125333110491435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,8,balanced,0.32102932532628375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,8,balanced,0.0872213343779246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,8,balanced,0.11834133664766948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,8,balanced,0.08935999870300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,8,balanced,0.6004853248596191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,8,balanced,0.11991999546686809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,8,balanced,0.09029866258303325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,8,balanced,0.3982773224512736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,8,balanced,0.12057600418726604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,8,balanced,0.09169600407282512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,8,balanced,0.13408533732096353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,8,balanced,0.10001599788665771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,8,balanced,0.13981866836547852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,8,balanced,0.10418132940928142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,8,balanced,0.1565546691417694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,8,balanced,0.5646986564000448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,8,balanced,0.7771626313527426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,8,balanced,0.10981866717338562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,8,balanced,0.1750453313191732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,8,balanced,0.12581866979599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,8,balanced,0.13369066516558328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,8,balanced,0.21041067441304526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,8,balanced,0.7277493476867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,8,balanced,0.15473066767056784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,8,power_law_1.01,0.051667201519012454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,8,balanced,0.24658666054407755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,8,balanced,0.17783466974894205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,8,power_law_1.01,0.06711680293083191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,8,balanced,1.188208023707072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,8,balanced,0.3142186601956685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,8,balanced,0.22955199082692465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,8,power_law_1.01,0.05578240156173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,8,power_law_1.01,0.04614399969577789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,8,balanced,0.8903413613637289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,8,power_law_1.01,0.05847039818763733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,8,balanced,0.2794506748517354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,8,power_law_1.01,0.05178880095481873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,8,balanced,0.3861120144526164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,8,power_law_1.01,0.06926079988479614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,8,power_law_1.01,0.04663679897785187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,8,balanced,0.37861867745717365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,8,power_law_1.01,0.0791487991809845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,8,power_law_1.01,0.05379840135574341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,8,balanced,0.5329440037409464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,8,power_law_1.01,0.0890496015548706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,8,power_law_1.01,0.057158398628234866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,8,balanced,0.4759146769841512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,8,balanced,1.3788800239562988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,8,power_law_1.01,0.0924351990222931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,8,power_law_1.01,0.06380800008773804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,8,balanced,0.6765493551890055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,8,power_law_1.01,0.09287040233612061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,8,balanced,2.3333120346069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,8,power_law_1.01,0.07287039756774902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,8,balanced,0.6789546807607015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,8,power_law_1.01,0.09799039959907532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,8,power_law_1.01,0.0779583990573883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,8,power_law_1.01,0.10053119659423829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,8,balanced,0.8181119759877523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,8,power_law_1.01,0.0791104018688202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,8,power_law_1.01,0.10306559801101685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,8,power_law_1.01,0.07832319736480713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,8,balanced,0.8770346641540527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,8,power_law_1.01,0.1046720027923584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,8,power_law_1.01,0.0825215995311737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,8,power_law_1.01,0.10478719472885131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,8,power_law_1.01,0.08589439988136291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,8,power_law_1.01,0.11279360055923462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,8,power_law_1.01,0.0889855980873108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,8,power_law_1.01,0.11473920345306396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,8,balanced,2.6961119969685874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,8,power_law_1.01,0.09488639831542969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,8,power_law_1.01,0.12606719732284546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,8,balanced,1.0731093088785808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,8,balanced,1.226912021636963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,8,power_law_1.01,0.10588799715042115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,8,power_law_1.01,0.13552000522613525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,8,power_law_1.01,0.10844800472259522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,8,power_law_1.01,0.15831680297851564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,8,power_law_1.01,0.07100800275802613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,8,power_law_1.01,0.11320960521697998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,8,power_law_1.01,0.02111999988555908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,8,power_law_1.01,0.18353919982910155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,8,power_law_1.01,0.08520320057868958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,8,power_law_1.01,0.12856320142745972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,8,power_law_1.01,0.023526400327682495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,8,balanced,1.6598933537801106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,8,power_law_1.01,0.2006527900695801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,8,power_law_1.01,0.06624000072479248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,8,power_law_1.01,0.1437824010848999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,8,power_law_1.01,0.023897600173950196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,8,power_law_1.01,0.25090560913085935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,8,power_law_1.01,0.06962559819221496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,8,power_law_1.01,0.17134720087051392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,8,power_law_1.01,0.02428800016641617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,8,balanced,2.373589356740316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,8,power_law_1.01,0.30271999835968016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,8,power_law_1.01,0.07910400032997131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,8,power_law_1.01,0.1982143998146057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,8,power_law_1.01,0.026188799738883974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,8,power_law_1.01,0.2570111989974976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,8,power_law_1.01,0.39650559425354004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,8,power_law_1.01,0.09423360228538513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,8,power_law_1.01,0.3275264024734497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,8,power_law_1.01,0.029216000437736513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,8,power_law_1.01,0.09792640209197997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,8,power_law_1.01,0.4901120185852051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,8,power_law_1.01,0.43061118125915526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,8,power_law_1.01,0.03364480137825012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,8,power_law_1.01,0.5638463973999024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,8,power_law_1.01,0.10143359899520873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,8,power_law_1.01,0.6761600017547608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,8,balanced,3.273248036702474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,8,power_law_1.01,0.0333759993314743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,8,power_law_1.01,0.7894720077514649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,8,power_law_1.01,0.10523519515991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,8,power_law_1.01,0.8762944221496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,8,power_law_1.01,0.0352512001991272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,8,power_law_1.01,1.0409728050231934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,8,power_law_1.01,0.10227199792861938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,8,power_law_1.01,1.034444808959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,8,power_law_1.01,0.03548159897327423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,8,power_law_1.01,1.311564826965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,8,power_law_1.01,0.10792319774627686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,8,power_law_1.01,1.7219839096069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,8,power_law_1.01,0.0377023994922638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,8,power_law_1.01,0.11122560501098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,8,power_law_1.01,2.0379072189331056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,8,power_law_1.01,2.9680192947387694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,8,power_law_1.01,0.03943679928779602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,8,power_law_1.01,0.1124351978302002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,8,power_law_1.01,4.329081726074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,8,power_law_1.01,0.03946239948272705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,8,power_law_1.01,0.11397119760513305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,8,power_law_1.01,0.04952319860458374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,8,power_law_1.01,0.12138240337371826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,8,power_law_1.01,0.05440639853477478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,8,power_law_1.01,0.12346880435943604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,8,power_law_1.01,0.05614079833030701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,8,power_law_1.01,0.1289471983909607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,8,power_law_1.01,0.055929601192474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,8,power_law_1.01,0.13850879669189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,8,power_law_1.01,0.06684799790382386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.01,0.15038080215454103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.01,0.10468480587005616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.01,0.18068480491638184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.01,0.12350080013275147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.01,0.20483200550079345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.01,0.11308159828186035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.01,0.2653183937072754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.01,0.13458559513092042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.01,0.30027520656585693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.01,0.15101439952850343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.01,0.4038080215454102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.01,0.20383360385894775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.01,0.5113791942596435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.01,0.24713599681854248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.01,0.7422912120819092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.01,0.3452415943145752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.01,0.9630144119262696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.01,0.45124478340148927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,8,power_law_1.01,1.1989888191223144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,8,power_law_1.01,0.5531968116760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,8,power_law_1.01,1.809107208251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,8,power_law_1.01,0.8714752197265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,8,power_law_1.01,3.7869121551513674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,8,power_law_1.01,1.6661376953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,8,power_law_1.2,0.0466048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,8,power_law_1.2,0.021049599349498748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,8,power_law_1.2,0.07253119945526124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,8,power_law_1.2,0.0500927984714508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,8,power_law_1.2,0.05111680030822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,8,power_law_1.2,0.02418559938669205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,8,power_law_1.2,0.08503680229187012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,8,power_law_1.2,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,8,power_law_1.2,0.06702719926834107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,8,power_law_1.2,0.023897600173950196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,8,power_law_1.2,0.06410239934921265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,8,power_law_1.2,0.052083200216293334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,8,power_law_1.2,0.048825600743293764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,8,power_law_1.2,0.024403199553489685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,8,power_law_1.2,0.07303040027618408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,8,power_law_1.2,0.05591040253639221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,8,power_law_1.2,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,8,power_law_1.2,0.027372801303863527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,8,power_law_1.2,0.06328960061073304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,8,power_law_1.2,0.07720320224761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,8,power_law_1.2,0.06394879817962647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,8,power_law_1.2,0.028940799832344054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,8,power_law_1.2,0.0711296021938324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,8,power_law_1.2,0.08644480109214783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,8,power_law_1.2,0.07530879974365234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,8,power_law_1.2,0.03250559866428375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,8,power_law_1.2,0.07580800056457519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,8,power_law_1.2,0.1021183967590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,8,power_law_1.2,0.032595199346542356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,8,power_law_1.2,0.08957440257072449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,8,power_law_1.2,0.0774399995803833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,8,power_law_1.2,0.1045375943183899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,8,power_law_1.2,0.03439359962940216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,8,power_law_1.2,0.09327999949455261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,8,power_law_1.2,0.07845119833946228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,8,power_law_1.2,0.10787839889526367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,8,power_law_1.2,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,8,power_law_1.2,0.09072639942169189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,8,power_law_1.2,0.08193280100822449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,8,power_law_1.2,0.10595200061798096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,8,power_law_1.2,0.0379584014415741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,8,power_law_1.2,0.09758719801902771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,8,power_law_1.2,0.08487679958343505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,8,power_law_1.2,0.10687359571456909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,8,power_law_1.2,0.038361600041389464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,8,power_law_1.2,0.0980288028717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,8,power_law_1.2,0.08887680172920227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,8,power_law_1.2,0.10722559690475464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,8,power_law_1.2,0.04046080112457275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,8,power_law_1.2,0.10115200281143188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,8,power_law_1.2,0.11476479768753052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,8,power_law_1.2,0.09483519792556763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,8,power_law_1.2,0.049004799127578734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,8,power_law_1.2,0.10410879850387574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,8,power_law_1.2,0.11823999881744385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,8,power_law_1.2,0.1080191969871521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,8,power_law_1.2,0.12253439426422119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,8,power_law_1.2,0.05411199927330017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,8,power_law_1.2,0.10400639772415161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,8,power_law_1.2,0.10825599431991577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,8,power_law_1.2,0.12615679502487182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,8,power_law_1.2,0.0566976010799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,8,power_law_1.2,0.11433600187301636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,8,power_law_1.2,0.11610239744186401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,8,power_law_1.2,0.13084800243377687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,8,power_law_1.2,0.05594239830970764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,8,power_law_1.2,0.11661440134048462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,4,balanced,0.021370666722456615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,8,power_law_1.2,0.13127039670944213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,8,power_law_1.2,0.1431040048599243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,8,power_law_1.2,0.06823679804801941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,8,power_law_1.2,0.1286080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,8,power_law_1.2,0.15038080215454103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,4,balanced,0.023120000958442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.2,0.16113920211791993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.2,0.11020159721374512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,8,power_law_1.2,0.14568320512771607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,8,power_law_1.2,0.1788480043411255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,4,balanced,0.025397333006064098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.2,0.18287359476089476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.2,0.12271360158920289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,8,power_law_1.2,0.15820800065994262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,8,power_law_1.2,0.21288321018218995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,4,balanced,0.029535998900731403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.2,0.21456000804901124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,4,balanced,0.05602666735649109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.2,0.11448960304260254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,8,power_law_1.2,0.19006719589233398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,8,power_law_1.2,0.274835205078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,4,balanced,0.031658666829268135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.2,0.26784000396728513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,4,balanced,0.05805333455403646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.2,0.13655680418014526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,8,power_law_1.2,0.21823360919952392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,8,power_law_1.2,0.37342081069946287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,4,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.2,0.3303168058395386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,4,balanced,0.05940799911816915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.2,0.15681920051574708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,8,power_law_1.2,0.25471360683441163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,8,power_law_1.2,0.47448320388793946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,4,balanced,0.04809066653251648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.2,0.4590400218963623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,4,balanced,0.06612800061702728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.2,0.21568639278411866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,8,power_law_1.2,0.3212287902832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,8,power_law_1.2,0.6276735782623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,4,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.2,0.5772480010986328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,4,balanced,0.07769600053628285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,8,power_law_1.2,0.8798272132873535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.2,0.2729599952697754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,8,power_law_1.2,0.43704957962036134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,4,balanced,0.050144001841545105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.2,0.7954815864562989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,4,balanced,0.1069546639919281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,8,power_law_1.2,1.1232447624206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.2,0.37804160118103025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,8,power_law_1.2,0.5111807823181153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,4,balanced,0.05206400156021118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.2,1.073094367980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,4,balanced,0.10814932982126872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,4,balanced,0.04789333542188009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,8,power_law_1.2,1.6262079238891602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,4,balanced,0.04987200101216634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.2,0.48894081115722654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,8,power_law_1.2,0.7599552154541016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,4,balanced,0.05418133238951365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,4,balanced,0.11134399970372517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,8,power_law_1.2,1.2041472434997558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,4,balanced,0.050954664746920265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,8,power_law_1.2,2.3017791748046874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,4,balanced,0.04993600149949392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,8,power_law_1.2,0.5655807971954345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,8,power_law_1.2,1.0493056297302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,4,balanced,0.05717866619427999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,4,balanced,0.11160000165303548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,8,power_law_1.2,2.0465215682983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,4,balanced,0.05146133402983347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,4,balanced,0.11365866661071777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,4,balanced,0.052069331208864846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,8,power_law_1.2,4.326137542724609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,8,power_law_1.2,0.8872511863708497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,8,power_law_1.2,1.1963071823120117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,4,balanced,0.05840533475081126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,4,balanced,0.055573334296544395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,8,power_law_1.2,4.0048065185546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,4,balanced,0.11251733700434367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,4,balanced,0.06005866825580597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,8,power_law_1.2,1.8283008575439452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,8,power_law_1.2,1.8678400039672851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,4,balanced,0.061280002196629844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,4,balanced,0.06588799754778545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,4,balanced,0.11532800396283467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,4,balanced,0.08162133395671844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,4,balanced,0.06446399788061778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,8,power_law_1.2,3.9722110748291017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,4,balanced,0.09453866879145305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,4,balanced,0.11551466584205627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,4,balanced,0.11252267162005107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,4,balanced,0.06849066913127899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,4,balanced,0.09604799747467041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,4,balanced,0.11873599886894226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,4,balanced,0.11565333604812622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,4,balanced,0.08084799846013387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,4,balanced,0.09798399607340495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,4,balanced,0.12613333264986673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,4,balanced,0.11633599797884624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,4,balanced,0.09063466389973958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,4,balanced,0.10089600086212158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,4,balanced,0.12829333543777466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,4,balanced,0.11781866351763408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,4,balanced,0.1276479959487915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,4,balanced,0.10016533732414246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,4,balanced,0.13518399993578592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,4,balanced,0.11868799726168315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,4,balanced,0.10525332887967427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,4,balanced,0.1432213286558787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,4,balanced,0.14915200074513754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,4,balanced,0.12276800473531087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,4,balanced,0.10707733035087585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,4,balanced,0.1623306671778361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,4,balanced,0.1586186687151591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,4,balanced,0.12109333276748657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,4,balanced,0.10754133264223735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,4,balanced,0.19393599033355713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,4,balanced,0.18685332934061685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,4,balanced,0.11089600125948589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,4,balanced,0.12184533476829529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,4,balanced,0.21742933988571167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,4,balanced,0.20992533365885416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,4,balanced,0.11999467015266418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,4,balanced,0.12492266297340393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,4,balanced,0.12317867080370586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,4,balanced,0.26024534304936725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,4,balanced,0.2787359952926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,4,balanced,0.13404800494511923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,4,balanced,0.13127999504407248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,4,balanced,0.13645866513252258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,4,balanced,0.321343998114268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,4,balanced,0.3204960028330485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,4,balanced,0.1402186652024587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,4,balanced,0.14747732877731323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,4,balanced,0.4573440154393514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,4,balanced,0.15621866782506308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,4,balanced,0.15770666797955832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,4,balanced,0.49971731503804523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,4,balanced,0.5738186836242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,4,balanced,0.164682666460673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,4,balanced,0.18634132544199625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,4,balanced,0.19083199898401895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,4,balanced,0.21423467000325522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,4,balanced,0.8366560141245524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,4,balanced,0.21221866210301718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,4,balanced,0.6080746650695801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,4,balanced,0.29283734162648517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,4,balanced,0.2767519950866699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,4,balanced,0.34572800000508624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,4,balanced,1.0868000189463298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,4,balanced,0.31571199496587116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,4,balanced,0.49486398696899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,4,balanced,0.7819946606953939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,4,balanced,0.43798398971557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,4,balanced,0.6200639804204305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,4,balanced,1.3514240582784016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,4,balanced,0.5476640065511068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,4,balanced,0.9042773246765137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,4,balanced,0.7750186920166016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,4,balanced,1.2024640242258708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,4,balanced,1.1713973681132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,4,balanced,2.060357411702474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,4,balanced,0.9782346884409586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,4,balanced,1.4391147295633953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,4,balanced,1.2207573254903157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,4,balanced,2.3723413149515786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,4,balanced,2.236015955607096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,4,balanced,1.8516213099161785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,4,balanced,4.012240091959636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,4,balanced,4.384538650512695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,4,balanced,3.5157012939453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,4,power_law_1.01,0.07265920042991639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,4,power_law_1.01,0.04849919974803925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,4,power_law_1.01,0.04652160108089447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,4,power_law_1.01,0.08451840281486511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,4,power_law_1.01,0.0208064004778862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,4,power_law_1.01,0.051635199785232545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,4,power_law_1.01,0.050470399856567386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,4,power_law_1.01,0.06770560145378113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,4,power_law_1.01,0.04976640045642853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,4,power_law_1.01,0.02279040068387985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,4,power_law_1.01,0.048179200291633605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,4,power_law_1.01,0.07827200293540955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,4,power_law_1.01,0.05868800282478333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,4,power_law_1.01,0.023507200181484222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,4,power_law_1.01,0.05617920160293579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,4,power_law_1.01,0.1001471996307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,4,power_law_1.01,0.07015680074691773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,4,power_law_1.01,0.027961599826812743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,4,power_law_1.01,0.061875200271606444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,4,power_law_1.01,0.11380480527877808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,4,power_law_1.01,0.0800383985042572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,4,power_law_1.01,0.030355200171470642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,4,power_law_1.01,0.07146239876747132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,4,power_law_1.01,0.13157119750976562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,4,power_law_1.01,0.09255040287971497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,4,power_law_1.01,0.033267199993133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,4,power_law_1.01,0.08326399922370911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,4,power_law_1.01,0.10020480155944825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,4,power_law_1.01,0.13974399566650392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,4,power_law_1.01,0.03670400083065033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,4,power_law_1.01,0.08758400082588196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,4,power_law_1.01,0.10120320320129395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,4,power_law_1.01,0.13872640132904052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,4,power_law_1.01,0.03938559889793396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,4,power_law_1.01,0.09041919708251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,4,power_law_1.01,0.10276479721069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,4,power_law_1.01,0.1442752003669739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,4,power_law_1.01,0.0412992000579834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,4,power_law_1.01,0.10699520111083985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,4,power_law_1.01,0.08933759927749634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,4,power_law_1.01,0.13908480405807494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,4,power_law_1.01,0.04216319918632507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,4,power_law_1.01,0.11121280193328857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,4,power_law_1.01,0.09481599926948547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,4,power_law_1.01,0.14449280500411987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,4,power_law_1.01,0.042694398760795595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,4,power_law_1.01,0.11259520053863525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,4,power_law_1.01,0.09949439764022827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,4,power_law_1.01,0.14958720207214354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,4,power_law_1.01,0.04376319944858551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,4,power_law_1.01,0.12269439697265624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,4,power_law_1.01,0.10565760135650634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,4,power_law_1.01,0.15378559827804567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,4,power_law_1.01,0.04560000002384186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,4,power_law_1.01,0.12997759580612184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,4,power_law_1.01,0.11176960468292237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,4,power_law_1.01,0.15679359436035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,4,power_law_1.01,0.04912639856338501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,4,power_law_1.01,0.1337407946586609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,4,power_law_1.01,0.12284159660339355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,4,power_law_1.01,0.16310399770736694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,4,power_law_1.01,0.05292159914970398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,4,power_law_1.01,0.14881279468536376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,4,power_law_1.01,0.1292736053466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,4,power_law_1.01,0.1688256025314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,4,power_law_1.01,0.06270719766616821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,4,power_law_1.01,0.16747519969940186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,4,power_law_1.01,0.12905600070953369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,4,power_law_1.01,0.18373759984970092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,4,power_law_1.01,0.06438400149345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,4,power_law_1.01,0.18595839738845826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,4,power_law_1.01,0.15422719717025757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.01,0.203385591506958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,4,power_law_1.01,0.07823359966278076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,4,power_law_1.01,0.2109312057495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,4,power_law_1.01,0.1722432017326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.01,0.2383359909057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.01,0.10607999563217163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,4,power_law_1.01,0.2450495958328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,4,power_law_1.01,0.20794880390167236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.01,0.26979200839996337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.01,0.13334399461746216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,4,power_law_1.01,0.3130496025085449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,4,power_law_1.01,0.24062080383300782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.01,0.34352641105651854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.01,0.13485440015792846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,4,power_law_1.01,0.38727679252624514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,4,power_law_1.01,0.3286400079727173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.01,0.4255104064941406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.01,0.1587007999420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,4,power_law_1.01,0.493836784362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,4,power_law_1.01,0.40636157989501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.01,0.5821055889129638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.01,0.18143999576568604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,4,power_law_1.01,0.6586431980133056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,4,power_law_1.01,0.5478911876678467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.01,0.2332927942276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.01,0.735148811340332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,4,power_law_1.01,0.8659135818481445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,4,power_law_1.01,0.7047103881835938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.01,0.28707840442657473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,4,power_law_1.2,0.049235200881958006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.01,1.0316672325134277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,4,power_law_1.01,1.144863986968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,4,power_law_1.01,1.0185855865478515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.01,1.3079360008239747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.01,0.39176321029663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,4,power_law_1.2,0.04447360038757324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,4,power_law_1.2,0.05121279954910278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,4,power_law_1.01,1.4676480293273926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,4,power_law_1.01,1.3457216262817382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,4,power_law_1.01,1.692460823059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.01,0.5139135837554931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,4,power_law_1.2,0.048230400681495665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,4,power_law_1.2,0.049030399322509764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,4,power_law_1.01,2.168230438232422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,4,power_law_1.01,1.5709888458251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,4,power_law_1.01,2.68088321685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,4,power_law_1.01,0.6171775817871094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,4,power_law_1.2,0.0453247994184494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,4,power_law_1.2,0.057017600536346434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,4,power_law_1.01,4.014656066894531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,4,power_law_1.01,0.9554240226745605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,4,power_law_1.01,2.517990493774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,4,power_law_1.01,4.942879867553711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,4,power_law_1.2,0.05311359763145447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,4,power_law_1.2,0.06833279728889466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,4,power_law_1.01,1.838764762878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,4,power_law_1.01,4.893324661254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,4,power_law_1.2,0.07793920040130616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,4,power_law_1.2,0.02040960043668747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,4,power_law_1.2,0.05797119736671448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,4,power_law_1.2,0.07505279779434204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,4,power_law_1.2,0.09466239809989929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,4,power_law_1.2,0.021996800601482392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,4,power_law_1.2,0.0664255976676941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,4,power_law_1.2,0.08272640109062195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,4,power_law_1.2,0.09819520115852357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,4,power_law_1.2,0.02314240038394928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,4,power_law_1.2,0.08042880296707153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,4,power_law_1.2,0.0694271981716156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,4,power_law_1.2,0.10032000541687011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,4,power_law_1.2,0.027456000447273254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,4,power_law_1.2,0.08539519906044006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,4,power_law_1.2,0.0833407998085022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,4,power_law_1.2,0.10225919485092164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,4,power_law_1.2,0.029523199796676634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,4,power_law_1.2,0.08528640270233154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,4,power_law_1.2,0.08986240029335021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,4,power_law_1.2,0.1074112057685852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,4,power_law_1.2,0.03281280100345611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,4,power_law_1.2,0.10721280574798583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,4,power_law_1.2,0.0896511971950531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,4,power_law_1.2,0.11143679618835449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,4,power_law_1.2,0.036601600050926206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,4,power_law_1.2,0.13169920444488525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,4,power_law_1.2,0.09654399752616882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,4,power_law_1.2,0.11734399795532227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,4,power_law_1.2,0.03960959911346436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,4,power_law_1.2,0.1347584009170532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,4,power_law_1.2,0.10182399749755859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,4,power_law_1.2,0.12400000095367432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,4,power_law_1.2,0.13630720376968383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,4,power_law_1.2,0.041075199842453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,4,power_law_1.2,0.10657279491424561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,4,power_law_1.2,0.1318336009979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,4,power_law_1.2,0.1429759979248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,4,power_law_1.2,0.04224640130996704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,4,power_law_1.2,0.11607040166854858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,4,power_law_1.2,0.1406399965286255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,4,power_law_1.2,0.14294400215148925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,4,power_law_1.2,0.04289279878139496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,4,power_law_1.2,0.1248960018157959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,4,power_law_1.2,0.15433599948883056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,4,power_law_1.2,0.14375679492950438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,4,power_law_1.2,0.04496000111103058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,4,power_law_1.2,0.13066879510879517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,4,power_law_1.2,0.1748095989227295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,4,power_law_1.2,0.14469120502471924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,4,power_law_1.2,0.04652799963951111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,4,power_law_1.2,0.13550080060958863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,4,power_law_1.2,0.18938239812850952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,4,power_law_1.2,0.1512895941734314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,4,power_law_1.2,0.04937599897384644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,4,power_law_1.2,0.15719679594039918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,4,power_law_1.2,0.220198392868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,4,power_law_1.2,0.1594879984855652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,4,power_law_1.2,0.06051200032234192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,4,power_law_1.2,0.17637759447097778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,4,power_law_1.2,0.2559999942779541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,4,power_law_1.2,0.16171519756317138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,4,power_law_1.2,0.06255999803543091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,4,power_law_1.2,0.21671679019927978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,4,power_law_1.2,0.33661439418792727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,4,power_law_1.2,0.16825599670410157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,4,power_law_1.2,0.06629120111465454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,4,power_law_1.2,0.24762239456176757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,4,power_law_1.2,0.39417600631713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,4,power_law_1.2,0.1903807997703552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,4,power_law_1.2,0.07856000065803528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,4,power_law_1.2,0.32968959808349607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,4,power_law_1.2,0.5631487846374512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.2,0.201964807510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.2,0.12210559844970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,4,power_law_1.2,0.4213247776031494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,4,power_law_1.2,0.6900352001190185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.2,0.24350719451904296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.2,0.13559039831161498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,4,power_law_1.2,0.5782783985137939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,4,power_law_1.2,1.0352512359619142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.2,0.2825920104980469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.2,0.14057600498199463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,4,power_law_1.2,0.7662079811096192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,4,power_law_1.2,1.2692864418029786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.2,0.37730560302734373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.2,0.1625920057296753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,4,power_law_1.2,1.0780287742614747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,4,power_law_1.2,1.517414379119873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.2,0.4366015911102295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.2,0.18372479677200318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,4,power_law_1.2,1.3815744400024415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,4,power_law_1.2,2.447078323364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.2,0.5800384044647217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.2,0.24301440715789796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,4,power_law_1.2,1.6561471939086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.2,0.7617919921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,4,power_law_1.2,4.676448059082031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.2,0.29799680709838866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,4,power_law_1.2,2.741817665100098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.2,1.0893568038940429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.2,0.4157567977905273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.2,1.4567551612854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,4,power_law_1.2,5.5338623046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.2,0.5215871810913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,4,power_law_1.2,1.8833728790283204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,4,power_law_1.2,0.6329728126525879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,4,power_law_1.2,2.598329544067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,4,power_law_1.2,1.0507136344909669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,4,power_law_1.2,5.609779357910156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,4,power_law_1.2,1.9099712371826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,64,balanced,0.027263998985290527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,64,balanced,0.032730666299661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,64,balanced,0.02733866622050603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,64,balanced,0.028912000358104706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,64,balanced,0.03136533250411352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,64,balanced,0.027690666417280834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,64,balanced,0.02932800104220708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,64,balanced,0.027669332921504974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,64,balanced,0.029722665747006733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,64,balanced,0.032885332902272545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,64,balanced,0.03156800071398417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,64,balanced,0.03266666581233343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,64,balanced,0.0334346666932106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,64,balanced,0.03158933420976003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,64,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,64,balanced,0.031541332602500916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,64,balanced,0.03149333347876867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,64,balanced,0.03313600023587545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,64,balanced,0.03339733431736628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,64,balanced,0.0312266672650973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,64,balanced,0.033402666449546814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,64,balanced,0.031504000226656594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,64,balanced,0.033530667424201965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,64,balanced,0.03251733382542928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,64,balanced,0.03431999931732813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,64,balanced,0.03349866718053818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,64,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,64,balanced,0.03632533301909765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,64,balanced,0.03535466641187668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,64,balanced,0.03765333443880081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,64,balanced,0.037989333271980286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,64,balanced,0.03640000025431315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,64,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,64,balanced,0.037621334195137024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,64,balanced,0.03941866755485535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,64,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,64,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,64,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,64,balanced,0.04148799926042557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,64,balanced,0.03990400085846583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,64,balanced,0.04331199824810028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,64,balanced,0.04171200096607208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,64,balanced,0.045663997530937195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,64,balanced,0.04384533564249674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,64,balanced,0.04974933465321859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,64,balanced,0.054245332876841225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,64,balanced,0.05100266635417938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,64,balanced,0.07223999996980031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,64,balanced,0.057760000228881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,64,balanced,0.09352533022562663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,64,balanced,0.06842133402824402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,64,balanced,0.12748266259829202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,64,balanced,0.08281599978605907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,64,balanced,0.1536853313446045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,64,balanced,0.09334933757781982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,64,balanced,0.18386133511861166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,64,balanced,0.12753066420555115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,64,balanced,0.2677706678708394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,2,balanced,0.0499946673711141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,2,power_law_1.01,0.04886400103569031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,2,balanced,0.05035733183224996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,2,power_law_1.01,0.05314559936523437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,2,balanced,0.05415999889373779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,2,power_law_1.2,0.04935680031776428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,2,power_law_1.01,0.05812479853630066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,2,balanced,0.06614399949709575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,2,power_law_1.01,0.06211199760437012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,2,power_law_1.2,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,2,balanced,0.09278399745623271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,2,power_law_1.01,0.07924479842185975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,2,power_law_1.2,0.05779839754104614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,2,balanced,0.1395146648089091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,2,power_law_1.2,0.045926401019096376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,2,balanced,0.05197333296140035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,2,power_law_1.01,0.09604480266571044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,2,power_law_1.2,0.06353279948234558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,2,balanced,0.14507200320561728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,2,power_law_1.01,0.049158400297164916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,2,power_law_1.2,0.05023360252380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,2,balanced,0.05178666611512502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,2,power_law_1.01,0.11485439538955688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,2,power_law_1.2,0.07324159741401673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,2,balanced,0.1367093324661255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,64,balanced,0.20215467611948648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,2,power_law_1.01,0.05195519924163818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,64,balanced,0.4797066847483317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,2,power_law_1.2,0.05120000243186951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,2,balanced,0.05593066910902659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,2,power_law_1.01,0.11927039623260498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,2,power_law_1.2,0.09140480160713196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,2,balanced,0.1372213363647461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,2,power_law_1.01,0.05443840026855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,2,power_law_1.2,0.0584384024143219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,2,balanced,0.06398933132489522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,2,power_law_1.01,0.12387200593948364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,2,power_law_1.2,0.11371519565582275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,2,balanced,0.13945066928863525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,2,power_law_1.01,0.06318079829216003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,2,power_law_1.2,0.06751360297203064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,2,balanced,0.08196799953778584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,2,power_law_1.01,0.12330880165100097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,2,power_law_1.2,0.1194815993309021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,2,balanced,0.14006933569908142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,2,power_law_1.01,0.07270399928092956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,2,power_law_1.2,0.08254079818725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,2,balanced,0.12181333700815837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,2,power_law_1.01,0.12915199995040894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,2,power_law_1.2,0.12168320417404174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,2,balanced,0.14404267072677612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,2,balanced,0.13040533661842346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,2,power_law_1.01,0.08463360071182251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,2,power_law_1.2,0.09940479993820191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,2,power_law_1.01,0.13642239570617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,2,power_law_1.2,0.12343039512634277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,2,balanced,0.14655466874440512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,2,balanced,0.12072533369064331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,2,power_law_1.01,0.10399359464645386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,2,power_law_1.2,0.10703999996185302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,2,power_law_1.01,0.13916800022125245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,2,power_law_1.2,0.12795519828796387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,2,balanced,0.14973333477973938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,2,balanced,0.1234773298104604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,2,power_law_1.01,0.1060863971710205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,2,power_law_1.2,0.10867840051651001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,2,balanced,0.1567520002524058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,2,power_law_1.01,0.14341119527816773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,2,power_law_1.2,0.13306879997253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,2,balanced,0.12492799758911133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,2,power_law_1.01,0.10966399908065796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,2,power_law_1.2,0.11224960088729859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,2,balanced,0.1604746679464976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,2,power_law_1.01,0.15023360252380372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,2,power_law_1.2,0.1357632040977478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,2,balanced,0.1262933313846588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,2,power_law_1.01,0.10999679565429688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,2,power_law_1.2,0.11802239418029785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,2,balanced,0.1684160033861796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,2,power_law_1.01,0.15345920324325563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,2,power_law_1.2,0.14163199663162232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,2,balanced,0.13109333316485086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,2,power_law_1.01,0.11616640090942383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,2,power_law_1.2,0.12545280456542968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,2,balanced,0.18866666158040366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,2,power_law_1.01,0.16810879707336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,2,power_law_1.2,0.1549504041671753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,2,balanced,0.13346133629480997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,2,power_law_1.01,0.12610559463500975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,2,power_law_1.2,0.12879999876022338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,2,power_law_1.01,0.19251840114593505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,2,balanced,0.20619734128316244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,2,power_law_1.2,0.13948160409927368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,2,power_law_1.2,0.1581055998802185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,2,balanced,0.13616533080736795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,2,power_law_1.01,0.12908159494400023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,2,power_law_1.01,0.21970560550689697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,2,balanced,0.2476373314857483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,2,power_law_1.2,0.15560319423675537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,2,power_law_1.2,0.17446399927139283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,2,balanced,0.14324800173441568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,2,power_law_1.01,0.14222079515457153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,2,power_law_1.01,0.2687488079071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,2,balanced,0.2832533319791158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,2,power_law_1.2,0.16273280382156372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,2,power_law_1.2,0.1991039991378784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,2,balanced,0.1495306690533956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,2,power_law_1.01,0.15205760002136232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,2,power_law_1.2,0.16046719551086425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,2,balanced,0.3950186570485433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,2,power_law_1.01,0.33219199180603026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,2,power_law_1.2,0.23333120346069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,2,balanced,0.1604586640993754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,2,power_law_1.01,0.16223360300064088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,2,power_law_1.2,0.19761919975280762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,2,power_law_1.01,0.42538881301879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,2,power_law_1.2,0.22505600452423097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,2,balanced,0.45450135072072345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,2,power_law_1.2,0.28602240085601804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,2,balanced,0.18557866414388022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,2,power_law_1.01,0.16316800117492675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,2,power_law_1.01,0.534662389755249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,2,power_law_1.2,0.27949440479278564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,2,power_law_1.2,0.3277760028839111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,2,balanced,0.6622133255004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,2,balanced,0.20364266633987427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,2,power_law_1.01,0.18776320219039916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,2,power_law_1.01,0.7230912208557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,2,balanced,0.8151146570841471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,2,power_law_1.2,0.33649280071258547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,2,power_law_1.2,0.4233727931976318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,2,balanced,0.24889065821965536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,2,power_law_1.01,0.21791999340057372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,2,power_law_1.01,0.8998335838317871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,2,power_law_1.2,0.4409152030944824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,2,power_law_1.01,0.27711360454559325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,2,balanced,0.28995732466379803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,2,power_law_1.2,0.5464320182800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,2,power_law_1.01,1.2781056404113769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,2,balanced,1.1985653241475422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,2,power_law_1.2,0.5661119937896728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,2,power_law_1.01,0.3274240016937256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,2,power_law_1.2,0.7296448230743409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,2,balanced,0.42027731736501056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,2,power_law_1.01,1.6466880798339845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,2,power_law_1.2,0.7856512069702148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,2,power_law_1.01,0.4433407783508301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,2,power_law_1.2,0.9481344223022461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,2,power_law_1.01,2.020377540588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,2,balanced,0.48955734570821124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,2,power_law_1.2,1.0237440109252929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,2,power_law_1.01,0.5469759941101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,2,balanced,1.555829366048177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,2,power_law_1.2,1.2666879653930665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,2,power_law_1.01,3.239353561401367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,2,power_law_1.2,1.5058048248291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,2,power_law_1.01,0.7677696228027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,2,balanced,0.7204533418019613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,2,power_law_1.2,1.7068927764892579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,2,power_law_1.01,6.568147277832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,2,power_law_1.2,1.8700607299804688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,2,power_law_1.01,0.9896127700805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,2,power_law_1.2,2.078175926208496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,2,balanced,1.873205343882243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,2,balanced,0.912874698638916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,2,power_law_1.2,2.2865407943725584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,2,power_law_1.01,1.4604031562805175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,2,power_law_1.2,3.1146047592163084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,2,power_law_1.01,1.8902143478393554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,2,power_law_1.2,3.5455169677734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,2,balanced,1.3332266807556152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,2,power_law_1.2,6.474758148193359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,2,power_law_1.01,2.2790592193603514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,2,power_law_1.2,7.6152702331542965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,2,balanced,2.924544016520182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,2,power_law_1.01,3.696268844604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,2,balanced,1.7357014020284016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,2,power_law_1.01,6.964915466308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,2,balanced,2.144053300221761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,2,balanced,3.3519414265950522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,2,balanced,5.736250559488933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,2,balanced,6.573498407999675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,64,balanced,0.01915733392039935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,64,balanced,0.01708799973130226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,64,balanced,0.018272000054518383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,64,balanced,0.03347733368476232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,64,balanced,0.018922666708628338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,64,balanced,0.034448000291983284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,64,balanced,0.01904533306757609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,64,balanced,0.03356266766786575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,64,balanced,0.021168000996112823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,64,balanced,0.03552533437808355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,64,balanced,0.02109866589307785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,64,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,64,balanced,0.027850667635599773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,64,balanced,0.03544000039498011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,64,balanced,0.028949332733949024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,64,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,64,balanced,0.029264000554879505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,64,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,64,balanced,0.027845333019892376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,64,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,64,balanced,0.029391999046007793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,64,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,64,balanced,0.029317334294319153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,64,balanced,0.035930665830771126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,64,balanced,0.029637334247430164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,64,balanced,0.037834666669368744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,64,balanced,0.03251733382542928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,64,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,64,balanced,0.03143999973932902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,64,balanced,0.0378560001651446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,64,balanced,0.03136533250411352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,64,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,64,balanced,0.03392533212900162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,64,balanced,0.04193066557248434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,64,balanced,0.035589332381884255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,64,balanced,0.04366933306058248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,64,balanced,0.04236799975236257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,64,balanced,0.04187199970086416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,64,balanced,0.05007466673851013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,64,balanced,0.043765331308046974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,64,balanced,0.05762133498986562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,64,balanced,0.04457599918047587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,64,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,64,balanced,0.06851199766000111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,64,balanced,0.04971733192602793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,64,balanced,0.09331732988357544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,64,balanced,0.053861334919929504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,64,balanced,0.11770666639010112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,64,balanced,0.06642666459083557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,64,balanced,0.07451733450094859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,64,balanced,0.16184533635775247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,64,balanced,0.09389866391817729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,64,balanced,0.20633065700531006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,64,balanced,0.11120532949765523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,64,balanced,0.2515733242034912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,64,power_law_1.01,0.025305598974227905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,64,balanced,0.12941867113113403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,64,power_law_1.01,0.02563199996948242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,64,power_law_1.01,0.02503040134906769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,64,power_law_1.01,0.029017600417137145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,64,power_law_1.01,0.031763198971748355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,64,power_law_1.01,0.026617598533630372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,64,power_law_1.01,0.03429119884967804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,64,balanced,0.3858879804611206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,64,power_law_1.01,0.02839039862155914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,64,power_law_1.01,0.030355200171470642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,64,power_law_1.01,0.017811200022697447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,64,power_law_1.01,0.03348479866981506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,64,power_law_1.01,0.028121599555015565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,64,balanced,0.18340800205866495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,64,power_law_1.01,0.030156800150871278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,64,power_law_1.01,0.01681919991970062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,64,power_law_1.01,0.0308351993560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,64,power_law_1.01,0.03222399950027466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,64,power_law_1.01,0.02998400032520294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,64,power_law_1.01,0.017151999473571777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,64,power_law_1.01,0.033004799485206605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,64,power_law_1.01,0.035129600763320924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,64,power_law_1.01,0.030297601222991945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,64,power_law_1.01,0.018041600286960603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,64,power_law_1.01,0.03402880132198334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,64,power_law_1.01,0.03555200099945068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,64,power_law_1.01,0.03128319978713989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,64,power_law_1.01,0.018246400356292724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,64,power_law_1.01,0.03704319894313812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,64,power_law_1.01,0.035743999481201175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,64,power_law_1.01,0.031615999341011045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,64,power_law_1.01,0.02088959962129593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,64,power_law_1.01,0.039552000164985654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,64,power_law_1.2,0.023500800132751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,64,power_law_1.01,0.03583999872207642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,64,power_law_1.01,0.031641599535942075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,64,balanced,0.7465600172678629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,64,power_law_1.01,0.0208064004778862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,64,power_law_1.01,0.040780800580978396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,64,power_law_1.2,0.024294400215148927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,64,power_law_1.01,0.03633280098438263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,64,balanced,0.3065013289451599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,64,power_law_1.01,0.03173120021820068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,64,power_law_1.01,0.028012800216674804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,64,power_law_1.01,0.03440000116825104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,64,power_law_1.2,0.024128000438213348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,64,power_law_1.01,0.03637759983539581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,64,power_law_1.01,0.0318336009979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,64,power_law_1.01,0.028217598795890808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,64,power_law_1.2,0.03178240060806274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,64,power_law_1.01,0.03669120073318481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,64,power_law_1.2,0.027526399493217467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,64,power_law_1.01,0.036499199271202085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,64,power_law_1.01,0.03278720080852508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,64,power_law_1.01,0.028377598524093627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,64,power_law_1.2,0.02794240117073059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,64,power_law_1.01,0.03834240138530731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,64,power_law_1.2,0.025804799795150758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,64,power_law_1.01,0.03751679956912994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,64,power_law_1.01,0.03318400084972382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,64,power_law_1.01,0.02807680070400238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,64,power_law_1.2,0.028288000822067262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,64,power_law_1.01,0.03840000033378601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,64,power_law_1.2,0.02917119860649109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,64,power_law_1.01,0.037503999471664426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,64,power_law_1.01,0.0343423992395401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,64,power_law_1.01,0.028575998544692994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,64,power_law_1.2,0.029843199253082275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,64,power_law_1.01,0.04224640130996704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,64,power_law_1.2,0.029049599170684816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,64,power_law_1.01,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,64,power_law_1.01,0.0285504013299942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,64,power_law_1.01,0.03578239977359772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,64,power_law_1.2,0.030623999238014222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,64,power_law_1.2,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,64,power_law_1.01,0.04429439902305603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,64,power_law_1.01,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,64,power_law_1.01,0.02839680016040802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,64,power_law_1.01,0.03911679983139038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,64,power_law_1.2,0.031673601269721983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,64,power_law_1.2,0.03251200020313263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,64,power_law_1.01,0.04981119930744171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,64,power_law_1.01,0.04099839925765991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,64,power_law_1.01,0.029292801022529603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,64,power_law_1.01,0.039647999405860904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,64,power_law_1.2,0.03150720000267029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,64,power_law_1.2,0.034067198634147644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,64,power_law_1.01,0.05244160294532776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,64,power_law_1.01,0.04160000085830688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,64,power_law_1.01,0.029471999406814574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,64,power_law_1.01,0.042361599206924436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,64,power_law_1.2,0.03164800107479095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,64,power_law_1.2,0.03811199963092804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,64,power_law_1.01,0.0632960021495819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,64,power_law_1.01,0.04415999948978424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,64,power_law_1.01,0.030969598889350893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,64,power_law_1.01,0.0447488009929657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,64,power_law_1.2,0.03223040103912354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,64,power_law_1.2,0.04078719913959503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,64,power_law_1.01,0.06865919828414917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,64,power_law_1.01,0.04746879935264588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,64,power_law_1.01,0.08860160112380981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,64,power_law_1.01,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,64,power_law_1.01,0.04891520142555237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,64,power_law_1.2,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,64,power_law_1.2,0.041126400232315063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,64,power_law_1.01,0.05030400156974792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,64,power_law_1.01,0.11627520322799682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,64,power_law_1.01,0.03461120128631592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,64,power_law_1.01,0.05560320019721985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,64,power_law_1.2,0.03277440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,64,power_law_1.01,0.15057920217514037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,64,power_law_1.01,0.05607039928436279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,64,power_law_1.2,0.03412480056285858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,64,power_law_1.01,0.05637120008468628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,64,power_law_1.01,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,64,power_law_1.2,0.03387520015239716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,64,power_law_1.01,0.18940800428390503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,64,power_law_1.01,0.07258880138397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,64,power_law_1.2,0.037088000774383546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,64,power_law_1.01,0.23107199668884276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,64,power_law_1.01,0.05960959792137146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,64,power_law_1.01,0.04465279877185822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,64,power_law_1.2,0.034668800234794614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,64,power_law_1.01,0.08566399812698364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,64,power_law_1.2,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,64,power_law_1.01,0.32933759689331055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,64,power_law_1.01,0.07187839746475219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,64,power_law_1.01,0.052691197395324706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,64,power_law_1.2,0.03620480000972748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,64,power_law_1.01,0.1164031982421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,64,power_law_1.2,0.041043201088905336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,64,power_law_1.01,0.6114687919616699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,64,power_law_1.01,0.08489599823951721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,64,power_law_1.01,0.06285439729690552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,64,power_law_1.2,0.039155200123786926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,64,power_law_1.01,0.14465919733047486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,64,power_law_1.2,0.04550400078296661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,64,power_law_1.01,0.11587200164794922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,64,power_law_1.01,0.08324480056762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,64,power_law_1.2,0.04067200124263763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,64,power_law_1.01,0.1320575952529907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,64,power_law_1.01,0.21203839778900146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,64,power_law_1.2,0.04869759976863861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,64,power_law_1.01,0.10068479776382447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,64,power_law_1.2,0.043942400813102724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,64,power_law_1.01,0.17957119941711425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,64,power_law_1.01,0.23929600715637206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,64,power_law_1.2,0.058764797449111936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,64,power_law_1.01,0.1418496012687683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,64,power_law_1.2,0.04715520143508911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,64,power_law_1.2,0.06223999857902527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,64,power_law_1.01,0.22894721031188964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,64,power_law_1.01,0.33720319271087645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,64,power_law_1.01,0.18807040452957152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,64,power_law_1.2,0.053932797908782956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,64,power_law_1.2,0.08421120047569275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,64,power_law_1.01,0.28878719806671144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,64,power_law_1.01,0.5256639957427979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,64,power_law_1.01,0.22661120891571046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,64,power_law_1.2,0.058355200290679934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,64,power_law_1.2,0.10078719854354859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,64,power_law_1.01,0.48592000007629393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,64,power_law_1.01,0.9961407661437989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,64,power_law_1.01,0.3429824113845825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,64,power_law_1.2,0.06744319796562195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,64,power_law_1.2,0.14235520362854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,64,power_law_1.01,0.8415424346923828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,64,power_law_1.01,0.6596543788909912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,64,power_law_1.2,0.08725759983062745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,64,power_law_1.2,0.17428480386734008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,64,power_law_1.2,0.1047808051109314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,64,power_law_1.2,0.2756927967071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,64,power_law_1.2,0.17365119457244874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,64,power_law_1.2,0.3263040065765381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,64,power_law_1.2,0.19646719694137574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,64,power_law_1.2,0.31678080558776855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,64,power_law_1.2,0.33048319816589355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,64,power_law_1.2,0.6056896209716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,64,power_law_1.2,0.3857919931411743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,64,power_law_1.2,1.2635647773742675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,64,power_law_1.2,0.5670591831207276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,64,power_law_1.2,0.8403008460998536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,64,power_law_1.2,1.7999103546142579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,64,power_law_1.2,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,64,power_law_1.2,0.01774719953536987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,64,power_law_1.2,0.03371520042419433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,64,power_law_1.2,0.016857600212097167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,64,power_law_1.2,0.032979199290275575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,64,power_law_1.2,0.017190399765968322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,64,power_law_1.2,0.03589119911193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,64,power_law_1.2,0.01817599982023239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,64,power_law_1.2,0.035699200630187986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,64,power_law_1.2,0.018355199694633485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,64,power_law_1.2,0.036345601081848145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,64,power_law_1.2,0.020313599705696107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,64,power_law_1.2,0.036364799737930296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,64,power_law_1.2,0.020556800067424774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,64,power_law_1.2,0.036371201276779175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,64,power_law_1.2,0.027622398734092713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,64,power_law_1.2,0.03718400001525879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,64,power_law_1.2,0.02812800109386444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,64,power_law_1.2,0.03717760145664215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,64,power_law_1.2,0.02797439992427826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,64,power_law_1.2,0.03726719915866852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,64,power_law_1.2,0.02746239900588989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,64,power_law_1.2,0.03760640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,64,power_law_1.2,0.02824319899082184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,64,power_law_1.2,0.038431999087333676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,64,power_law_1.2,0.028095999360084535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,64,power_law_1.2,0.03918719887733459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,64,power_law_1.2,0.028140801191329955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,64,power_law_1.2,0.04246399998664856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,64,power_law_1.2,0.029068800806999206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,64,power_law_1.2,0.042982399463653564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,64,power_law_1.2,0.029798400402069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,64,power_law_1.2,0.04529280066490173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,64,power_law_1.2,0.031436800956726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,64,power_law_1.2,0.04863359928131104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,64,power_law_1.2,0.033190399408340454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,64,power_law_1.2,0.053260797262191774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,64,power_law_1.2,0.03463680148124695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,64,power_law_1.2,0.06116480231285095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,64,power_law_1.2,0.0403328001499176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,64,power_law_1.2,0.06699519753456115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,64,power_law_1.2,0.04666880071163178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,64,power_law_1.2,0.07765759825706482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,64,power_law_1.2,0.05429120063781738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,64,power_law_1.2,0.10050560235977173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,64,power_law_1.2,0.06591359972953796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,64,power_law_1.2,0.13303680419921876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,64,power_law_1.2,0.09091200232505799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,64,power_law_1.2,0.17316479682922364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,64,power_law_1.2,0.11141760349273681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,64,power_law_1.2,0.2637120008468628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,64,power_law_1.2,0.1575808048248291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,64,power_law_1.2,0.357478404045105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,64,power_law_1.2,0.19326080083847047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,64,power_law_1.2,0.40700159072875974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,64,power_law_1.2,0.24602880477905273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,64,power_law_1.2,0.7571648120880127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,64,power_law_1.2,0.37624320983886717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,64,power_law_1.2,1.6369535446166992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,64,power_law_1.2,0.7409728050231934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,32,power_law_1.01,0.02372480034828186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,32,power_law_1.01,0.022163200378417968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,32,power_law_1.01,0.022086399793624877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,32,power_law_1.01,0.024268800020217897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,32,power_law_1.01,0.025011199712753295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,32,power_law_1.01,0.025817599892616273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,32,power_law_1.01,0.02619520127773285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,32,power_law_1.01,0.026995199918746948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,32,power_law_1.01,0.029683199524879456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,32,power_law_1.01,0.030380800366401672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,32,power_law_1.01,0.031167998909950256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,32,power_law_1.01,0.03207040131092072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,32,balanced,0.02665599932273229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,32,power_law_1.01,0.03240959942340851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,32,balanced,0.023823998868465424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,32,power_law_1.01,0.03285120129585266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,32,balanced,0.02510933329661687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,32,power_law_1.01,0.03597440123558045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,32,balanced,0.025040000677108765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,32,power_law_1.01,0.036934399604797365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,32,balanced,0.025402667621771496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,32,power_law_1.01,0.037299200892448425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,32,balanced,0.028149334092934925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,32,power_law_1.01,0.039315199851989745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,32,balanced,0.029167999823888142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,32,power_law_1.01,0.04519039988517761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,32,balanced,0.027877333263556164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,32,power_law_1.01,0.048281601071357726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,32,balanced,0.027376001079877216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,32,power_law_1.01,0.05167999863624573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,32,balanced,0.02741866558790207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,32,balanced,0.017136000096797943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,32,power_law_1.01,0.058899199962615965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,32,balanced,0.02939733366171519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,32,power_law_1.01,0.07167999744415283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,32,balanced,0.029130667448043823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,32,balanced,0.01701333373785019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,32,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,32,power_law_1.01,0.09630079865455628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,32,balanced,0.029738667110602062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,32,balanced,0.017082666357358296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,32,balanced,0.03494933247566223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,32,power_law_1.01,0.1177664041519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,32,balanced,0.0335413341720899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,32,balanced,0.017125333348910015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,32,balanced,0.033589333295822144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,32,power_law_1.01,0.15549440383911134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,32,balanced,0.03438399980465571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,32,balanced,0.018090666582187016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,32,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,32,power_law_1.01,0.20208640098571778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,32,balanced,0.035546667873859406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,32,balanced,0.01926933353145917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,32,balanced,0.035536001125971474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,32,power_law_1.01,0.28126718997955324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,32,balanced,0.037418665985266365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,32,balanced,0.032186667124430336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,32,balanced,0.019498666127522785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,32,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,32,power_law_1.01,0.35041279792785646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,32,balanced,0.030789333085219067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,32,balanced,0.03783999880154928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,32,balanced,0.021290667355060577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,32,balanced,0.0388373335202535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,32,balanced,0.037615999579429626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,32,power_law_1.01,0.7365888118743896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,32,balanced,0.02941333254178365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,32,balanced,0.022597332795461018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,32,balanced,0.03975466638803482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,32,balanced,0.03571200122435888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,32,balanced,0.031210665901501972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,32,balanced,0.02317333221435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,32,balanced,0.04161066561937332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,32,balanced,0.03773866593837738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,32,balanced,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,32,balanced,0.029370665550231934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,32,power_law_1.01,0.03504000008106232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,32,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,32,balanced,0.044112001856168113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,32,balanced,0.0334346666932106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,32,balanced,0.029253333806991577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,32,power_law_1.01,0.030611199140548707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,32,balanced,0.03734400123357773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,32,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,32,balanced,0.045893331368764244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,32,balanced,0.02922133356332779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,32,power_law_1.01,0.03056640028953552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,32,balanced,0.039733332892258964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,32,balanced,0.03288000077009201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,32,balanced,0.05592533449331919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,32,balanced,0.02916266769170761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,32,power_law_1.01,0.030105599761009218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,32,balanced,0.03950933367013931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,32,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,32,balanced,0.031354665756225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,32,power_law_1.01,0.030771198868751525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,32,balanced,0.06409599880377452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,32,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,32,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,32,balanced,0.031317333380381264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,32,power_law_1.01,0.03180800080299377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,32,balanced,0.04186666508515676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,32,balanced,0.03367999941110611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,32,balanced,0.0314026673634847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,32,balanced,0.08175466458002727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,32,balanced,0.037530665596326195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,32,power_law_1.01,0.03178240060806274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,32,balanced,0.04189866781234741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,32,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,32,balanced,0.0395359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,32,power_law_1.01,0.03221119940280914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,32,balanced,0.0436160018046697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,32,balanced,0.10368000467618306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,32,balanced,0.034314667185147606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,32,balanced,0.044639999667803444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,32,power_law_1.01,0.032358399033546446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,32,balanced,0.04257066547870636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,32,balanced,0.036144000788529716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,32,balanced,0.051914667089780174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,32,power_law_1.01,0.03218559920787811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,32,balanced,0.04401599864164988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,32,balanced,0.11728533109029134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,32,balanced,0.039621333281199135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,32,balanced,0.062309334675470986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,32,power_law_1.01,0.03318400084972382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,32,balanced,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,32,balanced,0.03832533210515976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,32,power_law_1.01,0.034118399024009705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,32,balanced,0.07807466884454091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,32,balanced,0.04987200101216634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,32,balanced,0.03949866692225138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,32,balanced,0.1530346671740214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,32,power_law_1.01,0.03449600040912628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,32,balanced,0.10453866918881734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,32,balanced,0.04159999887148539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,32,balanced,0.05203199883302053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,32,power_law_1.01,0.036320000886917114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,32,balanced,0.0418453315893809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,32,balanced,0.13166933258374533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,32,power_law_1.01,0.04000000059604645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,32,balanced,0.058143998185793556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,32,balanced,0.043925335009892784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,32,power_law_1.01,0.041740798950195314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,32,balanced,0.17942933241526285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,32,balanced,0.048021331429481506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,32,power_law_1.01,0.04180479943752289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,32,balanced,0.07567999760309856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,32,balanced,0.05402666827042898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,32,power_law_1.01,0.04581120014190674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,32,balanced,0.23228800296783447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,32,power_law_1.01,0.049670401215553286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,32,balanced,0.05972266693909963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,32,power_law_1.01,0.03518719971179962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,32,power_law_1.01,0.01817599982023239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,32,balanced,0.08481599887212117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,32,power_law_1.01,0.05368319749832153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,32,balanced,0.2539733250935872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,32,power_law_1.01,0.03489919900894165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,32,balanced,0.0786186655362447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,32,balanced,0.1076800028483073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,32,power_law_1.01,0.01658879965543747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,32,balanced,0.28168533245722455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,32,power_law_1.01,0.06453120112419128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,32,power_law_1.01,0.03317759931087494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,32,power_law_1.01,0.01637759953737259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,32,balanced,0.10718400279680888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,32,power_law_1.01,0.07255039811134338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,32,power_law_1.01,0.03551360070705414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,32,power_law_1.01,0.017215999960899352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,32,balanced,0.13800533612569174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,32,power_law_1.01,0.0933568000793457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,32,power_law_1.01,0.03665919899940491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,32,balanced,0.1453546682993571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,32,power_law_1.01,0.01701119989156723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,32,power_law_1.01,0.13049600124359131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,32,power_law_1.01,0.03828479945659637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,32,balanced,0.43516798814137775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,32,power_law_1.01,0.017740799486637114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,32,power_law_1.01,0.16228480339050294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,32,power_law_1.01,0.03824639916419983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,32,balanced,0.16270933548609415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,32,balanced,0.17482666174570718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,32,power_law_1.01,0.017798399925231932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,32,power_law_1.01,0.22053120136260987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,32,power_law_1.01,0.037625598907470706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,32,power_law_1.01,0.020524799823760986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,32,power_law_1.01,0.2756160020828247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,32,power_law_1.01,0.03763839900493622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,32,power_law_1.01,0.020582400262355804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,32,power_law_1.01,0.31826560497283934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,32,power_law_1.01,0.03824639916419983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,32,power_law_1.01,0.021881599724292756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,32,balanced,0.21171732743581137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,32,power_law_1.01,0.5412479877471924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,32,power_law_1.01,0.03901439905166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,32,balanced,0.22609599431355795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,32,power_law_1.01,0.02766079902648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,32,balanced,0.8460373083750407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,32,power_law_1.01,1.1392064094543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,32,power_law_1.01,0.03947519958019256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,32,power_law_1.01,0.028569599986076354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,32,power_law_1.01,0.04046719968318939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,32,power_law_1.01,0.02863999903202057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,32,power_law_1.01,0.04008319973945618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,32,balanced,0.30537599325180054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,32,power_law_1.01,0.028121599555015565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,32,power_law_1.01,0.04291839897632599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,32,power_law_1.01,0.028652799129486085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,32,power_law_1.01,0.04415999948978424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,32,power_law_1.01,0.028755199909210206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,32,power_law_1.01,0.04614399969577789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,32,power_law_1.01,0.029919999837875366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,32,power_law_1.01,0.05203199982643127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,32,power_law_1.01,0.03239679932594299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,32,power_law_1.01,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,32,power_law_1.01,0.035596799850463864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,32,power_law_1.01,0.05985919833183288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,32,power_law_1.01,0.039078399538993835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,32,power_law_1.01,0.06471040248870849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,32,power_law_1.01,0.04333440065383911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,32,balanced,0.391759991645813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,32,power_law_1.01,0.0833791971206665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,32,power_law_1.01,0.051283198595047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,32,power_law_1.01,0.09240959882736206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,32,power_law_1.01,0.061427199840545656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,32,power_law_1.01,0.12421760559082032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,32,power_law_1.01,0.08284800052642823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,32,balanced,0.5448799928029379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,32,power_law_1.01,0.15496959686279296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,32,power_law_1.01,0.09878399968147278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,32,power_law_1.01,0.19619840383529663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,32,power_law_1.01,0.1356735944747925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,32,power_law_1.01,0.2608511924743652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,32,power_law_1.01,0.17363840341567993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,32,power_law_1.01,0.3095616102218628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,32,power_law_1.01,0.21290240287780762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,32,power_law_1.01,0.4991487979888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,32,power_law_1.01,0.3298432111740112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,32,power_law_1.01,1.062828826904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,32,power_law_1.01,0.6135424137115478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,32,power_law_1.2,0.023552000522613525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,32,power_law_1.2,0.02237440049648285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,32,power_law_1.2,0.022726400196552275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,32,power_law_1.2,0.023923200368881226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,32,power_law_1.2,0.02499839961528778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,32,power_law_1.2,0.025836798548698425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,32,power_law_1.2,0.0265855997800827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,32,power_law_1.2,0.0301503986120224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,32,power_law_1.2,0.02971520125865936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,32,power_law_1.2,0.03061760067939758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,32,power_law_1.2,0.03184640109539032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,32,power_law_1.2,0.032364800572395325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,32,power_law_1.2,0.03290880024433136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,32,power_law_1.2,0.0332863986492157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,32,power_law_1.2,0.03616639971733093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,32,power_law_1.2,0.03701759874820709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,32,power_law_1.2,0.03825919926166534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,32,power_law_1.2,0.04523519873619079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,32,power_law_1.2,0.04818559885025024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,32,power_law_1.2,0.05060480237007141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,32,power_law_1.2,0.05803520083427429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,32,power_law_1.2,0.07242239713668823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,32,power_law_1.2,0.09189760088920593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,32,power_law_1.2,0.12436480522155761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,32,power_law_1.2,0.15808000564575195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,32,power_law_1.2,0.19062399864196777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,32,power_law_1.2,0.29453439712524415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,32,power_law_1.2,0.36087679862976074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,32,power_law_1.2,0.6240064144134522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,32,power_law_1.2,1.4833279609680177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,32,power_law_1.2,0.03614720106124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,32,power_law_1.2,0.028467199206352232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,32,power_law_1.2,0.028492799401283263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,32,power_law_1.2,0.030028799176216127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,32,power_law_1.2,0.030291199684143066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,32,power_law_1.2,0.03172479867935181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,32,power_law_1.2,0.03214719891548157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,32,power_law_1.2,0.03209599852561951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,32,power_law_1.2,0.03221760094165802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,32,power_law_1.2,0.03252480030059814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,32,power_law_1.2,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,32,power_law_1.2,0.033670398592948916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,32,power_law_1.2,0.03485440015792847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,32,power_law_1.2,0.03594239950180054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,32,power_law_1.2,0.041580799221992495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,32,power_law_1.2,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,32,power_law_1.2,0.04323840141296387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,32,power_law_1.2,0.046105599403381346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,32,power_law_1.2,0.05008640289306641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,32,power_law_1.2,0.06116480231285095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,32,power_law_1.2,0.06677759885787964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,32,power_law_1.2,0.09560959935188293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,32,power_law_1.2,0.11074559688568116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,32,power_law_1.2,0.1414080023765564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,32,power_law_1.2,0.19343359470367433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,32,power_law_1.2,0.29386239051818847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,32,power_law_1.2,0.3932607889175415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,32,power_law_1.2,0.47129597663879397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,32,power_law_1.2,0.9682623863220214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,32,power_law_1.2,1.9099071502685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,32,power_law_1.2,0.035488000512123107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,32,power_law_1.2,0.015967999398708344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,32,power_law_1.2,0.03351039886474609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,32,power_law_1.2,0.016358399391174318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,32,power_law_1.2,0.03292160034179688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,32,power_law_1.2,0.01598079949617386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,32,power_law_1.2,0.0353408008813858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,32,power_law_1.2,0.015897600352764128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,32,power_law_1.2,0.036108800768852235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,32,power_law_1.2,0.016927999258041383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,32,power_law_1.2,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,32,power_law_1.2,0.017343999445438386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,32,power_law_1.2,0.038047999143600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,32,power_law_1.2,0.017734399437904357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,32,power_law_1.2,0.037785598635673524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,32,power_law_1.2,0.020345599949359895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,32,power_law_1.2,0.037868800759315493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,32,power_law_1.2,0.020300799608230592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,32,power_law_1.2,0.03795199990272522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,32,power_law_1.2,0.02069759964942932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,32,power_law_1.2,0.038431999087333676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,32,power_law_1.2,0.027980801463127137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,32,power_law_1.2,0.039059200882911684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,32,power_law_1.2,0.028358399868011475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,32,power_law_1.2,0.03991039991378784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,32,power_law_1.2,0.02797439992427826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,32,power_law_1.2,0.040300801396369934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,32,power_law_1.2,0.027046400308609008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,32,power_law_1.2,0.044147199392318724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,32,power_law_1.2,0.028652799129486085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,32,power_law_1.2,0.0455808013677597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,32,power_law_1.2,0.028755199909210206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,32,power_law_1.2,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,32,power_law_1.2,0.029747200012207032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,32,power_law_1.2,0.05141119956970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,32,power_law_1.2,0.03306879997253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,32,power_law_1.2,0.05459840297698974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,32,power_law_1.2,0.03537279963493347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,32,power_law_1.2,0.06228479743003845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,32,power_law_1.2,0.041510400176048276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,32,power_law_1.2,0.06751360297203064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,32,power_law_1.2,0.0447488009929657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,32,power_law_1.2,0.09029759764671326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,32,power_law_1.2,0.054604798555374146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,32,power_law_1.2,0.11552640199661254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,32,power_law_1.2,0.06230400204658508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,32,power_law_1.2,0.15483520030975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,32,power_law_1.2,0.09153280258178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,32,power_law_1.2,0.18695679903030396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,32,power_law_1.2,0.11329280138015747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,32,power_law_1.2,0.31471359729766846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,32,power_law_1.2,0.14742399454116822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,32,power_law_1.2,0.3523263931274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,32,power_law_1.2,0.19052159786224365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,32,power_law_1.2,0.47526397705078127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,32,power_law_1.2,0.2255295991897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,32,power_law_1.2,0.7286079883575439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,32,power_law_1.2,0.3610944032669067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,32,power_law_1.2,2.015443229675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,32,power_law_1.2,0.7565824031829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,16,balanced,0.02447466552257538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,16,balanced,0.025290665527184803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,16,balanced,0.025397333006064098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,16,balanced,0.02402666707833608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,16,balanced,0.026047999660174053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,16,balanced,0.02810666710138321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,16,power_law_1.2,0.02414720058441162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,16,balanced,0.028565332293510437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,16,power_law_1.2,0.021638399362564086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,16,balanced,0.028575999041398365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,16,power_law_1.2,0.021971200406551362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,16,balanced,0.027477333943049114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,16,power_law_1.2,0.023769600689411162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,16,balanced,0.02922666569550832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,16,power_law_1.2,0.02487040013074875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,16,balanced,0.028618666032950085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,16,power_law_1.2,0.026240000128746034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,16,balanced,0.027376001079877216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,16,power_law_1.2,0.026931199431419372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,16,balanced,0.029296000798543293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,16,power_law_1.2,0.029356798529624938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,16,balanced,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,16,power_law_1.2,0.02959359884262085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,16,balanced,0.0317493329445521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,16,balanced,0.0352960005402565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,16,power_law_1.2,0.03190400004386902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,16,balanced,0.03143999973932902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,16,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,16,power_law_1.2,0.03230080008506775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,16,balanced,0.029680001238981884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,16,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,16,power_law_1.2,0.03203200101852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,16,balanced,0.03153600047032038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,16,balanced,0.037647999823093414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,16,power_law_1.2,0.03234559893608093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,16,balanced,0.032101333141326904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,16,balanced,0.03982399900754293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,16,power_law_1.2,0.0316864013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,16,balanced,0.033626665671666466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,16,power_law_1.01,0.02555519938468933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,16,balanced,0.04192533095677694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,16,power_law_1.2,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,16,balanced,0.03369600077470144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,16,power_law_1.01,0.023321600258350374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,16,power_law_1.01,0.03747200071811676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,16,balanced,0.045941332976023354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,16,power_law_1.01,0.024607999622821806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,16,power_law_1.2,0.03596799969673157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,16,balanced,0.03269333392381668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,16,power_law_1.01,0.031231999397277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,16,power_law_1.01,0.02497279942035675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,16,balanced,0.05009066561857859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,16,power_law_1.2,0.03863039910793305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,16,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,16,power_law_1.01,0.02971520125865936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,16,power_law_1.01,0.02540160119533539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,16,balanced,0.053770666321118675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,16,power_law_1.2,0.04379520118236542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,16,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,16,power_law_1.01,0.031014400720596313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,16,power_law_1.01,0.027321600914001466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,16,balanced,0.06611200173695882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,16,power_law_1.2,0.046515199542045596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,16,balanced,0.034245334565639496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,16,power_law_1.01,0.03115519881248474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,16,power_law_1.01,0.027296000719070436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,16,power_law_1.2,0.05081599950790405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,16,power_law_1.01,0.02818560004234314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,16,balanced,0.03572800010442734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,16,balanced,0.08332799871762593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,16,power_law_1.01,0.031948798894882204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,16,power_law_1.2,0.05728639960289002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,16,power_law_1.01,0.030559998750686646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,16,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,16,power_law_1.01,0.03219200074672699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,16,power_law_1.01,0.03118720054626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,16,power_law_1.2,0.07479680180549622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,16,balanced,0.03701333453257879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,16,power_law_1.01,0.03294079899787903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,16,balanced,0.10570666193962097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,16,power_law_1.01,0.03185920119285583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,16,power_law_1.2,0.08760960102081299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,16,balanced,0.039861333866914116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,16,power_law_1.01,0.032876798510551454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,16,power_law_1.01,0.03265919983386993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,16,power_law_1.2,0.1117184042930603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,16,balanced,0.12754666805267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,16,balanced,0.039493332306543984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,16,power_law_1.01,0.032927998900413515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,16,power_law_1.01,0.0322816014289856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,16,power_law_1.2,0.1715008020401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,16,balanced,0.039877332746982574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,16,power_law_1.01,0.03550080060958862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,16,power_law_1.01,0.031795200705528257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,16,power_law_1.2,0.2423680067062378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,16,balanced,0.1486293375492096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,16,balanced,0.0429066667954127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,16,power_law_1.01,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,16,power_law_1.01,0.03468799889087677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,16,power_law_1.2,0.30271999835968016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,16,balanced,0.044250667095184326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,16,power_law_1.01,0.03585920035839081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,16,power_law_1.01,0.0346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,16,power_law_1.2,0.34202239513397215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,16,balanced,0.047728002071380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,16,balanced,0.20914665857950845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,16,power_law_1.01,0.037920001149177554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,16,power_law_1.01,0.03904640078544617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,16,power_law_1.2,0.664192008972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,16,balanced,0.053770666321118675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,16,power_law_1.01,0.0412416011095047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,16,power_law_1.01,0.041280001401901245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,16,power_law_1.2,1.2912192344665527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,16,balanced,0.06226666768391927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,16,power_law_1.01,0.04273920059204102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,16,power_law_1.01,0.045849600434303285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,16,power_law_1.01,0.04364160001277924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,16,power_law_1.2,0.036582401394844054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,16,power_law_1.01,0.046374401450157164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,16,balanced,0.0682239979505539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,16,power_law_1.01,0.050297600030899045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,16,power_law_1.2,0.027859199047088622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,16,power_law_1.01,0.05118079781532288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,16,power_law_1.01,0.05555840134620667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,16,power_law_1.2,0.02766079902648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,16,balanced,0.09335999687512715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,16,balanced,0.37928001085917157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,16,power_law_1.01,0.059334397315979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,16,power_law_1.01,0.06726400256156921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,16,power_law_1.2,0.029651200771331786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,16,power_law_1.01,0.06727039813995361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,16,balanced,0.12545599540074667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,16,power_law_1.01,0.07642239928245545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,16,power_law_1.2,0.030854400992393494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,16,power_law_1.01,0.08269439935684204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,16,power_law_1.01,0.10903040170669556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,16,power_law_1.2,0.03223040103912354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,16,power_law_1.01,0.09349759817123413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,16,power_law_1.01,0.12029440402984619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,16,power_law_1.2,0.031692799925804135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,16,balanced,0.1689866582552592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,16,power_law_1.01,0.14312959909439088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,16,power_law_1.01,0.17137919664382933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,16,power_law_1.2,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,16,power_law_1.01,0.17598079442977904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,16,power_law_1.01,0.20170240402221679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,16,balanced,0.21364800135294595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,16,power_law_1.2,0.03289600014686585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,16,power_law_1.01,0.2726464033126831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,16,power_law_1.01,0.2889024019241333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,16,power_law_1.2,0.03410559892654419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,16,power_law_1.01,0.3251136064529419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,16,power_law_1.01,0.4676032066345215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,16,power_law_1.2,0.03451519906520843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,16,balanced,0.25678400198618573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,16,power_law_1.01,0.3976703882217407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,16,power_law_1.01,0.7623487949371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,16,power_law_1.2,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,16,power_law_1.01,0.5969088077545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,16,power_law_1.2,0.036556801199913024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,16,power_law_1.01,1.2045056343078613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,16,power_law_1.2,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,16,power_law_1.2,0.04213759899139404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,16,power_law_1.2,0.04319359958171844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,16,power_law_1.2,0.04429439902305603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,16,balanced,0.3736639817555745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,16,power_law_1.2,0.04723840057849884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,16,power_law_1.2,0.052288001775741576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,16,power_law_1.2,0.06366080045700073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,16,power_law_1.2,0.07099519968032837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,16,power_law_1.2,0.09230719804763794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,16,power_law_1.2,0.11415040493011475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,16,power_law_1.2,0.16000640392303467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,16,power_law_1.2,0.21123199462890624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,16,power_law_1.2,0.3268415927886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,16,power_law_1.2,0.3960319995880127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,16,power_law_1.2,0.5414976119995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,16,power_law_1.2,0.9226304054260254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,16,balanced,0.7063413461049398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,16,power_law_1.2,1.995302391052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,8,balanced,0.03345066557327906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,8,balanced,0.037392000357309975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,8,balanced,0.031685332457224526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,8,balanced,0.031231999397277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,8,balanced,0.03140799949566523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,8,balanced,0.030762667457262676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,8,balanced,0.03124266614516576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,8,balanced,0.03155199935038885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,8,balanced,0.03309333324432373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,8,balanced,0.03239466746648153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,8,balanced,0.03025600065787633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,8,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,8,balanced,0.029530666768550873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,8,balanced,0.03141866624355316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,8,balanced,0.0314026673634847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,8,balanced,0.03339199970165888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,8,balanced,0.03155199935038885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,8,balanced,0.03164266546567281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,8,balanced,0.033530667424201965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,8,balanced,0.033413333197434746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,8,balanced,0.03133333226044973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,8,balanced,0.03346133232116699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,8,balanced,0.03154666721820831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,8,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,8,balanced,0.0334346666932106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,8,balanced,0.0378560001651446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,8,balanced,0.03183999905983607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,8,balanced,0.038191998998324074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,8,balanced,0.033626665671666466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,8,balanced,0.040175999204317726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,8,balanced,0.03348266581694285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,8,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,8,balanced,0.033717334270477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,8,balanced,0.045696000258127846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,8,balanced,0.03979733337958654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,8,balanced,0.0460746685663859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,8,balanced,0.038106667498747505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,8,balanced,0.06227200229962667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,8,balanced,0.03967999915281931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,8,balanced,0.06394133468468984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,8,balanced,0.04367466767628988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,8,balanced,0.0886346697807312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,8,balanced,0.04553600152333578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,8,balanced,0.10159466663996379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,8,balanced,0.051728000243504844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,8,balanced,0.01916266605257988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,8,balanced,0.05573866764704386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,8,balanced,0.14257599910100302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,8,balanced,0.018976000448067982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,8,balanced,0.07603733241558075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,8,balanced,0.15051733454068503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,8,balanced,0.01905599981546402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,8,balanced,0.09549333651860555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,8,balanced,0.01930133377512296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,8,balanced,0.13219199577967325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,8,balanced,0.02027200038234393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,8,balanced,0.1674720048904419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,8,balanced,0.1606559952100118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,8,balanced,0.02107200026512146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,8,balanced,0.03374933451414108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,8,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,8,balanced,0.2257279952367147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,8,balanced,0.2584266662597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,8,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,8,balanced,0.03957333415746689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,8,balanced,0.2746613423029582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,8,balanced,0.0354720006386439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,8,balanced,0.02991466720898946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,8,balanced,0.03403199960788091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,8,balanced,0.3322826623916626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,8,balanced,0.029882666965325672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,8,balanced,0.4530506531397502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,8,balanced,0.0354666660229365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,8,power_law_1.01,0.03343360126018524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,8,power_law_1.01,0.02922239899635315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,8,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,8,balanced,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,8,balanced,0.5041600068410238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,8,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,8,power_law_1.01,0.03341439962387085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,8,power_law_1.01,0.029465600848197937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,8,balanced,0.033359999457995095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,8,balanced,0.0354720006386439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,8,power_law_1.01,0.028255999088287354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,8,power_law_1.01,0.02876160144805908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,8,balanced,0.9575573603312174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,8,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,8,balanced,0.0354666660229365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,8,power_law_1.01,0.02969599962234497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,8,power_law_1.01,0.031411200761795044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,8,balanced,0.03179199993610382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,8,power_law_1.01,0.029996800422668456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,8,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,8,power_law_1.01,0.030246400833129884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,8,balanced,0.03355200091997782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,8,power_law_1.01,0.01764480024576187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,8,power_law_1.01,0.03237760066986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,8,power_law_1.01,0.030681601166725157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,8,balanced,0.03645866612593333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,8,power_law_1.01,0.031353598833084105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,8,balanced,0.03363200028737386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,8,power_law_1.01,0.01772159934043884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,8,power_law_1.01,0.03358719944953918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,8,power_law_1.01,0.03089280128479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,8,balanced,0.03819733361403147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,8,power_law_1.01,0.03201920092105866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,8,balanced,0.03356799980004629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,8,power_law_1.01,0.01759359985589981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,8,power_law_1.01,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,8,power_law_1.01,0.03112959861755371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,8,power_law_1.01,0.031916800141334536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,8,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,8,balanced,0.038506666819254555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,8,power_law_1.01,0.018764799833297728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,8,power_law_1.01,0.033843201398849485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,8,power_law_1.01,0.031014400720596313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,8,power_law_1.01,0.03243519961833954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,8,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,8,balanced,0.04031466692686081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,8,power_law_1.01,0.03426559865474701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,8,power_law_1.01,0.01797119975090027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,8,power_law_1.01,0.03165439963340759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,8,power_law_1.01,0.03243519961833954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,8,balanced,0.043738668163617454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,8,balanced,0.0558186670144399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,8,power_law_1.01,0.035071998834609985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,8,power_law_1.01,0.01995519995689392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,8,power_law_1.01,0.03234559893608093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,8,power_law_1.01,0.03332479894161224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,8,balanced,0.04109866668780645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,8,balanced,0.06169066826502482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,8,power_law_1.01,0.03506560027599335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,8,power_law_1.01,0.037088000774383546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,8,power_law_1.01,0.03630079925060272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,8,power_law_1.01,0.03544960021972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,8,power_law_1.01,0.03609600067138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,8,balanced,0.04371733466784159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,8,power_law_1.01,0.03592959940433502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,8,balanced,0.08194666604201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,8,power_law_1.01,0.037529599666595456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,8,power_law_1.01,0.03591679930686951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,8,power_law_1.01,0.03706879913806915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,8,balanced,0.044031997521718345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,8,power_law_1.01,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,8,balanced,0.09714133540789287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,8,power_law_1.01,0.02834559977054596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,8,power_law_1.01,0.036236798763275145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,8,power_law_1.01,0.03726719915866852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,8,balanced,0.046725332736968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,8,power_law_1.01,0.03675520122051239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,8,power_law_1.01,0.028825598955154418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,8,balanced,0.127893328666687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,8,power_law_1.01,0.04271360039710999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,8,power_law_1.01,0.044038400053977966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,8,balanced,0.05218133330345154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,8,power_law_1.01,0.03834240138530731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,8,power_law_1.01,0.029862400889396668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,8,power_law_1.01,0.04688639938831329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,8,balanced,0.16105066736539206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,8,power_law_1.01,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,8,balanced,0.05401599903901418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,8,power_law_1.01,0.03780480027198792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,8,power_law_1.01,0.03159680068492889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,8,power_law_1.01,0.04810880124568939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,8,power_law_1.01,0.0482367992401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,8,balanced,0.22456000248591104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,8,balanced,0.06449066599210103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,8,power_law_1.01,0.03803519904613495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,8,power_law_1.01,0.030976000428199767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,8,power_law_1.01,0.053427201509475705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,8,power_law_1.01,0.056806397438049314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,8,balanced,0.07539199789365132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,8,balanced,0.28860799471537274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,8,power_law_1.01,0.0438975989818573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,8,power_law_1.01,0.03051519989967346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,8,power_law_1.01,0.05916799902915955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,8,power_law_1.01,0.05975040197372437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,8,balanced,0.09901866316795349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,8,balanced,0.3527359962463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,8,power_law_1.01,0.044870400428771974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,8,power_law_1.01,0.03297280073165894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,8,power_law_1.01,0.06347519755363465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,8,power_law_1.01,0.0700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,8,balanced,0.11709333459536235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,8,power_law_1.01,0.04710400104522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,8,power_law_1.01,0.03175039887428284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,8,balanced,0.5441866715749105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,8,power_law_1.01,0.08117120265960694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,8,power_law_1.01,0.08731520175933838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,8,power_law_1.01,0.05345919728279114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,8,power_law_1.01,0.03285120129585266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,8,balanced,0.1535253326098124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,8,power_law_1.01,0.10755840539932252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,8,power_law_1.01,0.11301120519638061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.01,0.05894399881362915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,8,power_law_1.01,0.03837440013885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,8,balanced,1.0567999680836995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,8,power_law_1.01,0.12312959432601929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,8,balanced,0.18542933464050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,8,power_law_1.01,0.15574400424957274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.01,0.06604160070419311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.01,0.04049279987812042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,8,power_law_1.01,0.17674239873886108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,8,power_law_1.01,0.21127040386199952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.01,0.05377280116081238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.01,0.07752959728240967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,8,balanced,0.22106132904688516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,8,power_law_1.01,0.2162303924560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,8,power_law_1.01,0.2618112087249756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.01,0.05937920212745666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,8,balanced,0.32690133651097614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.01,0.09561600089073181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,8,power_law_1.01,0.3031167984008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,8,power_law_1.01,0.38293120861053465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.01,0.07958400249481201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.01,0.10880639553070068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,8,power_law_1.01,0.38410239219665526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,8,power_law_1.01,0.46798081398010255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,16,8,power_law_1.2,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.01,0.09546239972114563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,8,balanced,0.5980480114618937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.01,0.1620479941368103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,8,power_law_1.01,0.42369918823242186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,8,power_law_1.01,0.5833983898162842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,16,8,power_law_1.2,0.033504000306129454
TRTLLM,1.2.0rc5,NVIDIA B200,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 B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.01,0.20646400451660157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,8,power_law_1.01,0.6439167976379394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,8,power_law_1.01,0.9056511878967285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,16,8,power_law_1.2,0.0281792014837265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.01,0.15809919834136962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.01,0.27445120811462403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,8,power_law_1.01,1.3419648170471192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,8,power_law_1.01,1.6700735092163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,16,8,power_law_1.2,0.028166401386260986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,16,8,power_law_1.2,0.029945600032806396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.01,0.2201472043991089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.01,0.32406399250030515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,16,8,power_law_1.2,0.027724799513816834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,16,8,power_law_1.2,0.029926401376724244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.01,0.28586881160736083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,8,power_law_1.01,0.45852160453796387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,16,8,power_law_1.2,0.02844800055027008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,16,8,power_law_1.2,0.03086079955101013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,8,power_law_1.01,0.3484607934951782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,8,power_law_1.01,0.7227583885192871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,16,8,power_law_1.2,0.029683199524879456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,16,8,power_law_1.2,0.030777600407600404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,8,power_law_1.01,0.5421887874603272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,8,power_law_1.01,1.3390015602111816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,16,8,power_law_1.2,0.030419200658798218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,16,8,power_law_1.2,0.03118720054626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,8,power_law_1.01,1.0553152084350585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,16,8,power_law_1.2,0.030611199140548707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,16,8,power_law_1.2,0.03223679959774017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,16,8,power_law_1.2,0.03130879998207092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,16,8,power_law_1.2,0.031692799925804135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,16,8,power_law_1.2,0.031225600838661195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,16,8,power_law_1.2,0.03244799971580505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,16,8,power_law_1.2,0.03133440017700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,16,8,power_law_1.2,0.036160001158714296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,16,8,power_law_1.2,0.032678401470184325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,16,8,power_law_1.2,0.03553920090198517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,16,8,power_law_1.2,0.03343360126018524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,16,8,power_law_1.2,0.03740800023078918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,16,8,power_law_1.2,0.04410879909992218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,16,8,power_law_1.2,0.03362559974193573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,16,8,power_law_1.2,0.048044800758361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,16,8,power_law_1.2,0.03630079925060272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,16,8,power_law_1.2,0.04836480021476745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,16,8,power_law_1.2,0.036447998881340024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,16,8,power_law_1.2,0.055980801582336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,16,8,power_law_1.2,0.04279040098190308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,16,8,power_law_1.2,0.057068800926208495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,16,8,power_law_1.2,0.04682239890098572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,16,8,power_law_1.2,0.07092480063438415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,16,8,power_law_1.2,0.04969600141048432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,16,8,power_law_1.2,0.08713600039482117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,16,8,power_law_1.2,0.058143997192382814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,16,8,power_law_1.2,0.10784640312194824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,16,8,power_law_1.2,0.05829120278358459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,16,8,power_law_1.2,0.13296639919281006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,16,8,power_law_1.2,0.07395200133323669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,16,8,power_law_1.2,0.1764799952507019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,16,8,power_law_1.2,0.08471680283546448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,16,8,power_law_1.2,0.22960000038146972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,16,8,power_law_1.2,0.11959680318832397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,16,8,power_law_1.2,0.3017983913421631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,16,8,power_law_1.2,0.1545215964317322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,16,8,power_law_1.2,0.36227200031280515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,16,8,power_law_1.2,0.2054975986480713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,16,8,power_law_1.2,0.4561471939086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,16,8,power_law_1.2,0.26929919719696044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,16,8,power_law_1.2,0.6741824150085449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,16,8,power_law_1.2,0.3699712038040161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,16,8,power_law_1.2,1.285548782348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,16,8,power_law_1.2,0.483244800567627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,16,8,power_law_1.2,0.6087103843688965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,16,8,power_law_1.2,0.9034815788269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,16,8,power_law_1.2,1.7838079452514648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,14336,2,8,16,8,power_law_1.2,0.032576000690460204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,14336,2,8,16,8,power_law_1.2,0.03331199884414673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,14336,2,8,16,8,power_law_1.2,0.03285120129585266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,14336,2,8,16,8,power_law_1.2,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,14336,2,8,16,8,power_law_1.2,0.03484799861907959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,14336,2,8,16,8,power_law_1.2,0.034860798716545106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,14336,2,8,16,8,power_law_1.2,0.03501439988613129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,14336,2,8,16,8,power_law_1.2,0.035257598757743834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,14336,2,8,16,8,power_law_1.2,0.03663359880447388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,14336,2,8,16,8,power_law_1.2,0.036339199542999266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,14336,2,8,16,8,power_law_1.2,0.036160001158714296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,14336,2,8,16,8,power_law_1.2,0.037964800000190736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,14336,2,8,16,8,power_law_1.2,0.03784320056438446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,14336,2,8,16,8,power_law_1.2,0.03775359988212586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,14336,2,8,16,8,power_law_1.2,0.04348160028457641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,14336,2,8,16,8,power_law_1.2,0.016979199647903443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,14336,2,8,16,8,power_law_1.2,0.045561599731445315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,14336,2,8,16,8,power_law_1.2,0.04699519872665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,14336,2,8,16,8,power_law_1.2,0.017343999445438386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,14336,2,8,16,8,power_law_1.2,0.05368319749832153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,14336,2,8,16,8,power_law_1.2,0.01733759939670563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.2,0.058220797777175905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,14336,2,8,16,8,power_law_1.2,0.017343999445438386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.2,0.06967679858207702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,14336,2,8,16,8,power_law_1.2,0.01820160001516342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.2,0.07589120268821717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,14336,2,8,16,8,power_law_1.2,0.019948799908161164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.2,0.10062079429626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,14336,2,8,16,8,power_law_1.2,0.037088000774383546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.2,0.12145919799804687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,14336,2,8,16,8,power_law_1.2,0.037561601400375365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.2,0.16792320013046264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,14336,2,8,16,8,power_law_1.2,0.028147199749946596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.2,0.2001471996307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,14336,2,8,16,8,power_law_1.2,0.0286655992269516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.2,0.27301759719848634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,14336,2,8,16,8,power_law_1.2,0.028428798913955687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.2,0.3752768039703369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,14336,2,8,16,8,power_law_1.2,0.030905601382255555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,14336,2,8,16,8,power_law_1.2,0.4659008026123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,14336,2,8,16,8,power_law_1.2,0.03057279884815216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,14336,2,8,16,8,power_law_1.2,0.6573567867279053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,14336,2,8,16,8,power_law_1.2,0.030214399099349976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,14336,2,8,16,8,power_law_1.2,1.2484095573425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,14336,2,8,16,8,power_law_1.2,0.03209599852561951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,14336,2,8,16,8,power_law_1.2,0.032179200649261476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,14336,2,8,16,8,power_law_1.2,0.032902398705482484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,14336,2,8,16,8,power_law_1.2,0.03731200098991394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.2,0.04043520092964172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.2,0.05322880148887634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.2,0.05926399827003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.2,0.07983999848365783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.2,0.09523839950561523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.2,0.125491201877594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.2,0.1588863968849182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.2,0.22183680534362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.2,0.28499839305877683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,14336,2,8,16,8,power_law_1.2,0.3496896028518677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,14336,2,8,16,8,power_law_1.2,0.5420928001403809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,14336,2,8,16,8,power_law_1.2,1.0566783905029298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,4,balanced,0.03446399917205175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,4,balanced,0.02995733420054118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,4,balanced,0.029663999875386555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,4,balanced,0.031285333136717476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,4,balanced,0.031717332700888314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,4,balanced,0.03136533250411352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,4,power_law_1.2,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,4,balanced,0.030608000854651134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,4,balanced,0.031471999982992806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,4,power_law_1.2,0.02800000011920929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,4,balanced,0.0317546675602595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,4,balanced,0.03339199970165888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,4,power_law_1.2,0.02805120050907135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,4,balanced,0.03154666721820831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,4,power_law_1.2,0.02937600016593933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,4,balanced,0.033610666791598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,4,power_law_1.2,0.029702401161193846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,4,balanced,0.03183999905983607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,4,power_law_1.2,0.028364801406860353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,4,balanced,0.031770666440327965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,4,power_law_1.2,0.030239999294281006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,4,balanced,0.031658666829268135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,4,power_law_1.01,0.031590399146080014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,4,power_law_1.2,0.028947201371192933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,4,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,4,power_law_1.2,0.03073279857635498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,4,balanced,0.03344533344109853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,4,balanced,0.033770665526390076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,4,power_law_1.01,0.03127039968967438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,4,power_law_1.01,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,4,power_law_1.2,0.030118399858474733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,4,power_law_1.2,0.03078399896621704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,4,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,4,balanced,0.033743999898433685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,4,power_law_1.01,0.029631999135017396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,4,power_law_1.01,0.030169600248336793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,4,power_law_1.2,0.03078399896621704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,4,power_law_1.2,0.030739200115203858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,4,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,4,balanced,0.03350399931271871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,4,power_law_1.01,0.029657599329948426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,4,power_law_1.01,0.030124801397323608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,4,power_law_1.2,0.03144960105419159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,4,power_law_1.2,0.03086720108985901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,4,balanced,0.03985599925120672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,4,balanced,0.03380800038576126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,4,power_law_1.01,0.029747200012207032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,4,power_law_1.01,0.03084160089492798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,4,power_law_1.2,0.03247359991073608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,4,power_law_1.2,0.03189760148525238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,4,balanced,0.039919999738534294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,4,balanced,0.03566399961709976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,4,power_law_1.01,0.030553600192070006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,4,power_law_1.01,0.03128960132598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,4,power_law_1.2,0.032601600885391234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,4,power_law_1.2,0.03283199965953827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,4,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,4,balanced,0.03475733349720637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,4,power_law_1.01,0.030425599217414855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,4,power_law_1.01,0.03207040131092072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,4,power_law_1.2,0.03311359882354736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,4,power_law_1.2,0.033452799916267394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,4,balanced,0.04404800136884054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,4,balanced,0.03548266738653183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,4,power_law_1.01,0.030943998694419862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,4,power_law_1.01,0.03213439881801605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,4,power_law_1.2,0.03406080007553101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,4,power_law_1.2,0.03351039886474609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,4,balanced,0.04399466514587402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,4,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,4,power_law_1.01,0.032313600182533264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,4,power_law_1.01,0.03357439935207367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,4,power_law_1.2,0.034771201014518735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,4,power_law_1.2,0.035776001214981076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,4,balanced,0.04996266464392344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,4,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,4,power_law_1.01,0.031788799166679385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,4,power_law_1.01,0.03375999927520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,4,power_law_1.2,0.035097599029541016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,4,power_law_1.2,0.04094719886779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,4,balanced,0.05208000044027964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,4,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,4,power_law_1.01,0.032518398761749265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,4,power_law_1.01,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,4,power_law_1.2,0.036646398901939395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,4,power_law_1.2,0.0440447986125946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,4,balanced,0.07038400073846181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,4,balanced,0.0400693342089653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,4,power_law_1.01,0.03318400084972382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,4,power_law_1.01,0.03497599959373474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,4,power_law_1.2,0.0379584014415741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,4,power_law_1.2,0.052275198698043826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,4,balanced,0.08130666613578796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,4,balanced,0.04154666761557261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,4,power_law_1.01,0.033452799916267394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,4,power_law_1.01,0.03583360016345978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,4,power_law_1.2,0.04382080137729645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,4,power_law_1.2,0.05811840295791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,4,balanced,0.11086400349934895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,4,balanced,0.04637333254019419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,4,power_law_1.01,0.03501439988613129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,4,power_law_1.01,0.036774399876594546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,4,power_law_1.2,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,4,power_law_1.2,0.05878400206565857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,4,balanced,0.0499946673711141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,4,balanced,0.12507733702659607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,4,power_law_1.2,0.05203199982643127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,4,power_law_1.2,0.07281280159950257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,4,balanced,0.05588266750176748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,4,power_law_1.01,0.04069760143756866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,4,power_law_1.01,0.03854719996452331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,4,power_law_1.2,0.057004797458648684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,4,power_law_1.2,0.08205440044403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,4,balanced,0.1661066710948944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,4,balanced,0.06271466612815857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,4,power_law_1.01,0.045126399397850035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,4,power_law_1.01,0.04456959962844849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,4,power_law_1.2,0.05802239775657654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,4,power_law_1.2,0.10338560342788697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,4,balanced,0.08692266543706258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,4,power_law_1.01,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,4,power_law_1.01,0.046623998880386354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,4,power_law_1.2,0.07648000121116638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,4,balanced,0.19614932934443155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,4,power_law_1.2,0.1425984025001526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,4,balanced,0.11364266276359558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,4,power_law_1.01,0.05482239723205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,4,balanced,0.24423466126124063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,4,power_law_1.01,0.04873600006103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,4,power_law_1.2,0.09955840110778809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,4,power_law_1.2,0.19566719532012938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,4,balanced,0.15550399820009866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,4,power_law_1.01,0.05707520246505737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,4,power_law_1.01,0.06160640120506287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,4,power_law_1.2,0.1322368025779724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,4,power_law_1.2,0.23326079845428466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,4,power_law_1.01,0.07534080147743225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,4,balanced,0.34707732995351154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,4,balanced,0.19129600127538046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,4,power_law_1.01,0.05936639904975891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,4,power_law_1.2,0.176147198677063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,4,power_law_1.2,0.3719360113143921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,4,power_law_1.01,0.08260480165481568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,4,power_law_1.01,0.07637760043144226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,4,balanced,0.26501333713531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,4,power_law_1.2,0.2407167911529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,4,power_law_1.2,0.38047358989715574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,4,power_law_1.01,0.1053056001663208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,4,power_law_1.01,0.0881663978099823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,4,balanced,0.6223680178324381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,4,power_law_1.2,0.31381759643554685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,4,balanced,0.3254400094350179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,4,power_law_1.2,0.553926420211792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,8,balanced,0.04171733558177948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,4,power_law_1.01,0.12743040323257446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,4,power_law_1.01,0.11319680213928222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,4,power_law_1.2,0.46530561447143554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,8,balanced,0.04178133110205332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,4,power_law_1.2,0.8526975631713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,4,power_law_1.01,0.1626431941986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,4,power_law_1.01,0.15624959468841554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,8,balanced,0.033514666060606636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,4,balanced,0.40092265605926514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,8,balanced,0.03411199897527695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,4,power_law_1.2,0.5493504047393799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,4,power_law_1.2,1.6499584197998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,4,power_law_1.01,0.2279360055923462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,4,power_law_1.01,0.23500800132751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,8,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,8,balanced,0.03507733345031738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,4,balanced,0.6053226788838705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,4,power_law_1.2,0.638431978225708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,4,power_law_1.01,0.3535295963287354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,4,power_law_1.01,0.2567807912826538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,8,balanced,0.03180799881617228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,8,balanced,0.03376533339420954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,4,power_law_1.2,1.1360768318176269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,4,power_law_1.01,0.38645119667053224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,4,power_law_1.01,0.4310592174530029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,8,balanced,0.031898667414983116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,8,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,4,balanced,1.1604053179423015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,4,power_law_1.2,2.3860095977783202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,4,power_law_1.01,0.5792640209197998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,4,power_law_1.01,0.4288832187652588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,8,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,8,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,4,power_law_1.01,0.6167232036590576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,8,balanced,0.033759998778502144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,4,power_law_1.01,0.7875199794769288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,8,balanced,0.03344533344109853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,8,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,4,power_law_1.01,1.0536319732666015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,8,balanced,0.03594133257865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,8,balanced,0.03551999976237615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,4,power_law_1.01,1.4048959732055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,4,power_law_1.01,2.0296255111694337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,8,balanced,0.03339199970165888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,8,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,8,balanced,0.035818666219711304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,8,balanced,0.03379733363787333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,8,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,8,balanced,0.036415999134381614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,8,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,8,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,8,balanced,0.04179200033346812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,8,balanced,0.03588266670703888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,8,balanced,0.04179200033346812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,8,balanced,0.03513066718975703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,8,balanced,0.04213866591453552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,8,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,8,balanced,0.045909335215886436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,8,balanced,0.043791999419530235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,8,balanced,0.048842668533325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,8,balanced,0.041696002086003624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,8,balanced,0.052815998593966164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,8,balanced,0.04422933359940847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,8,balanced,0.06235733131567637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,8,balanced,0.04822400212287903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,8,balanced,0.08333333333333333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,8,balanced,0.05193066596984863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,8,balanced,0.08430400490760803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,8,balanced,0.05816000203291575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,8,balanced,0.11188266674677531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,8,balanced,0.076773335536321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,8,balanced,0.10950932900110881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,8,balanced,0.12893866499265036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,8,balanced,0.12946133812268576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,8,balanced,0.16750399271647134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,8,balanced,0.17523199319839478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,8,balanced,0.20466132958730063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,8,balanced,0.21366933981577554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,8,balanced,0.25459200143814087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,8,balanced,0.30461867650349933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,8,balanced,0.34984532992045086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,8,balanced,0.38206398487091064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,8,balanced,0.470634659131368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,8,balanced,0.6197066704432169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,8,balanced,0.7079359690348307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,8,balanced,1.350586732228597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,8,balanced,0.03756266583998998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,8,balanced,0.0379573330283165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,8,balanced,0.03748800108830134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,8,balanced,0.035786665976047516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,8,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,8,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,8,balanced,0.037647999823093414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,8,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,8,balanced,0.01915733392039935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,8,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,8,balanced,0.019365333020687103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,8,balanced,0.039749334255854286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,8,balanced,0.021045332153638203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,8,balanced,0.03798400113979975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,8,balanced,0.019472000499566395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,8,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,8,balanced,0.019823999454577763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,8,balanced,0.03955733279387156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,8,balanced,0.021146667500336964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,8,balanced,0.039706667264302574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,8,balanced,0.021269333859284718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,8,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,8,balanced,0.02350933353106181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,8,balanced,0.04367466767628988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,8,balanced,0.025429333249727886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,8,balanced,0.04578666885693868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,8,balanced,0.02534399926662445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,8,power_law_1.01,0.038489601016044615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,8,power_law_1.01,0.030847999453544616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,8,balanced,0.04799466828505198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,8,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,8,power_law_1.01,0.018918399512767792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,8,power_law_1.01,0.03449600040912628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,8,power_law_1.01,0.038726401329040525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,8,power_law_1.01,0.030585598945617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,8,balanced,0.05204799771308899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,8,power_law_1.01,0.030937600135803222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,8,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,8,power_law_1.01,0.0341376006603241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,8,power_law_1.01,0.018719999492168425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,8,balanced,0.05421866476535797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,8,power_law_1.01,0.03028480112552643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,8,power_law_1.01,0.032927998900413515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,8,balanced,0.03619199991226196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,8,power_law_1.01,0.034457600116729735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,8,power_law_1.01,0.019270400702953338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,8,power_law_1.01,0.032652801275253295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,8,power_law_1.01,0.03206399977207184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,8,balanced,0.06418666740258534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,8,balanced,0.03659733384847641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,8,power_law_1.01,0.03573119938373566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,8,power_law_1.01,0.01884160041809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,16,8,power_law_1.2,0.03916800022125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,8,power_law_1.01,0.03296639919281006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,8,power_law_1.01,0.03313280045986176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,8,balanced,0.08453866839408875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,8,balanced,0.03955200066169103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,8,power_law_1.01,0.036236798763275145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,8,power_law_1.01,0.019487999379634857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,8,power_law_1.01,0.03368319869041443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,16,8,power_law_1.2,0.038771200180053714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,8,power_law_1.01,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,8,balanced,0.038848000268141426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,8,balanced,0.09195199608802795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,8,power_law_1.01,0.036294400691986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,8,power_law_1.01,0.021830399334430695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,8,power_law_1.01,0.033267199993133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,16,8,power_law_1.2,0.0318336009979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,8,power_law_1.01,0.0345984011888504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,8,balanced,0.041536000867684685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,8,power_law_1.01,0.03694719970226288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,8,power_law_1.01,0.0456063985824585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,8,power_law_1.01,0.03461120128631592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,16,8,power_law_1.2,0.033030399680137636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,8,power_law_1.01,0.03462400138378143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,8,balanced,0.12069867054621379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,8,balanced,0.04433066646258036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,8,power_law_1.01,0.03711360096931458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,8,power_law_1.01,0.037363201379776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,8,power_law_1.01,0.04559360146522522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,16,8,power_law_1.2,0.03326080143451691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,8,power_law_1.01,0.03509120047092438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,8,power_law_1.01,0.03903360068798065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,8,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,8,balanced,0.14269333084424338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,16,8,power_law_1.2,0.02998400032520294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,8,power_law_1.01,0.037574398517608645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,8,power_law_1.01,0.03185279965400696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,16,8,power_law_1.2,0.033081600069999696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,8,power_law_1.01,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,8,power_law_1.01,0.03891839981079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,8,balanced,0.06786666810512543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,8,balanced,0.19278399149576822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,8,power_law_1.01,0.03783040046691895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,8,power_law_1.01,0.03206399977207184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,16,8,power_law_1.2,0.0293503999710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,16,8,power_law_1.2,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,8,power_law_1.01,0.03648000061511993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,8,power_law_1.01,0.04015359878540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,8,balanced,0.07669866581757863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,8,balanced,0.24109333753585815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,8,power_law_1.01,0.052121597528457644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,8,power_law_1.01,0.036774399876594546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,8,power_law_1.01,0.038675200939178464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,8,power_law_1.01,0.03407999873161316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,16,8,power_law_1.2,0.030214399099349976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,16,8,power_law_1.2,0.03421440124511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,8,balanced,0.10130133231480916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,8,power_law_1.01,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,8,balanced,0.2902933359146118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,8,power_law_1.01,0.03914879858493805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,8,power_law_1.01,0.03951359987258911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,8,power_law_1.01,0.034835198521614076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,16,8,power_law_1.2,0.030880001187324525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,16,8,power_law_1.2,0.03464959859848023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,8,balanced,0.1269493301709493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,8,power_law_1.01,0.059724801778793336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,8,power_law_1.01,0.03932160139083862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,8,power_law_1.01,0.04056960046291351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,8,balanced,0.42394665877024335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,8,power_law_1.01,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,16,8,power_law_1.2,0.0321152001619339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,16,8,power_law_1.2,0.036550399661064145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,8,power_law_1.01,0.07166079878807068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,8,balanced,0.1739199956258138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,8,power_law_1.01,0.040966400504112245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,8,power_law_1.01,0.042182400822639465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,8,power_law_1.01,0.035257598757743834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,8,power_law_1.01,0.07340800166130065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,16,8,power_law_1.2,0.038227200508117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,16,8,power_law_1.2,0.032671999931335446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,8,power_law_1.01,0.0516543984413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,8,balanced,0.7942079703013102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,8,power_law_1.01,0.047142401337623596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,8,balanced,0.212501327196757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,8,power_law_1.01,0.03723520040512085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,8,power_law_1.01,0.09188479781150818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,16,8,power_law_1.2,0.03438720107078552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,16,8,power_law_1.2,0.03898879885673523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,8,power_law_1.01,0.05198720097541809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,8,power_law_1.01,0.04737919867038727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,8,balanced,0.3088746666908264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,8,power_law_1.01,0.12021759748458863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,8,power_law_1.01,0.03716480135917664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,16,8,power_law_1.2,0.034143999218940735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,8,power_law_1.01,0.05441920161247253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,16,8,power_law_1.2,0.039705601334571836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,8,power_law_1.01,0.169868803024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,8,power_law_1.01,0.05425919890403748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,8,power_law_1.01,0.03948799967765808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,8,balanced,0.3953973452250163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,16,8,power_law_1.2,0.03549439907073974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,8,power_law_1.01,0.06452479958534241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,16,8,power_law_1.2,0.04267520010471344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,8,power_law_1.01,0.165555202960968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,8,power_law_1.01,0.06232960224151611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,8,power_law_1.01,0.04352000057697296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,16,8,power_law_1.2,0.035820800065994265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,8,power_law_1.01,0.07002879977226258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,8,power_law_1.01,0.2503232002258301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,16,8,power_law_1.2,0.050521600246429446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.01,0.0654911994934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.01,0.046188798546791074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,16,8,power_law_1.2,0.03680639863014221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,8,power_law_1.01,0.2879103899002075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,8,power_law_1.01,0.09056000113487243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,8,balanced,0.4917440017064412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.01,0.07788159847259521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.01,0.06616320013999939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,8,power_law_1.01,0.42406401634216306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,16,8,power_law_1.2,0.03863680064678192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,16,8,power_law_1.2,0.055769598484039305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,8,power_law_1.01,0.12478079795837402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,8,power_law_1.01,0.5030335903167724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.01,0.09358720183372497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.01,0.07464320063591004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,16,8,power_law_1.2,0.03776000142097473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,8,power_law_1.01,0.6084735870361329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,8,power_law_1.01,0.15707520246505738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,16,8,power_law_1.2,0.05928320288658142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,8,balanced,0.7635680039723715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.01,0.12629760503768922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.01,0.09880959987640381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,16,8,power_law_1.2,0.04281600117683411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,8,power_law_1.01,1.026912021636963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,8,power_law_1.01,0.20312960147857667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,16,8,power_law_1.2,0.0728767991065979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.01,0.1472640037536621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.01,0.12412799596786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,16,8,power_law_1.2,0.050348800420761106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,8,power_law_1.01,1.8202367782592774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,8,power_law_1.01,0.27429120540618895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,8,balanced,1.485904057820638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,16,8,power_law_1.2,0.07367039918899536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.01,0.20706560611724853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.01,0.17107199430465697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,16,8,power_law_1.2,0.05308160185813904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,8,power_law_1.01,0.35024640560150144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,16,8,power_law_1.2,0.0976639986038208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.01,0.2536639928817749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.01,0.20979840755462648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,16,8,power_law_1.2,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,8,power_law_1.01,0.5141632080078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,16,8,power_law_1.2,0.11709439754486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.01,0.370796799659729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.01,0.30589439868927004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,16,8,power_law_1.2,0.0656063973903656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,16,8,power_law_1.2,0.17280639410018922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.01,0.4809408187866211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,8,power_law_1.01,0.6315328121185303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.01,0.3935296058654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,16,8,power_law_1.2,0.06840320229530335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,16,8,power_law_1.2,0.18042240142822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,8,power_law_1.01,0.5467135906219482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,8,power_law_1.01,0.7517824172973633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,8,power_law_1.01,0.48942079544067385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,16,8,power_law_1.2,0.09712640047073365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,16,8,power_law_1.2,0.24158079624176027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,8,power_law_1.01,0.8897727966308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,8,power_law_1.01,1.2300671577453612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,8,power_law_1.01,0.7592383861541748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,16,8,power_law_1.2,0.12607359886169434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,8,power_law_1.01,1.7560575485229493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,16,8,power_law_1.2,0.29998719692230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,8,power_law_1.01,2.402739143371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,16,8,power_law_1.2,0.16636159420013427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,8,power_law_1.01,1.483942413330078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,16,8,power_law_1.2,0.03444480001926422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,16,8,power_law_1.2,0.42739200592041016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,16,8,power_law_1.2,0.19856640100479125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,16,8,power_law_1.2,0.0342848002910614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,16,8,power_law_1.2,0.5222784042358398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,16,8,power_law_1.2,0.26810240745544434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,16,8,power_law_1.2,0.03432320058345795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,16,8,power_law_1.2,0.3457279920578003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,16,8,power_law_1.2,0.6185599803924561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,16,8,power_law_1.2,0.035641598701477054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,16,8,power_law_1.2,0.519539213180542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,16,8,power_law_1.2,1.0317952156066894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,16,8,power_law_1.2,0.0358271986246109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,16,8,power_law_1.2,0.6408703804016114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,16,8,power_law_1.2,1.8753215789794921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,16,8,power_law_1.2,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,16,8,power_law_1.2,0.7824512004852295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,16,8,power_law_1.2,0.03705599904060364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,16,8,power_law_1.2,1.2639039993286132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,16,8,power_law_1.2,0.037011200189590455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,16,8,power_law_1.2,2.4656768798828126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,16,8,power_law_1.2,0.037363201379776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,16,8,power_law_1.2,0.03818880021572113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,16,8,power_law_1.2,0.038431999087333676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,16,8,power_law_1.2,0.039724799990653994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,16,8,power_law_1.2,0.04042240083217621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,16,8,power_law_1.2,0.04252800047397613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,16,8,power_law_1.2,0.0469184011220932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,16,8,power_law_1.2,0.04737919867038727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,16,8,power_law_1.2,0.053651201725006106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,16,8,power_law_1.2,0.0639360010623932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.2,0.06716799736022949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.2,0.07912960052490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.2,0.0933247983455658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.2,0.1285055994987488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.2,0.15368959903717042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.2,0.199891197681427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.2,0.2651648044586182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.2,0.377676796913147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.2,0.4618879795074463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,16,8,power_law_1.2,0.5614912033081054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,16,8,power_law_1.2,0.9064512252807617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,16,8,power_law_1.2,1.755571174621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,16,8,power_law_1.2,0.0190528005361557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,16,8,power_law_1.2,0.019200000166893005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,16,8,power_law_1.2,0.019167999923229217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,16,8,power_law_1.2,0.018636800348758698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,16,8,power_law_1.2,0.01932159960269928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,16,8,power_law_1.2,0.021305599808692934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,16,8,power_law_1.2,0.044998401403427125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,16,8,power_law_1.2,0.044659200310707095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,16,8,power_law_1.2,0.03192960023880005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,16,8,power_law_1.2,0.03238399922847748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,16,8,power_law_1.2,0.03416320085525513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,16,8,power_law_1.2,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,16,8,power_law_1.2,0.03550719916820526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,16,8,power_law_1.2,0.03496319949626923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,16,8,power_law_1.2,0.037036800384521486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,16,8,power_law_1.2,0.03694080114364624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,16,8,power_law_1.2,0.04099839925765991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,16,8,power_law_1.2,0.04270719885826111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.2,0.04588159918785095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.2,0.06554880142211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.2,0.07432320117950439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.2,0.0998528003692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.2,0.12341120243072509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.2,0.17091200351715088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.2,0.20976641178131103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.2,0.30551040172576904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.2,0.3930367946624756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,16,8,power_law_1.2,0.4884160041809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,16,8,power_law_1.2,0.7583360195159912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,16,8,power_law_1.2,1.4816896438598632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,4,balanced,0.019039999693632126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,4,balanced,0.01922133316596349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,4,balanced,0.01931200052301089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,4,balanced,0.019365333020687103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,4,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,4,balanced,0.019445333629846573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,4,balanced,0.03833599885304769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,4,balanced,0.02094399929046631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,4,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,4,balanced,0.021205333371957142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,4,balanced,0.03739733248949051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,4,balanced,0.02128000060717265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,4,power_law_1.01,0.0394239991903305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,4,balanced,0.03772799919048945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,4,balanced,0.022090665996074677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,4,balanced,0.037871999045213066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,4,power_law_1.01,0.039340800046920775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,4,balanced,0.023178666830062866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,4,balanced,0.03938666731119156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,4,power_law_1.01,0.03941119909286499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,4,balanced,0.02515200028816859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,4,balanced,0.03900266687075297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,4,power_law_1.01,0.04131839871406555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,4,balanced,0.02555199960867564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,4,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,4,power_law_1.01,0.04197759926319122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,4,balanced,0.02717333287000656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,4,balanced,0.03961600114901861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,4,power_law_1.01,0.04291200041770935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,4,balanced,0.035573333501815796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,4,balanced,0.03931200007597605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,4,power_law_1.01,0.04202240109443665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,4,balanced,0.03728533287843069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,4,balanced,0.04012266546487808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,4,power_law_1.01,0.04287999868392944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,4,balanced,0.037632000943024956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,4,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,4,power_law_1.01,0.04289920032024384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,4,balanced,0.03997333347797394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,4,balanced,0.041509332756201424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,4,power_law_1.01,0.04238080084323883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,4,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,4,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,4,power_law_1.01,0.04318079948425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,4,balanced,0.046570668617884316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,4,balanced,0.047872001926104225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,4,power_law_1.01,0.04471679925918579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,4,balanced,0.05507733424504598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,4,balanced,0.04796266555786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,4,power_law_1.01,0.044921600818634035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,4,balanced,0.0643146683772405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,4,balanced,0.052015999952952065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,4,power_law_1.01,0.04524160027503967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,4,balanced,0.08885332942008972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,4,balanced,0.05622399846712748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,4,power_law_1.01,0.05061119794845581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,4,balanced,0.06084266801675161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,4,balanced,0.10177600383758545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,4,power_law_1.01,0.051283198595047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,4,balanced,0.07149333258469899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,4,power_law_1.01,0.053504002094268796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,4,balanced,0.0888320008913676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,4,balanced,0.13797866304715475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,4,balanced,0.04354133208592733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,4,power_law_1.01,0.06207360029220581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,4,balanced,0.11348799864451091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,4,balanced,0.03336533407370249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,4,balanced,0.17468800147374472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.01,0.07078400254249573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,4,balanced,0.03350399931271871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,4,balanced,0.14750400185585022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,4,balanced,0.24672534068425497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.01,0.08948479890823365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,4,balanced,0.03369066615899404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,4,balanced,0.1739573280016581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,4,balanced,0.30798933903376263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.01,0.10220160484313964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,4,balanced,0.03363200028737386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,4,balanced,0.2405973275502523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.01,0.13904639482498168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,4,balanced,0.3803199927012126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,4,balanced,0.033973333736260734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.01,0.15781760215759277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,4,balanced,0.30343466997146606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,4,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.01,0.24785919189453126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,4,balanced,0.5879306793212891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,4,balanced,0.03557866563399633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.01,0.2949120044708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,4,balanced,0.3677866856257121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,4,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,4,balanced,0.03551999976237615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.01,0.434771203994751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,4,balanced,0.034832000732421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,4,power_law_1.01,0.03978239893913269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,4,balanced,0.035530666510264076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,4,balanced,0.033728001018365227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.01,0.5248191833496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,4,balanced,1.1463946501413982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,4,power_law_1.01,0.03468799889087677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,4,balanced,0.5595680077870687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,4,balanced,0.035786665976047516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,4,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,4,power_law_1.01,0.663212776184082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,4,power_law_1.01,0.032364800572395325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,4,power_law_1.01,0.03115519881248474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,4,balanced,0.035717333356539406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,4,balanced,0.033813332517941795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,4,power_law_1.01,1.0163968086242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,4,power_law_1.01,0.03315840065479279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,4,balanced,1.0579520066579182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,4,balanced,0.03565333286921183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,4,balanced,0.03714133302370707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,4,power_law_1.01,0.03192960023880005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,4,power_law_1.01,2.206617546081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,4,power_law_1.01,0.033190399408340454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,4,balanced,0.03367999941110611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,4,balanced,0.03808533400297165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,4,power_law_1.01,0.031033599376678468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,4,power_law_1.01,0.03369599878787995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,4,balanced,0.03585066646337509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,4,balanced,0.04190400242805481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,4,power_law_1.01,0.031744000315666196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,4,power_law_1.01,0.03371520042419433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,4,balanced,0.035605333745479584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,4,power_law_1.01,0.03514240086078644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,4,balanced,0.04363200068473816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,4,power_law_1.01,0.03239679932594299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,4,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,4,power_law_1.01,0.018508799374103546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,4,power_law_1.01,0.0356799989938736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,4,balanced,0.044010668992996216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,4,power_law_1.01,0.01834239959716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,4,power_law_1.01,0.03306879997253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,4,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,4,power_law_1.01,0.037241598963737486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,4,power_law_1.01,0.034092798829078674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,4,balanced,0.05182399849096934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,4,power_law_1.01,0.018086400628089905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,4,balanced,0.037621334195137024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,4,power_law_1.01,0.036627200245857236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,4,power_law_1.01,0.03446399867534637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,4,balanced,0.0518453319867452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,4,power_law_1.01,0.018675200641155243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,4,balanced,0.03967999915281931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,4,power_law_1.01,0.03746559917926788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,4,power_law_1.01,0.035417601466178894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,4,balanced,0.06132799883683523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,4,power_law_1.01,0.01870719939470291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,4,balanced,0.0377813329299291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,4,power_law_1.01,0.03899520039558411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,4,balanced,0.06910400092601776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,4,power_law_1.01,0.035648000240325925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,4,power_law_1.01,0.019827200472354888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,4,balanced,0.0455626646677653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,4,power_law_1.01,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,4,balanced,0.0888320008913676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,4,power_law_1.01,0.03609600067138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,4,power_law_1.01,0.02024960070848465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,4,balanced,0.045696000258127846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,4,power_law_1.01,0.04572800099849701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,4,balanced,0.10337066650390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,4,power_law_1.01,0.03819519877433777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,4,power_law_1.01,0.020268799364566804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,4,balanced,0.04560533165931702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,4,power_law_1.01,0.04690560102462769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,4,balanced,0.13477333386739096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,4,power_law_1.01,0.03746559917926788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,4,power_law_1.01,0.020857599377632142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,4,balanced,0.05433600147565206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,4,power_law_1.01,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,4,power_law_1.01,0.021465599536895752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,4,power_law_1.01,0.04195840060710907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,4,balanced,0.1555466651916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,4,balanced,0.05587733288606008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,4,power_law_1.01,0.02396800071001053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,4,power_law_1.01,0.06520959734916687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,4,power_law_1.01,0.04688639938831329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,4,balanced,0.21845867236455283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,4,balanced,0.06436266501744588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,4,power_law_1.01,0.023865599930286408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,4,power_law_1.01,0.07797120213508606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,4,power_law_1.01,0.04636160135269165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,4,balanced,0.08477866649627686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,4,balanced,0.26128000020980835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,4,power_law_1.01,0.024831999838352204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,4,power_law_1.01,0.05634559988975525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,4,power_law_1.01,0.08647040128707886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,4,balanced,0.12036266922950745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,4,power_law_1.01,0.03488639891147614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,4,power_law_1.01,0.06037759780883789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,4,power_law_1.01,0.11297279596328735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,4,balanced,0.32390934228897095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,4,power_law_1.01,0.03500159978866577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,4,balanced,0.1490239997704824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,4,power_law_1.01,0.07074559926986694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,4,power_law_1.01,0.13867520093917846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,4,power_law_1.01,0.036422398686408994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,4,balanced,0.2036906679471334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,4,balanced,0.4657919804255168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,4,power_law_1.01,0.15653120279312133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,4,balanced,0.24971733490626016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,4,power_law_1.01,0.09811840057373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,4,power_law_1.01,0.04038400053977966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,4,power_law_1.01,0.2348543882369995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,4,balanced,0.3521226644515991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,4,power_law_1.01,0.12013440132141114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,4,power_law_1.01,0.043993601202964784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,4,power_law_1.01,0.26803839206695557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,4,balanced,0.8461546897888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,4,balanced,0.44041601816813153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,4,power_law_1.01,0.1667904019355774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.01,0.04584319889545441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,4,power_law_1.01,0.4375999927520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,4,power_law_1.01,0.19571199417114257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.01,0.05312640070915222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,4,balanced,0.5435893138249716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,4,power_law_1.01,0.5425983905792237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,4,power_law_1.01,0.30465919971466066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.01,0.06119040250778198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,4,power_law_1.01,0.6082880020141601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,4,power_law_1.01,0.3379839897155762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.01,0.08552320003509521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,4,balanced,0.8324426809946696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,4,power_law_1.01,1.1324671745300292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,4,power_law_1.01,0.529856014251709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.01,0.0992255985736847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,4,power_law_1.01,1.6878719329833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,4,power_law_1.01,0.7338304042816162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.01,0.13453439474105836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,4,power_law_1.01,0.8663104057312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.01,0.17138559818267823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,4,power_law_1.01,1.320467185974121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.01,0.2437567949295044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,4,balanced,1.6028639475504558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.01,0.30472960472106936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,4,power_law_1.01,2.877280044555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,4,power_law_1.01,0.3788480043411255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,4,power_law_1.01,0.5852416038513184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,4,power_law_1.01,1.1422464370727539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,4,power_law_1.2,0.04108160138130188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,4,power_law_1.2,0.030854400992393494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,4,power_law_1.2,0.03283199965953827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,4,power_law_1.2,0.032518398761749265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,4,power_law_1.2,0.03304960131645203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,4,power_law_1.2,0.033292800188064575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,4,power_law_1.2,0.03427839875221252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,4,power_law_1.2,0.031007999181747438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,4,power_law_1.2,0.03500159978866577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,4,power_law_1.2,0.03195520043373108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,4,power_law_1.2,0.03559040129184723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,4,power_law_1.2,0.03054080009460449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,4,power_law_1.2,0.036025598645210266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,4,power_law_1.2,0.03320960104465485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,4,power_law_1.2,0.037248000502586365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,4,power_law_1.2,0.041222399473190306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,4,power_law_1.2,0.032569599151611325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,4,power_law_1.2,0.04012160003185272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,4,power_law_1.2,0.0389055997133255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,4,power_law_1.2,0.03338240087032318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,4,power_law_1.2,0.040550398826599124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,4,power_law_1.2,0.039136001467704774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,4,power_law_1.2,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,4,power_law_1.2,0.0421887993812561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,4,power_law_1.2,0.041920000314712526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,4,power_law_1.2,0.03473280072212219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,4,power_law_1.2,0.04645119905471802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,4,power_law_1.2,0.04201599955558777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,4,power_law_1.2,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,4,power_law_1.2,0.0555840015411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,4,power_law_1.2,0.04257279932498932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,4,power_law_1.2,0.03601279854774475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,4,power_law_1.2,0.056704002618789676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,4,power_law_1.2,0.04291839897632599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,4,power_law_1.2,0.03749119937419891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,4,power_law_1.2,0.06705920100212097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,4,power_law_1.2,0.042777600884437564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,4,power_law_1.2,0.03845120072364807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,4,power_law_1.2,0.07809919714927674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,4,power_law_1.2,0.04254719913005829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,4,power_law_1.2,0.03991039991378784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,4,power_law_1.2,0.09932159781455993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,4,power_law_1.2,0.04287999868392944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,4,power_law_1.2,0.040889599919319154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,4,power_law_1.2,0.11323519945144653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,4,power_law_1.2,0.04312959909439087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,4,power_law_1.2,0.04780159890651703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,4,power_law_1.2,0.13848320245742798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,4,power_law_1.2,0.04456959962844849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,4,power_law_1.2,0.04747520089149475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,4,power_law_1.2,0.167084801197052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,4,power_law_1.2,0.045337599515914914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,4,power_law_1.2,0.05379840135574341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,4,power_law_1.2,0.23116800785064698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,4,power_law_1.2,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,4,power_law_1.2,0.06663039922714234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,4,power_law_1.2,0.32658560276031495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,4,power_law_1.2,0.07211520075798035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,4,power_law_1.2,0.051948797702789304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,4,power_law_1.2,0.47693438529968263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,4,power_law_1.2,0.10133119821548461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,4,power_law_1.2,0.052832001447677614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,4,power_law_1.2,0.5776256084442138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,4,power_law_1.2,0.14520319700241088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,4,power_law_1.2,0.05573760271072388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,4,power_law_1.2,0.6584832191467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,4,power_law_1.2,0.17341439723968505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,4,power_law_1.2,0.06478080153465271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,4,power_law_1.2,1.3338047981262207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,4,power_law_1.2,0.22597761154174806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,4,power_law_1.2,2.146623992919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.2,0.06722559928894042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,4,power_law_1.2,0.2866431951522827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.2,0.09300479888916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,4,power_law_1.2,0.39237759113311765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.2,0.11661440134048462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,4,power_law_1.2,0.6244031906127929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.2,0.1292415976524353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,4,power_law_1.2,0.858790397644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.2,0.18171520233154298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,4,power_law_1.2,0.9736448287963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.2,0.23868799209594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,4,power_law_1.2,1.557151985168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.2,0.35608959197998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,4,power_law_1.2,2.836454391479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.2,0.44196481704711915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.2,0.5666751861572266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,4,power_law_1.2,0.802188777923584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,4,power_law_1.2,1.130502414703369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,4,power_law_1.2,2.376486396789551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,4,power_law_1.2,0.018540799617767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,4,power_law_1.2,0.018182399868965148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,4,power_law_1.2,0.01852159947156906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,4,power_law_1.2,0.01845120042562485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,4,power_law_1.2,0.018592000007629395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,4,power_law_1.2,0.01969279944896698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,4,power_law_1.2,0.020236800611019134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,4,power_law_1.2,0.020153599977493285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,4,power_law_1.2,0.02139520049095154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,4,power_law_1.2,0.021580800414085388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,4,power_law_1.2,0.023571200668811798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,4,power_law_1.2,0.025331199169158936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,4,power_law_1.2,0.02531839907169342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,4,power_law_1.2,0.03468160033226013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,4,power_law_1.2,0.036185601353645326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,4,power_law_1.2,0.036697599291801455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,4,power_law_1.2,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,128,balanced,0.0823466678460439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,128,balanced,0.06250666578610738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,128,balanced,0.06020266811052958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,4,power_law_1.2,0.042342400550842284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,128,balanced,0.023413332800070446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,128,balanced,0.08213866750399272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,128,balanced,0.06267199913660686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,128,balanced,0.06057066718737284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.2,0.04576640129089356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,128,balanced,0.023765332996845245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,128,balanced,0.08180266618728638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,128,balanced,0.06242666641871134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,128,balanced,0.06047466893990835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.2,0.05294079780578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,128,balanced,0.025093334416548412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,128,balanced,0.08277333279450734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,128,balanced,0.06425066788991292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,128,balanced,0.06198933223883311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,128,power_law_1.01,0.09115520119667053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.2,0.06158080101013184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,128,balanced,0.02741333345572154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,128,balanced,0.08468799789746602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,128,balanced,0.06526400148868561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,128,power_law_1.01,0.06600959897041321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,128,balanced,0.0645653357108434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,128,power_law_1.01,0.09698560237884521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.2,0.0858560025691986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,128,balanced,0.03523733218510946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,128,balanced,0.09715200463930766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,128,balanced,0.07855999966462453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,128,power_law_1.01,0.0710528016090393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,128,balanced,0.07375466823577881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,128,power_law_1.01,0.06554880142211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,128,power_law_1.01,0.0873856008052826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.2,0.09939839839935302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,128,balanced,0.055215999484062195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,128,balanced,0.09744000434875488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,128,balanced,0.07226666808128357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,128,balanced,0.09904533624649048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,128,balanced,0.07455466687679291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,128,power_law_1.01,0.07008640170097351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,128,power_law_1.01,0.06469759941101075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,128,power_law_1.01,0.09562879800796509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.2,0.1349120020866394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,128,balanced,0.051514665285746254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,128,balanced,0.07363733152548473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,128,balanced,0.10439466436704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,128,balanced,0.07449066638946533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,128,power_law_1.01,0.07026559710502625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,128,power_law_1.01,0.06908159852027893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,128,power_law_1.01,0.09444479942321778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.2,0.17210240364074708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,128,balanced,0.0536053329706192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,128,balanced,0.07579199969768524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,128,balanced,0.10128000378608704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,128,balanced,0.07249600191911061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,128,power_law_1.01,0.07586560249328614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,128,power_law_1.01,0.06901760101318359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,128,power_law_1.01,0.09774079918861389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.2,0.24367361068725585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,128,balanced,0.07470400134722392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,128,balanced,0.052757332722345986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,128,balanced,0.10531199971834819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,128,balanced,0.0765066643555959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,128,power_law_1.01,0.0722815990447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,128,power_law_1.01,0.07288960218429566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,128,power_law_1.01,0.0978879988193512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.2,0.3051584005355835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,128,balanced,0.07675733168919881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,4,power_law_1.2,0.37879679203033445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,128,balanced,0.05269333223501841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,128,balanced,0.09916266798973083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,128,balanced,0.07474133372306824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,128,power_law_1.01,0.0738431990146637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,128,power_law_1.01,0.0744704008102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,128,power_law_1.01,0.09773439764976502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,128,balanced,0.07992533346017201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,4,power_law_1.2,0.5859776020050049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,128,balanced,0.0516480008761088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,128,balanced,0.1088213324546814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,128,balanced,0.07820266485214233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,128,power_law_1.01,0.07308800220489502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,128,power_law_1.01,0.07308160066604615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,128,power_law_1.01,0.10046080350875855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,128,balanced,0.08090666433175404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,4,power_law_1.2,1.1426624298095702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,128,balanced,0.048783997694651283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,128,balanced,0.09930133819580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,128,balanced,0.07852800190448761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,128,power_law_1.01,0.07373440265655518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,128,power_law_1.01,0.0765504002571106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,128,power_law_1.01,0.09950720071792603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,128,balanced,0.08197866876920064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,128,balanced,0.04394133388996124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,128,balanced,0.11425600449244182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,128,balanced,0.08458667000134786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,128,power_law_1.01,0.07630079984664917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,128,power_law_1.01,0.07437440156936645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,128,power_law_1.01,0.10097279548645019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,128,balanced,0.08331733445326488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,128,balanced,0.04218666752179464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,128,balanced,0.105103999376297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,128,balanced,0.08169066905975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,128,power_law_1.01,0.07396479845046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,128,power_law_1.01,0.0753920018672943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,128,power_law_1.01,0.10122239589691162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,128,balanced,0.09715200463930766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,128,balanced,0.06063466767470042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,128,balanced,0.10697600245475769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,128,balanced,0.08288000027338664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,128,power_law_1.01,0.07482879757881164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,128,power_law_1.01,0.07570559978485107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,128,power_law_1.01,0.10147199630737305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,128,balanced,0.09400000174840291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,128,balanced,0.06264000137646993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,128,balanced,0.1222879985968272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,128,balanced,0.08429333567619324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,128,power_law_1.01,0.0757695972919464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,128,power_law_1.01,0.0756991982460022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,128,power_law_1.01,0.09907199740409851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,128,balanced,0.10963199536005656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,128,power_law_1.01,0.07864959836006165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,128,balanced,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,128,balanced,0.11559466520945232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,128,balanced,0.09084799885749817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,128,power_law_1.01,0.07461119890213012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,128,power_law_1.01,0.10189440250396728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,128,balanced,0.11038399736086528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,128,power_law_1.01,0.08581119775772095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,128,balanced,0.06859733164310455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,128,balanced,0.1295199990272522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,128,balanced,0.1279253363609314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,128,power_law_1.01,0.07653120160102844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,128,balanced,0.09512533744176228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,128,power_law_1.01,0.10350079536437988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,128,power_law_1.01,0.08609920144081115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,128,balanced,0.07638399799664815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,128,balanced,0.14774933457374573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,128,balanced,0.13572266697883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,128,power_law_1.01,0.0807807981967926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,128,balanced,0.10962133606274922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,128,power_law_1.01,0.10135680437088013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,128,power_law_1.01,0.09041919708251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,128,balanced,0.10228266318639119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,128,power_law_1.01,0.07919359803199769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,128,balanced,0.15687466661135355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,128,balanced,0.18249066670735678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,128,balanced,0.1255573332309723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,128,power_law_1.01,0.1181056022644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,128,power_law_1.01,0.10665600299835205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,128,balanced,0.12504532933235168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,128,power_law_1.01,0.08184959888458251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,128,balanced,0.17914666732152304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,128,balanced,0.21802133321762085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,128,power_law_1.01,0.12190719842910766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,128,balanced,0.14752533038457236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,128,power_law_1.01,0.11509120464324951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,128,power_law_1.01,0.09534720182418824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,128,balanced,0.17409600814183554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,128,power_law_1.01,0.16405760049819945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,128,power_law_1.01,0.14455039501190187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.01,0.10606720447540283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,128,balanced,0.2251840035120646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,128,balanced,0.28988800446192425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,128,balanced,0.17362133661905924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,128,balanced,0.21626132726669312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,128,power_law_1.01,0.17216639518737792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,128,power_law_1.01,0.16958719491958618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.01,0.12839679718017577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,128,balanced,0.26525332530339557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,128,power_law_1.01,0.21979520320892335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,128,balanced,0.359061320622762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,128,balanced,0.2330133318901062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,128,power_law_1.01,0.22339200973510742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,128,balanced,0.31698666016260785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.01,0.14448000192642213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,128,power_law_1.01,0.27915520668029786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,128,balanced,0.37056533495585126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,128,power_law_1.01,0.2735935926437378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,128,balanced,0.5091466506322225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,128,balanced,0.2767680088678996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.01,0.18712320327758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,128,power_law_1.01,0.37436161041259763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,128,balanced,0.4179840087890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,128,power_law_1.01,0.3724735975265503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,128,power_law_1.01,0.477510404586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.01,0.22218880653381348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,128,power_law_1.01,0.4704576015472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,128,power_law_1.01,0.6973887920379639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.01,0.3059135913848877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,128,power_law_1.01,0.8797439575195313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.01,0.37857279777526853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,128,power_law_1.01,0.6832320213317871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,128,balanced,0.4479733308156331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,128,balanced,0.38234134515126544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,128,power_law_1.01,1.0844608306884767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.01,0.5490303993225097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,128,balanced,0.6452213525772095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,128,balanced,0.6122293472290039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,128,power_law_1.01,0.8722816467285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,128,power_law_1.01,1.6130687713623046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.01,0.6939328193664551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,128,power_law_1.01,1.0799296379089356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,128,balanced,0.4837119976679484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,128,power_law_1.01,3.3966209411621096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,128,power_law_1.01,0.8601920127868652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,128,balanced,0.5511519908905029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,128,power_law_1.01,1.7334720611572265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,128,power_law_1.01,1.3137280464172363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,128,balanced,0.8053599993387858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,128,balanced,0.7911626497904459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,128,power_law_1.01,3.255712127685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,128,power_law_1.01,2.6014144897460936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,128,balanced,0.5868853330612183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,128,balanced,1.0007200241088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,128,balanced,0.8240373134613037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,128,balanced,1.216421365737915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,128,balanced,0.8957440058390299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,128,power_law_1.01,0.02280319929122925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,128,power_law_1.01,0.023078399896621703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,128,balanced,1.583024024963379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,128,power_law_1.01,0.02282879948616028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,128,power_law_1.01,0.02502399981021881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,128,power_law_1.01,0.030943998694419862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,128,power_law_1.01,0.05364480018615723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,128,power_law_1.01,0.05100160241127014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,128,power_law_1.01,0.05246719717979431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,128,power_law_1.01,0.05279359817504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,128,power_law_1.01,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,128,power_law_1.01,0.05159040093421936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,128,balanced,1.5907039642333984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,128,power_law_1.01,0.05069440007209778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,128,balanced,2.3881279627482095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,128,power_law_1.01,0.048895999789237976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,128,power_law_1.01,0.0470335990190506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,128,power_law_1.01,0.0465472012758255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,128,balanced,1.7610400517781575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,128,power_law_1.01,0.046751999855041505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,128,power_law_1.01,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,128,power_law_1.01,0.05612159967422485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,128,balanced,3.139333407084147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.01,0.06755840182304382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.01,0.08755840063095092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.01,0.09940479993820191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.01,0.13781119585037233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.01,0.16969599723815917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.01,0.2297152042388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.01,0.3146496057510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.01,0.45183358192443845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.01,0.6200191974639893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,128,power_law_1.01,0.7416639804840088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,128,power_law_1.01,1.175001621246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,128,power_law_1.01,2.3429183959960938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,128,power_law_1.2,0.09121919870376587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,128,power_law_1.2,0.09693440198898315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,128,power_law_1.2,0.08366720080375671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,128,power_law_1.2,0.09595519900321961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,128,power_law_1.2,0.09569919705390931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,128,power_law_1.2,0.09838079810142517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,128,power_law_1.2,0.06607360243797303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,128,power_law_1.2,0.07077760100364686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,128,power_law_1.2,0.10019199848175049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,128,power_law_1.2,0.06574079990386963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,128,power_law_1.2,0.07026559710502625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,128,power_law_1.2,0.09866880178451538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,128,power_law_1.2,0.0639743983745575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,128,power_law_1.2,0.06554880142211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,128,power_law_1.2,0.10263680219650269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,128,power_law_1.2,0.0704576015472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,128,power_law_1.2,0.07329919934272766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,128,power_law_1.2,0.09843199849128723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,128,power_law_1.2,0.06811519861221313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,128,power_law_1.2,0.07038080096244811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,128,power_law_1.2,0.10061440467834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,128,power_law_1.2,0.07111039757728577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,128,power_law_1.2,0.0717248022556305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,128,power_law_1.2,0.09745280146598816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,128,power_law_1.2,0.07176960110664368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,128,power_law_1.2,0.07276800274848938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,128,power_law_1.2,0.09953280091285706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,128,power_law_1.2,0.0735871970653534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,128,power_law_1.2,0.07504640221595764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,128,power_law_1.2,0.09947519898414611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,128,power_law_1.2,0.07249919772148132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,128,power_law_1.2,0.07552639842033386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,128,power_law_1.2,0.10311039686203002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,128,power_law_1.2,0.07416319847106934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,128,power_law_1.2,0.07365760207176208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,128,power_law_1.2,0.10101120471954346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,128,power_law_1.2,0.07400320172309875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,128,power_law_1.2,0.07416319847106934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,128,power_law_1.2,0.10960639715194702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,128,power_law_1.2,0.07572479844093323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,128,power_law_1.2,0.07801600098609925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,128,power_law_1.2,0.11809279918670654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,128,power_law_1.2,0.0732096016407013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,128,power_law_1.2,0.0755840003490448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,128,power_law_1.2,0.14035840034484864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,128,power_law_1.2,0.07486720085144043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,128,power_law_1.2,0.07694079875946044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,64,balanced,0.06365333497524261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,128,power_law_1.2,0.022668799757957457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,128,power_law_1.2,0.16977280378341675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,128,power_law_1.2,0.0794048011302948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,128,power_law_1.2,0.08419839739799499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,64,balanced,0.0545066644748052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,64,balanced,0.06389866769313812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,128,power_law_1.2,0.023174400627613067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,128,power_law_1.2,0.20377600193023682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,128,power_law_1.2,0.07973120212554932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,128,power_law_1.2,0.08633599877357483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,128,power_law_1.2,0.02351360023021698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,64,balanced,0.055642664432525635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,128,power_law_1.2,0.09848319888114929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,64,balanced,0.05436799923578898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,128,power_law_1.2,0.026150399446487428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,64,balanced,0.0636106679836909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,128,power_law_1.2,0.28958079814910886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,128,power_law_1.2,0.09313279986381531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,64,balanced,0.055573334296544395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,128,power_law_1.2,0.1099392056465149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,64,balanced,0.05491200089454651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,128,power_law_1.2,0.033011201024055484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,64,balanced,0.06493333478768666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,128,power_law_1.2,0.34960639476776123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,64,balanced,0.05606933434804281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,128,power_law_1.2,0.09992960095405579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,128,power_law_1.2,0.11934720277786255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,64,balanced,0.05628266433874766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,128,power_law_1.2,0.05375999808311462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,64,balanced,0.0743146687746048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,128,power_law_1.2,0.46706562042236327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,64,balanced,0.05618133147557577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.2,0.1142016053199768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,64,balanced,0.023391999304294586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,128,power_law_1.2,0.15686399936676027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,64,balanced,0.06424533327420552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,128,power_law_1.2,0.18223999738693236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,128,power_law_1.2,0.052697598934173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,64,balanced,0.09619733691215515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,64,balanced,0.07231999933719635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,128,power_law_1.2,0.678764820098877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,64,balanced,0.06632533172766368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.2,0.13363200426101685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,64,balanced,0.0233599990606308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,64,balanced,0.07061866422494252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,128,power_law_1.2,0.24595201015472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,128,power_law_1.2,0.053337597846984865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,64,balanced,0.09169066945711772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,64,balanced,0.07027733325958252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,128,power_law_1.2,0.8876416206359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.2,0.15710079669952393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,64,balanced,0.023872000475724537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,64,balanced,0.07230400045712788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,128,power_law_1.2,0.31363840103149415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,128,power_law_1.2,0.0543936014175415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,64,balanced,0.09309867024421692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,64,balanced,0.07098666826883952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,128,power_law_1.2,1.2328384399414063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.2,0.20177919864654542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,64,balanced,0.0232640008131663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,64,balanced,0.07229333122571309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,128,power_law_1.2,0.4363711833953857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,128,power_law_1.2,0.053830397129058835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,64,balanced,0.10296533505121867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,64,balanced,0.07411733269691467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,128,power_law_1.2,1.6008447647094726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.2,0.25214080810546874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,64,balanced,0.025770666698614757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,64,balanced,0.07433066765467326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,128,power_law_1.2,0.5661439895629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.2,0.35287039279937743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,128,power_law_1.2,0.053964799642562865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,64,balanced,0.09936533371607463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,64,balanced,0.07230400045712788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,128,power_law_1.2,2.6734079360961913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,64,balanced,0.030154667794704437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,64,balanced,0.07230933507283528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,128,power_law_1.2,0.8766336441040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.2,0.4630144119262695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,128,power_law_1.2,0.05198720097541809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,64,balanced,0.10105066498120625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,64,balanced,0.07472000022729237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,64,balanced,0.05530133346716563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,128,power_law_1.2,6.375507354736328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,64,balanced,0.07443200051784515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,128,power_law_1.2,1.0912320137023925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.2,0.6789824008941651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,128,power_law_1.2,0.05245440006256104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,64,balanced,0.10036800305048625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,64,balanced,0.07634666562080383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,64,balanced,0.05584000051021576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,64,balanced,0.0740533322095871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,128,power_law_1.2,1.3907072067260742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,128,power_law_1.2,0.043059200048446655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.2,1.0141823768615723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,64,balanced,0.10366400082906087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,64,balanced,0.0781173308690389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,64,balanced,0.05357333521048228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,64,balanced,0.07521600027879079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,128,power_law_1.2,2.5821760177612303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,128,power_law_1.2,0.04478079974651337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,128,power_law_1.2,1.0776320457458497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,64,balanced,0.09890666604042053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,64,balanced,0.08201600114504497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,64,balanced,0.05229333539803823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,64,balanced,0.0774239997069041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,128,power_law_1.2,5.191104125976563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,128,power_law_1.2,0.04678399860858917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,128,power_law_1.2,1.8604415893554687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,64,balanced,0.10616532961527507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,64,balanced,0.0846506655216217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,64,balanced,0.05146666864554087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,64,balanced,0.08052266637484233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,128,power_law_1.2,0.055353599786758426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,128,power_law_1.2,3.796275329589844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,64,balanced,0.10177600383758545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,64,balanced,0.0845973292986552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,64,balanced,0.04986133178075155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,64,balanced,0.08081600069999695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,128,power_law_1.2,0.06286079883575439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,64,balanced,0.09314133723576863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,64,balanced,0.1079253355662028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,64,balanced,0.0465280016263326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.2,0.07200000286102295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,64,balanced,0.08363200227419536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,64,balanced,0.10426132877667744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,64,balanced,0.11392533779144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,64,balanced,0.04475200176239014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.2,0.08594560027122497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,64,balanced,0.08984000484148662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,64,balanced,0.1104853351910909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,64,balanced,0.11618133385976155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,64,balanced,0.05993066728115082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.2,0.10903040170669556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,64,balanced,0.09291733304659526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,64,balanced,0.12999467055002847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,64,balanced,0.12602133552233377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.2,0.14515199661254882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,64,balanced,0.06489600241184235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,64,balanced,0.11109333237012227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,64,balanced,0.14888532956441244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,64,balanced,0.1358453333377838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.2,0.18473600149154662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,64,balanced,0.05187733471393585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,64,balanced,0.1253653367360433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.2,0.2670079946517944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,64,balanced,0.1492693324883779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,64,balanced,0.18584533532460532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,64,balanced,0.1586720049381256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,64,balanced,0.07640000184377034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.2,0.3302079916000366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,64,balanced,0.17568532625834146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,64,balanced,0.1811573306719462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,64,balanced,0.08262933293978374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,64,balanced,0.2218559980392456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.2,0.502342414855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,64,balanced,0.2358400026957194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,64,balanced,0.10925867160161336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,64,balanced,0.22760534286499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.2,0.7022848129272461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,64,balanced,0.2951146761576335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,64,balanced,0.13430933157602945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,64,balanced,0.2824746568997701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,64,balanced,0.26879467566808063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,128,power_law_1.2,0.8469568252563476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,64,balanced,0.1879840095837911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,64,balanced,0.36368000507354736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,128,power_law_1.2,1.290937614440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,64,balanced,0.38996267318725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,64,balanced,0.37220267454783124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,64,balanced,0.23146132628122965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,128,power_law_1.2,2.841049575805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,64,balanced,0.33691732088724774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,64,balanced,0.5196853478749593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,64,balanced,0.49456532796223956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,64,balanced,0.4573386510213216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,64,balanced,0.4376159906387329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,64,balanced,0.656496008237203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,64,balanced,0.6578239997227987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,64,balanced,0.602944016456604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,64,balanced,0.5647840102513632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,64,balanced,0.8072960376739502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,64,balanced,0.8754133383433024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,64,balanced,0.9245493412017822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,64,balanced,0.8391573429107666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,64,balanced,1.084554672241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,64,balanced,1.2423466841379802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,64,balanced,1.7187520662943523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,64,balanced,1.810863971710205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,64,balanced,1.6221760114034016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,64,balanced,2.4426453908284507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,64,balanced,3.4050238927205405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,64,power_law_1.01,0.066457599401474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,64,power_law_1.01,0.09203199744224548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,64,power_law_1.01,0.07557759881019592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,64,power_law_1.01,0.022873599827289582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,64,power_law_1.01,0.0649728000164032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,64,power_law_1.01,0.08912000060081482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,64,power_law_1.01,0.07363839745521546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,64,power_law_1.01,0.022758400440216063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,64,power_law_1.01,0.0576960027217865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,64,power_law_1.01,0.08442879915237426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,64,power_law_1.01,0.0695360004901886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,64,power_law_1.01,0.022777600586414336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,64,power_law_1.01,0.060601598024368285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,64,power_law_1.01,0.08580480217933655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,64,power_law_1.01,0.0705344021320343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,64,power_law_1.01,0.02284799963235855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,64,power_law_1.01,0.06376320123672485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,64,power_law_1.01,0.025996801257133485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,64,power_law_1.01,0.08786560297012329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,64,power_law_1.01,0.06773759722709656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,64,power_law_1.01,0.06842880249023438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,64,power_law_1.01,0.03207040131092072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,64,power_law_1.01,0.08981119990348815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,64,power_law_1.01,0.06887680292129517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,64,power_law_1.01,0.06939520239830017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,64,power_law_1.01,0.05454080104827881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,64,power_law_1.01,0.0946175992488861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,64,power_law_1.01,0.0693120002746582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,64,power_law_1.01,0.07439360022544861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,64,power_law_1.01,0.0542464017868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,64,power_law_1.01,0.09393280148506164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,64,power_law_1.01,0.07162240147590637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,64,power_law_1.01,0.07360000014305115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,64,power_law_1.01,0.05443199872970581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,64,power_law_1.01,0.0942911982536316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,64,power_law_1.01,0.07254400253295898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,64,power_law_1.01,0.07101439833641052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,64,power_law_1.01,0.054028797149658206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,64,power_law_1.01,0.09566079974174499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,64,power_law_1.01,0.07336959838867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,64,power_law_1.01,0.07304319739341736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,64,power_law_1.2,0.0649407982826233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,64,power_law_1.2,0.0919871985912323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,64,power_law_1.01,0.054048001766204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,64,power_law_1.01,0.09526399970054626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,64,power_law_1.01,0.07608960270881653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,64,power_law_1.01,0.07564160227775574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,64,power_law_1.2,0.063155198097229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,64,power_law_1.2,0.07574399709701538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,64,power_law_1.2,0.09095039963722229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,64,power_law_1.2,0.059436798095703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,64,power_law_1.01,0.05331839919090271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,64,power_law_1.01,0.09687680006027222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,64,power_law_1.01,0.051686400175094606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,64,power_law_1.01,0.07571200132369996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,64,power_law_1.01,0.07491840124130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,64,power_law_1.2,0.07425280213356018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,64,power_law_1.2,0.07198079824447631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,64,power_law_1.2,0.05979520082473755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,64,power_law_1.01,0.10001280307769775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,64,power_law_1.01,0.04935680031776428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,64,power_law_1.01,0.07281919717788696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,64,power_law_1.2,0.02279040068387985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,64,power_law_1.01,0.08081279993057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,64,power_law_1.2,0.06542720198631287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,64,power_law_1.2,0.0819711983203888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,64,power_law_1.2,0.06373760104179382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,64,power_law_1.01,0.0967743992805481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,64,power_law_1.01,0.050374400615692136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,64,power_law_1.01,0.07553920149803162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,64,power_law_1.2,0.022873599827289582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,64,power_law_1.01,0.08186240196228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,64,power_law_1.2,0.06567680239677429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,64,power_law_1.2,0.08810880184173583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,64,power_law_1.2,0.06598399877548218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,64,power_law_1.01,0.1011072039604187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,64,power_law_1.01,0.051846402883529666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,64,power_law_1.01,0.07994880080223084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,64,power_law_1.2,0.022809599339962006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,64,power_law_1.01,0.08453760147094727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,64,power_law_1.2,0.06484479904174804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,64,power_law_1.2,0.09043200016021728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,64,power_law_1.2,0.06766719818115234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,64,power_law_1.01,0.1014847993850708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,64,power_law_1.01,0.05143679976463318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,64,power_law_1.01,0.08146560192108154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,64,power_law_1.2,0.021849599480628968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,64,power_law_1.01,0.09200639724731445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,64,power_law_1.2,0.06795520186424256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,64,power_law_1.2,0.09828479886054993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,64,power_law_1.2,0.0720192015171051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,64,power_law_1.01,0.1047551989555359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,64,power_law_1.01,0.05454080104827881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,64,power_law_1.01,0.08416000008583069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,64,power_law_1.2,0.024953599274158477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,64,power_law_1.01,0.10558079481124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,64,power_law_1.2,0.06808320283889771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,64,power_law_1.2,0.09233279824256897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,64,power_law_1.2,0.07189120054244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,64,power_law_1.01,0.11336959600448608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.01,0.06028159856796265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,64,power_law_1.01,0.09810559749603272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,64,power_law_1.2,0.031430399417877196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,64,power_law_1.01,0.1146623969078064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,64,power_law_1.2,0.07249280214309692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,64,power_law_1.2,0.0942080020904541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,64,power_law_1.2,0.07048959732055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,64,power_law_1.01,0.12048640251159667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.01,0.07802240252494812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.01,0.10622080564498901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,64,power_law_1.2,0.054176002740859985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,64,power_law_1.01,0.1437440037727356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.01,0.12400640249252319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,64,power_law_1.2,0.05420799851417542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,64,power_law_1.2,0.07301759719848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,64,power_law_1.2,0.0950976014137268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,64,power_law_1.2,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,64,power_law_1.01,0.14624639749526977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.01,0.09081599712371827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,64,power_law_1.01,0.15592960119247437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,64,power_law_1.2,0.0713919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,64,power_law_1.01,0.15603840351104736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.01,0.14194560050964355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,64,power_law_1.2,0.053491199016571046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,64,power_law_1.2,0.07516800165176392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,64,power_law_1.2,0.07433599829673768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,64,power_law_1.2,0.09681919813156128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.01,0.12055039405822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,64,power_law_1.01,0.21551361083984374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,64,power_law_1.01,0.19967999458312988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.01,0.18056319952011107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,64,power_law_1.2,0.05260800123214722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,64,power_law_1.2,0.08141440153121948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,64,power_law_1.2,0.07313920259475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,64,power_law_1.2,0.09288319945335388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.01,0.1507904052734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,64,power_law_1.01,0.256441593170166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,64,power_law_1.01,0.25418241024017335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.01,0.21639039516448974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,64,power_law_1.2,0.05203840136528015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,64,power_law_1.2,0.08199040293693542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,64,power_law_1.2,0.07338240146636962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,64,power_law_1.2,0.09710720181465149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.01,0.21091198921203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,64,power_law_1.01,0.34929919242858887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,64,power_law_1.01,0.31930880546569823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,64,power_law_1.2,0.09514880180358887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.01,0.2866751909255981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,64,power_law_1.2,0.05140479803085327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,64,power_law_1.01,0.3999871969223022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,64,power_law_1.2,0.08652160167694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,64,power_law_1.2,0.07494400143623352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,64,power_law_1.01,0.4456319808959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.01,0.23629438877105713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,64,power_law_1.2,0.10062719583511352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.01,0.36874239444732665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,64,power_law_1.2,0.05018240213394165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,64,power_law_1.01,0.5623040199279785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,64,power_law_1.2,0.09066879749298096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,64,power_law_1.2,0.07528960108757018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,64,power_law_1.01,0.6301439762115478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.01,0.35717120170593264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,64,power_law_1.2,0.09746559858322143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.01,0.509881591796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,64,power_law_1.2,0.04904960095882416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,64,power_law_1.01,0.6781824111938477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,64,power_law_1.2,0.10786559581756591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,64,power_law_1.2,0.07984640002250672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,64,power_law_1.01,0.8181632041931153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.01,0.48643198013305666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,64,power_law_1.2,0.10453120470046998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.01,0.6708672046661377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,64,power_law_1.2,0.05025280117988586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,64,power_law_1.01,0.8835264205932617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,64,power_law_1.2,0.11831040382385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,64,power_law_1.2,0.08023040294647217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,64,power_law_1.01,1.0259200096130372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,64,power_law_1.01,0.6321343898773193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,64,power_law_1.2,0.08798080086708068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,64,power_law_1.2,0.11552640199661254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,64,power_law_1.01,0.8122624397277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,64,power_law_1.2,0.054745602607727054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,64,power_law_1.01,1.3739904403686523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,64,power_law_1.2,0.14591360092163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,64,power_law_1.01,1.6984128952026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,64,power_law_1.01,0.9232640266418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,64,power_law_1.2,0.10226559638977051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,64,power_law_1.2,0.12560640573501586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,64,power_law_1.01,1.3004863739013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,64,power_law_1.2,0.05438079833984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,64,power_law_1.01,2.6856639862060545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,64,power_law_1.2,0.1762112021446228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,64,power_law_1.01,3.3874431610107423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,64,power_law_1.01,2.0436223983764648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.2,0.11218559741973877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,64,power_law_1.2,0.15243519544601442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,64,power_law_1.2,0.05214080214500427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,64,power_law_1.01,2.505945587158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,64,power_law_1.2,0.23722879886627196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.2,0.1301568031311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,64,power_law_1.2,0.17929600477218627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.2,0.06287999749183655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,64,power_law_1.2,0.2821376085281372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.2,0.1505728006362915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,64,power_law_1.2,0.23725440502166747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.2,0.08499199748039246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,64,power_law_1.2,0.3987584114074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,64,power_law_1.2,0.2733504056930542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.2,0.19211519956588746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.2,0.0993664026260376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,64,power_law_1.2,0.5031424045562745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,64,power_law_1.2,0.3936064004898071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.2,0.2352447986602783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.2,0.1318400025367737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,64,power_law_1.2,0.7830207824707032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,64,power_law_1.2,0.5119999885559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.2,0.15673600435256957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.2,0.3136192083358765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,64,power_law_1.2,0.9861248016357422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,64,power_law_1.2,0.688972806930542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.2,0.2254784107208252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.2,0.394598388671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,64,power_law_1.2,1.251046371459961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,64,power_law_1.2,0.8458432197570801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.2,0.28285439014434816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.2,0.6341760158538818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,64,power_law_1.2,1.8959552764892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,64,power_law_1.2,1.199289608001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.2,0.4458752155303955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.2,0.7670527935028076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,64,power_law_1.2,3.840755081176758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,64,power_law_1.2,1.6844480514526368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.2,0.5823423862457275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,64,power_law_1.2,1.01527681350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,64,power_law_1.2,4.400537490844727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,64,power_law_1.2,0.6931519985198975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,64,power_law_1.2,1.5129664421081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,64,power_law_1.2,1.0728447914123536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,64,power_law_1.2,3.239750289916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,64,power_law_1.2,2.463852882385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,32,balanced,0.022117334107557934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,32,power_law_1.01,0.02276480048894882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,32,balanced,0.027263998985290527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,32,power_law_1.01,0.02245119959115982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,32,balanced,0.0230880007147789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,32,power_law_1.01,0.05735679864883423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,32,balanced,0.0521919975678126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,32,power_law_1.01,0.021516799926757812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,32,balanced,0.023056000471115112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,32,power_law_1.01,0.05588480234146118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,32,balanced,0.05399466554323832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,32,power_law_1.01,0.021644799411296843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,32,balanced,0.023210667073726654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,32,power_law_1.01,0.05544959902763367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,32,balanced,0.04994666576385498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,32,power_law_1.01,0.021907199919223786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,32,balanced,0.02533866713444392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,32,power_law_1.01,0.053299200534820554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,32,power_law_1.01,0.023417599499225616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,32,balanced,0.052485331892967224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,32,balanced,0.027248000105222065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,32,power_law_1.01,0.054636800289154054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,32,power_law_1.01,0.024959999322891235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,32,balanced,0.055999999245007835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,32,balanced,0.02757866680622101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,32,power_law_1.01,0.05576320290565491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,32,power_law_1.01,0.025964799523353576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,32,balanced,0.05725333094596863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,32,balanced,0.03155199935038885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,32,power_law_1.01,0.05699840188026428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,32,power_law_1.01,0.027225598692893982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,32,balanced,0.059248000383377075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,32,balanced,0.03127466638882955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,32,power_law_1.01,0.05806080102920532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,32,power_law_1.01,0.027340799570083618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,32,balanced,0.060271998246510826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,32,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,32,power_law_1.01,0.058323198556900026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,32,power_law_1.01,0.03203200101852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,32,balanced,0.0603359987338384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,32,power_law_1.01,0.03243519961833954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,32,balanced,0.038922667503356934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,32,power_law_1.01,0.03195520043373108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,32,power_law_1.01,0.0588096022605896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,32,balanced,0.05851200222969055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,32,balanced,0.037151999771595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,32,power_law_1.01,0.053446400165557864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,32,power_law_1.01,0.0591871976852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,32,balanced,0.06039999922116598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,32,balanced,0.05376533170541128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,32,power_law_1.01,0.053260797262191774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,32,power_law_1.01,0.06097279787063599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,32,balanced,0.06035199761390686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,32,balanced,0.06859200199445088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,32,power_law_1.01,0.05667200088500977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,32,balanced,0.045791998505592346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,32,power_law_1.01,0.061343997716903687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,32,balanced,0.062277331948280334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,32,balanced,0.06830933193365733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,32,power_law_1.01,0.0575872004032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,32,balanced,0.05667200187842051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,32,power_law_1.01,0.09594240188598632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,32,power_law_1.01,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,32,balanced,0.04562666515509287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,32,power_law_1.01,0.06283519864082336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,32,balanced,0.06312533219655354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,32,balanced,0.062133332093556724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,32,power_law_1.01,0.05742719769477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,32,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,32,power_law_1.01,0.06838399767875672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,32,power_law_1.01,0.06096640229225159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,32,balanced,0.04557333389918009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,32,power_law_1.01,0.07000319957733155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,32,balanced,0.06855999926726024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,32,balanced,0.06917333106199901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,32,power_law_1.01,0.047007998824119566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,32,balanced,0.04886400202910105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,32,power_law_1.01,0.05991680026054382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.01,0.0652351975440979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,32,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,32,power_law_1.01,0.06958079934120179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,32,balanced,0.06844266752401988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,32,balanced,0.08458133538564046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,32,power_law_1.01,0.04896000027656555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,32,balanced,0.04791999856630961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,32,power_law_1.01,0.05345919728279114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.01,0.07189120054244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,32,balanced,0.047983999053637184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,32,power_law_1.01,0.07567359805107117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,32,balanced,0.052111998200416565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,32,balanced,0.07268266876538594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,32,balanced,0.09946133693059285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,32,power_law_1.01,0.04989440143108368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,32,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,32,power_law_1.01,0.06085759997367859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,32,power_law_1.01,0.05237759947776795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,32,balanced,0.07463466624418895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.01,0.09312000274658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,32,power_law_1.01,0.08740479946136474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,32,balanced,0.053946668903032936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,32,balanced,0.07890666524569194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,32,balanced,0.11594133575757344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,32,power_law_1.01,0.06517760157585144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,32,balanced,0.08497066299120586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,32,power_law_1.01,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,32,balanced,0.07438933352629344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.01,0.11237759590148926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.01,0.0993664026260376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,32,balanced,0.055957332253456116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,32,balanced,0.1544533371925354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,32,power_law_1.01,0.06922240257263183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,32,balanced,0.10265066226323445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,32,power_law_1.01,0.056627202033996585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,32,balanced,0.07497066756089528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.01,0.13137279748916625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.01,0.11801600456237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,32,balanced,0.056373332937558494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,32,power_law_1.01,0.06736639738082886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,32,balanced,0.19454934199651083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,32,balanced,0.11798399686813354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,32,power_law_1.01,0.05715199708938599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,32,balanced,0.07657066484292348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.01,0.18259199857711791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.01,0.13381119966506957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,32,balanced,0.05610666672388712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,32,power_law_1.01,0.06898559927940369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,32,balanced,0.2676266630490621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,32,balanced,0.14144532879193625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,32,power_law_1.01,0.057120001316070555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,32,balanced,0.07663999994595845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.01,0.22714240550994874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.01,0.1676095962524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,32,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,32,power_law_1.01,0.06978560090065003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,32,balanced,0.3466293414433797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,32,balanced,0.1668213407198588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,32,power_law_1.01,0.05983999967575073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,32,balanced,0.07634666562080383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.01,0.32094080448150636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.01,0.19742079973220825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,32,balanced,0.059487998485565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,32,power_law_1.01,0.0715071976184845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,32,power_law_1.01,0.059699201583862306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,32,balanced,0.07843733330567677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,32,balanced,0.21528534094492593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.01,0.4168896198272705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,32,balanced,0.5103679895401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.01,0.28365440368652345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,32,balanced,0.05825600028038025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,32,power_law_1.01,0.07340800166130065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.01,0.3458240032196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,32,power_law_1.01,0.06339200139045716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,32,balanced,0.07770133515199025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,32,power_law_1.01,0.5018112182617187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,32,balanced,0.27111466725667316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,32,balanced,0.06404266754786174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,32,power_law_1.01,0.07451519966125489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.01,0.4672832012176514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,32,power_law_1.01,0.06872320175170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,32,balanced,0.07864533364772797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,32,balanced,0.6683946450551351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,32,power_law_1.01,0.7872447967529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,32,balanced,0.07047999898592631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,32,power_law_1.01,0.0750656008720398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,32,balanced,0.3640213410059611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.01,0.6222527980804443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,32,power_law_1.01,0.07548159956932068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,32,balanced,0.0846560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,32,power_law_1.01,1.6181440353393555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,32,balanced,0.07369600236415863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,32,power_law_1.01,0.08121600151062011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,32,balanced,0.08284799754619598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,32,power_law_1.01,0.7386688232421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,32,power_law_1.01,0.07854719758033753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,32,balanced,0.08454400300979614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,32,balanced,0.8394666512807211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,32,balanced,0.4546933174133301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,32,balanced,0.08679999907811482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,32,power_law_1.01,0.08388479948043823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,32,balanced,0.09240532914797465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,32,power_law_1.01,1.1847807884216308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,32,power_law_1.01,0.08579199910163879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,32,balanced,0.09309333562850952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,32,power_law_1.01,0.09318400025367737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,32,balanced,0.09974933664004008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,32,power_law_1.01,2.4951168060302735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,32,power_law_1.01,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,32,balanced,0.09669333696365356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,32,balanced,0.5585546493530273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,32,power_law_1.01,0.10215679407119752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,32,balanced,1.3134933312733967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,32,balanced,0.1181813379128774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,32,power_law_1.01,0.10506880283355713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,32,balanced,0.10618666807810466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,32,power_law_1.01,0.10625920295715333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,32,balanced,0.13454399506251016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,32,power_law_1.01,0.13222399950027466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,32,power_law_1.01,0.12389119863510131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,32,balanced,0.11866133411725362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,32,power_law_1.01,0.1517184019088745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,32,balanced,0.17185600598653158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,32,power_law_1.01,0.14033279418945313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,32,power_law_1.01,0.1984511971473694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,32,balanced,0.13672000169754028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,32,balanced,0.8375786940256754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,32,power_law_1.01,0.17715840339660643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,32,balanced,0.2050666610399882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,32,power_law_1.01,0.24314239025115966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,32,balanced,0.16150400042533875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,32,power_law_1.01,0.20932478904724122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,32,balanced,0.27372799317042035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,32,power_law_1.01,0.33917438983917236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,32,balanced,0.20020800828933716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,32,power_law_1.01,0.27368960380554197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,32,balanced,2.5963679949442544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,32,power_law_1.01,0.43194241523742677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,32,balanced,0.3447413444519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,32,power_law_1.01,0.35320320129394533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,32,balanced,0.24058133363723755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,32,power_law_1.01,0.6008448123931884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,32,power_law_1.01,0.48677759170532225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,32,power_law_1.01,0.7886911869049072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,32,balanced,1.6410400072733562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,32,balanced,0.48288532098134357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,32,power_law_1.01,0.6211264133453369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,32,power_law_1.01,0.9761728286743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,32,balanced,0.3264639973640442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,32,power_law_1.01,0.6765567779541015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,32,power_law_1.01,1.4739775657653809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,32,balanced,0.618506669998169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,32,power_law_1.01,1.1622528076171874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,32,power_law_1.01,3.040377616882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,32,power_law_1.01,2.073414421081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,32,balanced,0.4089333216349284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,32,balanced,0.7585386435190836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,32,balanced,0.4975200096766154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,32,balanced,1.1545920372009277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,32,balanced,0.7180426915486654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,32,balanced,2.2678346633911133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,32,balanced,1.3987785975138347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,32,power_law_1.2,0.09327999949455261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,32,power_law_1.2,0.06487039923667907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,32,power_law_1.2,0.055852800607681274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,32,power_law_1.2,0.04750080108642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,32,power_law_1.2,0.05514240264892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,32,power_law_1.2,0.05386880040168762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,32,power_law_1.2,0.04451839923858643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,32,power_law_1.2,0.05799040198326111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,32,power_law_1.2,0.04986239969730377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,32,power_law_1.2,0.05980160236358643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,32,power_law_1.2,0.050988799333572386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,32,power_law_1.2,0.06466559767723083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,32,power_law_1.2,0.05072000026702881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,32,power_law_1.2,0.06784639954566955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,32,power_law_1.2,0.054681599140167236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,32,power_law_1.2,0.07026559710502625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,32,power_law_1.2,0.05631359815597534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,32,power_law_1.2,0.07040640115737914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,32,power_law_1.2,0.05794559717178345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,32,power_law_1.2,0.07071999907493591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,32,power_law_1.2,0.0571008026599884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,32,power_law_1.2,0.07521920204162598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,32,power_law_1.2,0.061504000425338747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,32,power_law_1.2,0.07691519856452941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,32,power_law_1.2,0.06250240206718445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,32,power_law_1.2,0.07914239764213563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,32,power_law_1.2,0.0657472014427185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,32,power_law_1.2,0.08702719807624817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,32,power_law_1.2,0.07102720141410827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,32,power_law_1.2,0.08728320002555848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,32,power_law_1.2,0.07796480059623719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,32,power_law_1.2,0.09535999894142151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,32,power_law_1.2,0.08204799890518188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,32,power_law_1.2,0.10799360275268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,32,power_law_1.2,0.08945279717445373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,32,power_law_1.2,0.10962560176849365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,32,power_law_1.2,0.10372480154037475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,32,power_law_1.2,0.13656959533691407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,32,power_law_1.2,0.08115839958190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,32,power_law_1.2,0.11048320531845093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,32,power_law_1.2,0.07699840068817139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,32,power_law_1.2,0.15961600542068483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,32,power_law_1.2,0.1466815948486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,32,power_law_1.2,0.059680002927780154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,32,power_law_1.2,0.20648961067199706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,32,power_law_1.2,0.16551680564880372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,32,power_law_1.2,0.06261119842529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,32,power_law_1.2,0.23064959049224854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,32,power_law_1.2,0.21020159721374512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,32,power_law_1.2,0.0675711989402771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,32,power_law_1.2,0.3178623914718628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,32,power_law_1.2,0.024063999950885772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,32,power_law_1.2,0.262060809135437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,32,power_law_1.2,0.06572160124778748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,32,power_law_1.2,0.4257535934448242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,32,power_law_1.2,0.03346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,32,power_law_1.2,0.07466880083084107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,32,power_law_1.2,0.3786815881729126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,32,power_law_1.2,0.5887296199798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,32,power_law_1.2,0.021324799954891206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,32,power_law_1.2,0.49977598190307615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,32,power_law_1.2,0.07185279726982116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,32,power_law_1.2,0.6925824165344239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,32,power_law_1.2,0.021113599836826324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,32,power_law_1.2,0.6771455764770508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,32,power_law_1.2,0.022252799570560457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,32,power_law_1.2,0.07384960055351257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,32,power_law_1.2,0.9009792327880859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,32,power_law_1.2,0.9164608001708985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,32,power_law_1.2,0.023296000063419343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,32,power_law_1.2,0.07628160119056701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,32,power_law_1.2,1.4416383743286132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,32,power_law_1.2,0.07534080147743225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,32,power_law_1.2,1.148857593536377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,32,power_law_1.2,0.02499839961528778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,32,power_law_1.2,0.07708160281181335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,32,power_law_1.2,2.7058048248291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,32,power_law_1.2,1.792243194580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,32,power_law_1.2,0.02723200023174286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,32,power_law_1.2,0.07747200131416321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,32,power_law_1.2,0.026425600051879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,32,power_law_1.2,3.6414913177490233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,32,power_law_1.2,0.0785535991191864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,32,power_law_1.2,0.0271807998418808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,32,power_law_1.2,0.08232319951057435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,32,power_law_1.2,0.030873599648475646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,32,power_law_1.2,0.08451200127601624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,32,power_law_1.2,0.03073279857635498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,32,power_law_1.2,0.08926079869270324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,32,power_law_1.2,0.03223040103912354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,32,power_law_1.2,0.10375039577484131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,32,power_law_1.2,0.054150402545928955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.2,0.11444480419158935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,32,power_law_1.2,0.05416319966316223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.2,0.1316480040550232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,32,power_law_1.2,0.0579584002494812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.2,0.15070719718933107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,32,power_law_1.2,0.05702400207519531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.2,0.19113600254058838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,32,power_law_1.2,0.058182400465011594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.2,0.22897920608520508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.2,0.06259199976921082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.2,0.31960320472717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.2,0.07191680073738098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.2,0.09597439765930176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.2,0.41280641555786135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.2,0.11696000099182129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.2,0.6032703876495361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.2,0.1416767954826355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.2,0.7850815773010253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.2,0.1951807975769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,32,power_law_1.2,0.9394047737121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.2,0.24910080432891846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,32,power_law_1.2,1.604800033569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.2,0.3535680055618286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.2,0.49904642105102537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,32,power_law_1.2,3.0022783279418945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,32,power_law_1.2,0.6214591979980468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,32,power_law_1.2,0.9161919593811035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,32,power_law_1.2,1.9340095520019531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,16,balanced,0.02272533377011617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,16,balanced,0.02126399924357732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,16,balanced,0.05193600058555603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,16,balanced,0.0230880007147789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,16,balanced,0.023002666731675465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,16,balanced,0.05343466500441233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,16,balanced,0.023445333043734234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,16,balanced,0.051813334226608276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,16,balanced,0.025477332373460133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,16,balanced,0.056090667843818665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,16,balanced,0.02759466568628947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,16,balanced,0.058186665177345276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,16,balanced,0.02958933264017105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,16,balanced,0.06235733131567637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,16,balanced,0.03139200061559677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,16,balanced,0.06481599807739258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,16,balanced,0.03333866596221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,16,balanced,0.06551999847094218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,16,balanced,0.03207999964555105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,16,balanced,0.06628799935181935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,16,balanced,0.05449600021044413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,16,balanced,0.03548266738653183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,16,balanced,0.06637333333492279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,16,balanced,0.044351999958356224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,16,balanced,0.043968002001444496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,16,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,16,balanced,0.06824000179767609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,16,balanced,0.04562666515509287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,16,balanced,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,16,balanced,0.042261332273483276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,16,balanced,0.06764799853165944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,16,balanced,0.046480000019073486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,16,balanced,0.04574933151404063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,16,balanced,0.04594666759173075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,16,balanced,0.06905066470305125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,16,balanced,0.05251200000445048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,16,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,16,balanced,0.04868266483147939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,16,balanced,0.07025066514809926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,16,balanced,0.07419200241565704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,16,balanced,0.04966933528582255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,16,balanced,0.06685866912206014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,16,balanced,0.0745600014925003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,16,balanced,0.07594666878382365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,16,balanced,0.057162667314211525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,16,balanced,0.07679999868075053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,16,balanced,0.07439466814200084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,16,balanced,0.07666133344173431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,16,balanced,0.057616000374158226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,16,balanced,0.08261866867542267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,16,balanced,0.07877866427103679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,16,balanced,0.07833600044250488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,16,balanced,0.058058664202690125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,16,balanced,0.09814400474230449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,16,balanced,0.08731733759244283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,16,balanced,0.07699200014273326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,16,balanced,0.05950400233268738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,16,balanced,0.11967999736467998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,16,balanced,0.09487467010815938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,16,balanced,0.07788266738255818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,16,balanced,0.060789331793785095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,16,balanced,0.14629333217938742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,16,balanced,0.114847997824351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,16,balanced,0.07868266602357228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,16,balanced,0.06166933476924896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,16,balanced,0.18069867293039957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,16,balanced,0.06258133550484975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,16,balanced,0.13155733545621237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,16,balanced,0.07884799937407176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,16,balanced,0.06434133152167003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,16,balanced,0.1583146651585897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,16,balanced,0.08054399987061818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,16,balanced,0.23085333903630575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,16,balanced,0.0681279997030894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,16,balanced,0.08356266220410664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,16,balanced,0.1917546590169271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,16,balanced,0.08788266777992249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,16,balanced,0.07316266496976216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,16,balanced,0.2910240093866984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,16,balanced,0.08999466896057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,16,balanced,0.07640533149242401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,16,balanced,0.25097066164016724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,16,balanced,0.09574400385220845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,16,balanced,0.43398932615915936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,16,balanced,0.08680533369382222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,16,balanced,0.10289067029953003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,16,balanced,0.30798933903376263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,16,balanced,0.09737599889437358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,16,balanced,0.11320533355077107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,16,balanced,0.5566293398539225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,16,balanced,0.10593600074450175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,16,balanced,0.12626133362452188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,16,balanced,0.4227413336435954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,16,balanced,0.125791996717453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,16,balanced,0.15403733650843301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,16,balanced,0.7121386528015137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,16,balanced,0.14320000012715658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,16,balanced,0.18048532803853354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,16,power_law_1.01,0.08058239817619324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,16,balanced,0.18717867136001587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,16,balanced,0.5385440190633138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,16,balanced,0.2291626731554667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,16,power_law_1.01,0.051097601652145386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,16,power_law_1.01,0.06985599994659424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,16,balanced,0.2238346735636393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,16,power_law_1.01,0.06617599725723267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,16,balanced,0.28488532702128094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,16,power_law_1.01,0.04668799936771393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,16,balanced,1.1110986868540447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,16,power_law_1.01,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,16,balanced,0.303658664226532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,16,balanced,0.6673706372578939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,16,power_law_1.01,0.05143679976463318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,16,power_law_1.01,0.0492031991481781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,16,balanced,0.37823466459910077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,16,balanced,0.38518933455149335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,16,power_law_1.01,0.05596799850463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,16,power_law_1.01,0.051558399200439455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,16,power_law_1.01,0.06007680296897888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,16,power_law_1.01,0.05586559772491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,16,power_law_1.01,0.06791679859161377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,16,balanced,0.5466560125350952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,16,balanced,0.47705598672231037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,16,power_law_1.01,0.05750399827957153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,16,balanced,1.0080479780832927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,16,power_law_1.01,0.0694271981716156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,16,power_law_1.01,0.058764797449111936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,16,power_law_1.01,0.07196800112724304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,16,power_law_1.01,0.0581055998802185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,16,power_law_1.01,0.07274879813194275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,16,balanced,2.1889707247416177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,16,balanced,0.6928373177846273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,16,power_law_1.01,0.059334397315979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,16,power_law_1.01,0.0742464005947113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,16,balanced,0.6047039826711019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,16,power_law_1.01,0.06182399988174438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,16,power_law_1.01,0.07607679963111877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,16,balanced,0.8532479604085287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,16,power_law_1.01,0.06478719711303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,16,power_law_1.01,0.07626240253448487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,16,power_law_1.01,0.06636800169944763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,16,power_law_1.01,0.08154240250587463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,16,power_law_1.01,0.07059839963912964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,16,power_law_1.01,0.07342720031738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,16,power_law_1.01,0.09066879749298096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,16,power_law_1.01,0.08711680173873901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,16,power_law_1.01,0.08184959888458251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,16,balanced,1.969264030456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,16,power_law_1.01,0.09216639995574952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,16,balanced,0.8649600346883138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,16,power_law_1.01,0.022047999501228332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,16,power_law_1.01,0.06787840127944947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,16,balanced,1.3108479976654053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,16,power_law_1.01,0.061382400989532473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,16,power_law_1.01,0.08476160168647766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,16,power_law_1.01,0.09854080080986023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,16,power_law_1.01,0.024409599602222443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,16,power_law_1.01,0.06547840237617493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,16,power_law_1.01,0.09182080030441284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,16,power_law_1.01,0.10609279870986939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,16,power_law_1.01,0.022067199647426605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,16,power_law_1.01,0.06924160122871399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,16,power_law_1.01,0.10958720445632934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,16,power_law_1.01,0.11232000589370728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,16,power_law_1.01,0.02202879935503006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,16,power_law_1.01,0.07808640003204345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,16,power_law_1.01,0.11657600402832032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,16,power_law_1.01,0.12935680150985718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,16,power_law_1.01,0.022393600642681123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,16,power_law_1.01,0.07898880243301391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,16,power_law_1.01,0.1461951971054077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,16,power_law_1.01,0.1494976043701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,16,power_law_1.01,0.02409600019454956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,16,power_law_1.01,0.07454720139503479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,16,power_law_1.01,0.16577279567718506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,16,power_law_1.01,0.1851583957672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,16,power_law_1.01,0.02568959891796112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,16,power_law_1.01,0.08093439936637878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,16,power_law_1.01,0.21154561042785644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,16,balanced,1.6771574020385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,16,power_law_1.01,0.08139520287513732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,16,power_law_1.01,0.2317568063735962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,16,power_law_1.01,0.027084800601005554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,16,power_law_1.01,0.26090240478515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,16,balanced,2.5779199600219727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,16,power_law_1.01,0.08437119722366333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,16,power_law_1.01,0.2891648054122925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,16,power_law_1.01,0.028198400139808656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,16,power_law_1.01,0.38613119125366213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,16,power_law_1.01,0.08167679905891419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,16,power_law_1.01,0.3793344020843506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,16,power_law_1.01,0.02929919958114624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,16,power_law_1.01,0.4633920192718506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,16,power_law_1.01,0.08652160167694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,16,power_law_1.01,0.5035776138305664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,16,power_law_1.01,0.028307199478149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,16,power_law_1.01,0.722431993484497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,16,power_law_1.01,0.09059200286865235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,16,power_law_1.01,0.6870975971221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,16,power_law_1.01,0.0298880010843277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,16,power_law_1.01,0.8625727653503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,16,power_law_1.01,0.09198079705238342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,16,power_law_1.01,0.8174655914306641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,16,power_law_1.01,0.031014400720596313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,16,power_law_1.01,0.10059520006179809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,16,power_law_1.01,0.035750401020050046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,16,power_law_1.01,1.045695972442627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,16,power_law_1.01,1.2880319595336913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,16,power_law_1.01,0.11115520000457764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,16,power_law_1.01,0.03739520013332367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,16,power_law_1.01,1.6863231658935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,16,power_law_1.01,2.346963119506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.01,0.12076159715652465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,16,power_law_1.01,0.0398144006729126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,16,power_law_1.01,3.3558273315429688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.01,0.14468480348587037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,16,power_law_1.01,0.06130560040473938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.01,0.15978879928588868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,16,power_law_1.01,0.06406400203704835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.01,0.20127360820770263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.01,0.070387202501297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.01,0.23694078922271727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.01,0.3392575979232788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.01,0.07680000066757202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.01,0.41699838638305664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.01,0.0868224024772644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.01,0.6084671974182129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.01,0.10920959711074829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.01,0.7511680126190186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.01,0.1383296012878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,16,power_law_1.01,0.9483776092529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.01,0.17470719814300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,16,power_law_1.01,1.3689215660095215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.01,0.22910718917846679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,16,power_law_1.01,3.029817581176758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.01,0.30053119659423827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.01,0.40416641235351564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,16,power_law_1.01,0.4916224002838135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,16,power_law_1.01,0.7417407989501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,16,power_law_1.01,1.517740821838379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,16,power_law_1.2,0.07850880026817322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,16,power_law_1.2,0.06865280270576476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,16,power_law_1.2,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,16,power_law_1.2,0.05183359980583191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,16,power_law_1.2,0.0542527973651886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,16,power_law_1.2,0.05799040198326111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,16,power_law_1.2,0.06707839965820313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,16,power_law_1.2,0.06902400255203248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,16,power_law_1.2,0.07006719708442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,16,power_law_1.2,0.07297919988632202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,16,power_law_1.2,0.07386879920959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,16,power_law_1.2,0.07590399980545044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,16,power_law_1.2,0.05077760219573975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,16,power_law_1.2,0.07953280210494995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,16,power_law_1.2,0.06442880034446716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,16,power_law_1.2,0.0784063994884491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,16,power_law_1.2,0.04492799937725067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,16,power_law_1.2,0.09303680062294006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,16,power_law_1.2,0.04775680005550385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,16,power_law_1.2,0.09000319838523865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,16,power_law_1.2,0.05200639963150024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,16,power_law_1.2,0.0969215989112854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,16,power_law_1.2,0.054176002740859985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,16,power_law_1.2,0.10952320098876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,16,power_law_1.2,0.05847039818763733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,16,power_law_1.2,0.1155776023864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,16,power_law_1.2,0.05805439949035644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,16,power_law_1.2,0.07155200242996215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,16,power_law_1.2,0.13763840198516847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,16,power_law_1.2,0.05917440056800842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,16,power_law_1.2,0.08604159951210022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,16,power_law_1.2,0.16017279624938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,16,power_law_1.2,0.05986559987068176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,16,power_law_1.2,0.022368000447750093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,16,power_law_1.2,0.05851519703865051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,16,power_law_1.2,0.21040000915527343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,16,power_law_1.2,0.06240000128746033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,16,power_law_1.2,0.02407039999961853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,16,power_law_1.2,0.061388802528381345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,16,power_law_1.2,0.0656063973903656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,16,power_law_1.2,0.25615360736846926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,16,power_law_1.2,0.06388480067253113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,16,power_law_1.2,0.021049599349498748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,16,power_law_1.2,0.06564480066299438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,16,power_law_1.2,0.32994558811187746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,16,power_law_1.2,0.06705920100212097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,16,power_law_1.2,0.02160000056028366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,16,power_law_1.2,0.07397119998931885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,16,power_law_1.2,0.07497599720954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,16,power_law_1.2,0.40404481887817384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,16,power_law_1.2,0.075135999917984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,16,power_law_1.2,0.08469759821891784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,16,power_law_1.2,0.022649599611759184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,16,power_law_1.2,0.5849023818969726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,16,power_law_1.2,0.07919999957084656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,16,power_law_1.2,0.08450559973716736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,16,power_law_1.2,0.023647999763488768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,16,power_law_1.2,0.801318359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,16,power_law_1.2,0.08093439936637878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,16,power_law_1.2,0.09427199959754944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,16,power_law_1.2,0.025158399343490602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,16,power_law_1.2,0.0812991976737976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,16,power_law_1.2,1.0467328071594237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,16,power_law_1.2,0.11272319555282592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,16,power_law_1.2,0.0263808012008667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,16,power_law_1.2,0.08352000117301941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,16,power_law_1.2,1.4115839958190919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,16,power_law_1.2,0.12131839990615845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,16,power_law_1.2,0.02807680070400238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,16,power_law_1.2,0.0803391993045807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,16,power_law_1.2,0.1476032018661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,16,power_law_1.2,0.029606398940086365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,16,power_law_1.2,3.5044288635253906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,16,power_law_1.2,0.08487039804458618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,16,power_law_1.2,0.02901119887828827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,16,power_law_1.2,0.17341439723968505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,16,power_law_1.2,0.09048320055007934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,16,power_law_1.2,0.029465600848197937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,16,power_law_1.2,0.21910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,16,power_law_1.2,0.030905601382255555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,16,power_law_1.2,0.09057279825210571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,16,power_law_1.2,0.2853503942489624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,16,power_law_1.2,0.03526400029659271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,16,power_law_1.2,0.09770879745483399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,16,power_law_1.2,0.3948863983154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,16,power_law_1.2,0.03766399919986725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,8,balanced,0.021231998999913532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,16,power_law_1.2,0.11273599863052368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,16,power_law_1.2,0.51976318359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,16,power_law_1.2,0.0403328001499176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.2,0.12510080337524415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,8,balanced,0.022848000129063923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,16,power_law_1.2,0.7934271812438964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,16,power_law_1.2,0.061452800035476686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.2,0.1447487950325012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,8,balanced,0.02372266600529353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,16,power_law_1.2,1.0238080024719238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,16,power_law_1.2,0.06211199760437012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.2,0.17221759557723998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,8,balanced,0.025392000873883564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,16,power_law_1.2,1.204806423187256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.2,0.06871680021286011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,8,balanced,0.05299733579158783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.2,0.20320639610290528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,8,balanced,0.025221332907676697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.2,0.0772159993648529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,16,power_law_1.2,1.9609600067138673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.2,0.25893120765686034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,8,balanced,0.05576533575852712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,8,balanced,0.029274667302767437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.2,0.0878271996974945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.2,0.3524672031402588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,16,power_law_1.2,3.9878719329833983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,8,balanced,0.058005332946777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,8,balanced,0.030559999247392017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.2,0.11430399417877198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.2,0.4368192195892334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,8,balanced,0.058389330903689064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,8,balanced,0.031727999448776245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.2,0.13953280448913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.2,0.6654592037200928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,8,balanced,0.06489066779613495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,8,balanced,0.033615998923778534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.2,0.1769215941429138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.2,0.8474623680114746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,8,balanced,0.035530666510264076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,8,balanced,0.07628266513347626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.2,0.22970879077911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,8,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,16,power_law_1.2,1.1014464378356934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,8,balanced,0.07639466722806294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.2,0.3302848100662231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,8,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,16,power_law_1.2,1.6004608154296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,8,balanced,0.07807466884454091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.2,0.4229440212249756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,8,balanced,0.038746667404969536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,16,power_law_1.2,3.1965824127197267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,8,balanced,0.07804800073305766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,16,power_law_1.2,0.5303296089172364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,8,balanced,0.04174399872620901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,8,balanced,0.0788320004940033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,16,power_law_1.2,0.8821056365966797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,8,balanced,0.048010667165120445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,8,balanced,0.07877333462238312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,16,power_law_1.2,1.672038459777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,8,balanced,0.05403733253479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,8,balanced,0.08116266628106435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,8,balanced,0.05797333518664042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,8,balanced,0.08286400139331818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,8,balanced,0.07738133271535237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,8,balanced,0.08322666585445404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,8,balanced,0.09482133388519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,8,balanced,0.08868267138799031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,8,balanced,0.09106133381525676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,8,balanced,0.1285866697629293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,8,balanced,0.09540800253550212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,8,balanced,0.120688001314799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,8,balanced,0.047930667797724404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,8,balanced,0.10719466209411621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,8,power_law_1.01,0.05045120120048523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,8,balanced,0.047877331574757896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,8,balanced,0.15482667088508606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,8,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,8,power_law_1.01,0.05204480290412903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,8,balanced,0.1176479955514272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,8,balanced,0.05121066669623057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,8,power_law_1.01,0.04830079972743988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,8,balanced,0.18441067139307657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,8,balanced,0.060229331254959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,8,balanced,0.13821333646774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,8,balanced,0.04576533536116282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,8,power_law_1.01,0.05425919890403748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,8,balanced,0.08111999928951263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,8,balanced,0.1560533344745636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,8,balanced,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,8,balanced,0.23093867301940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,8,power_law_1.01,0.05720319747924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,8,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,8,balanced,0.08057599763075511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,8,balanced,0.19141334295272827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,8,power_law_1.01,0.06105599999427795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,8,balanced,0.04980266590913137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,8,balanced,0.29480000336964923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,8,balanced,0.08269866804281871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,8,balanced,0.05406400064627329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,8,power_law_1.01,0.07081599831581116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,8,balanced,0.08452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,8,balanced,0.06367999811967213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,8,power_law_1.01,0.0753600001335144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,8,balanced,0.23307200272878012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,8,balanced,0.08475200335184734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,8,balanced,0.06260799864927928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,8,power_law_1.01,0.07468799948692321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,8,balanced,0.4320106506347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,8,power_law_1.01,0.04684160053730011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,8,balanced,0.08505066235860188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,8,balanced,0.32215466101964313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,8,balanced,0.0663679987192154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,8,power_law_1.01,0.0756608009338379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,8,power_law_1.01,0.053600001335144046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,8,balanced,0.08752533793449402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,8,balanced,0.06569600105285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,8,balanced,0.39715735117594403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,8,balanced,0.5449653466542562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,8,balanced,0.06775466601053874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,8,power_law_1.01,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,8,balanced,0.08725866675376892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,8,power_law_1.01,0.0775871992111206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,8,balanced,0.07015466690063477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,8,balanced,0.08927999933560689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,8,power_law_1.01,0.0494592010974884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,8,power_law_1.01,0.0811456024646759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,8,balanced,0.07146133482456207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,8,balanced,0.0950933297475179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,8,power_law_1.01,0.052928000688552856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,8,balanced,0.5633600155512491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,8,power_law_1.01,0.07095680236816407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,8,balanced,0.6838453610738119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,8,power_law_1.01,0.0834176003932953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,8,balanced,0.09513066212336223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,8,balanced,0.07397866745789845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,8,power_law_1.01,0.02083200067281723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,8,power_law_1.01,0.05867519974708557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,8,power_law_1.01,0.08168320059776306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,8,power_law_1.01,0.08739839792251587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,8,balanced,0.07735466460386912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,8,balanced,0.10115200281143188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,8,balanced,0.7242186864217123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,8,power_law_1.01,0.02263679951429367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,8,power_law_1.01,0.06256639957427979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,8,power_law_1.01,0.062457597255706786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,8,power_law_1.01,0.09076480269432068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,8,balanced,0.08499733606974284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,8,balanced,0.11241599917411804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,8,power_law_1.01,0.02232320010662079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,8,power_law_1.01,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,8,balanced,1.0686506430308025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,8,balanced,0.08737599849700928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,8,power_law_1.01,0.06707839965820313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,8,power_law_1.01,0.09654399752616882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,8,balanced,0.11385599772135417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,8,power_law_1.01,0.023475199937820435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,8,power_law_1.01,0.06641280055046081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,8,balanced,0.8885599772135416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,8,balanced,0.09708266456921895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,8,power_law_1.01,0.07434880137443542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,8,power_law_1.01,0.10619519948959351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,8,power_law_1.01,0.024716800451278685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,8,balanced,0.13167466719945273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,8,power_law_1.01,0.06564480066299438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,8,balanced,0.11040000120798747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,8,power_law_1.01,0.08465279936790467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,8,power_law_1.01,0.1181056022644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,8,power_law_1.01,0.026041600108146667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,8,balanced,0.14682666460673013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,8,power_law_1.01,0.06803839802742004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,8,balanced,0.11952533324559529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,8,power_law_1.01,0.0942911982536316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,8,power_law_1.01,0.12997759580612184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,8,balanced,1.3767414093017578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,8,power_law_1.01,0.027609598636627198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,8,balanced,0.18460800250371298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,8,power_law_1.01,0.07048959732055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,8,balanced,0.14378666877746582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,8,power_law_1.01,0.09724159836769104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,8,power_law_1.01,0.15276800394058226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,8,power_law_1.01,0.028883200883865357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,8,balanced,2.119311968485514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,8,power_law_1.01,0.07225599884986877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,8,power_law_1.01,0.030425599217414855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,8,balanced,0.2218773365020752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,8,power_law_1.01,0.09886720180511474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,8,power_law_1.01,0.17829760313034057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,8,balanced,0.16234667102495828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,8,power_law_1.01,0.07888000011444092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,8,power_law_1.01,0.03147520124912262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,8,power_law_1.01,0.08831999897956848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,8,power_law_1.01,0.21313281059265138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,8,power_law_1.01,0.09499520063400269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,8,power_law_1.01,0.03237119913101196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,8,balanced,0.2163040041923523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,8,power_law_1.01,0.09407359957695008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,8,balanced,0.2878986597061157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,8,power_law_1.01,0.24756479263305664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,8,power_law_1.01,0.09472000002861022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,8,power_law_1.01,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,8,power_law_1.01,0.09903360009193421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,8,balanced,0.26320000489552814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,8,balanced,0.35739731788635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,8,power_law_1.01,0.11593600511550903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,8,power_law_1.01,0.09688959717750549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,8,power_law_1.01,0.3757184028625488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,8,power_law_1.01,0.033670398592948916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,8,power_law_1.01,0.1256767988204956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,8,balanced,0.36161065101623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,8,power_law_1.01,0.10174720287322998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,8,power_law_1.01,0.4253824234008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,8,power_law_1.01,0.03581439852714539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,8,balanced,0.49405864874521893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,8,power_law_1.01,0.15841280221939086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,8,balanced,2.706149419148763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,8,power_law_1.01,0.10016000270843506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,8,power_law_1.01,0.6488383769989013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,8,balanced,0.4606613318125407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,8,power_law_1.01,0.039801600575447085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,8,power_law_1.01,0.18442239761352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,8,power_law_1.01,0.10996479988098144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,8,power_law_1.01,0.8066944122314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,8,balanced,0.6278026501337687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,8,power_law_1.01,0.04249599874019623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,8,power_law_1.01,0.11322879791259766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,8,power_law_1.01,0.25107200145721437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,8,balanced,0.6603519916534424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,8,power_law_1.01,0.9922240257263184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,8,power_law_1.01,0.04759039878845215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,8,power_law_1.01,0.11719679832458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,8,power_law_1.01,0.3219072103500366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,8,power_law_1.01,0.057062399387359616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,8,power_law_1.01,1.5673407554626464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,8,power_law_1.01,0.12865920066833497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,8,power_law_1.01,0.44551677703857423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.01,0.08483200073242188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.01,0.14929280281066895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,8,power_law_1.01,0.5453248023986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,8,power_law_1.01,2.885158348083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.01,0.09799039959907532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.01,0.1696320056915283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,8,balanced,0.7901706695556641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,8,power_law_1.01,0.7469503879547119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.01,0.09742079973220825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,8,balanced,0.8426720301310221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.01,0.19471360445022584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,8,power_law_1.01,1.0332608222961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.01,0.1193727970123291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.01,0.2520064115524292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,8,power_law_1.01,1.209062385559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.01,0.1408959984779358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.01,0.2992768049240112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.01,0.19649920463562012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,8,power_law_1.01,1.924083137512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.01,0.39781761169433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.01,0.24196479320526124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,8,balanced,1.0457173188527424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.01,0.49911041259765626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,8,power_law_1.01,4.000652694702149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.01,0.32785279750823976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.01,0.7017407894134522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.01,0.42755842208862305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,8,balanced,1.1698986689249675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.01,0.9600192070007324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,8,power_law_1.01,0.5201151847839356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,8,power_law_1.01,1.1736767768859864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,8,power_law_1.01,0.7960639953613281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,8,power_law_1.01,1.8138368606567383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,8,power_law_1.01,1.5689215660095215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,8,power_law_1.01,3.6834945678710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,8,balanced,1.6164533297220867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,8,power_law_1.2,0.05063040256500244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,8,power_law_1.2,0.05211520195007324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,8,power_law_1.2,0.04575360119342804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,8,power_law_1.2,0.05100799798965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,8,power_law_1.2,0.05644800066947937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,8,power_law_1.2,0.061247998476028444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,8,balanced,2.2791360219319663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,8,power_law_1.2,0.0704576015472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,8,power_law_1.2,0.07162879705429077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,8,power_law_1.2,0.0735040009021759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,8,power_law_1.2,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,8,power_law_1.2,0.07819520235061646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,8,power_law_1.2,0.08094720244407654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,8,balanced,3.192970593770345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,8,power_law_1.2,0.08206719756126404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,8,power_law_1.2,0.08494719862937927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,8,power_law_1.2,0.09768319725990296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,8,power_law_1.2,0.1023743987083435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,8,power_law_1.2,0.10936959981918334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,8,power_law_1.2,0.11921919584274292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,8,power_law_1.2,0.13081599473953248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,8,power_law_1.2,0.048102399706840514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,8,power_law_1.2,0.1587839961051941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,8,power_law_1.2,0.053830397129058835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,8,power_law_1.2,0.17730560302734374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,8,power_law_1.2,0.0444927990436554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,8,power_law_1.2,0.2270143985748291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,8,power_law_1.2,0.04883840084075928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,8,power_law_1.2,0.3018752098083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,8,power_law_1.2,0.05146880149841308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,8,power_law_1.2,0.40188159942626955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,8,power_law_1.2,0.058278399705886844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,8,power_law_1.2,0.5732672214508057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,8,power_law_1.2,0.06272640228271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,8,power_law_1.2,0.688588809967041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,8,power_law_1.2,0.06394240260124207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,8,power_law_1.2,0.8656448364257813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,8,power_law_1.2,0.06472960114479065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,8,power_law_1.2,0.07055360078811646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,8,power_law_1.2,1.114310359954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,8,power_law_1.2,0.06563199758529663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,8,power_law_1.2,0.07791360020637512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,8,power_law_1.2,1.7983104705810546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,8,power_law_1.2,0.06900479793548583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,8,power_law_1.2,0.06223999857902527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,8,power_law_1.2,0.07154560089111328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,8,power_law_1.2,3.641222381591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,8,power_law_1.2,0.0677183985710144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,8,power_law_1.2,0.07481600046157837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,8,power_law_1.2,0.07206400036811829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,8,power_law_1.2,0.08069120049476623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,8,power_law_1.2,0.08290560245513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,8,power_law_1.2,0.09459199905395507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,8,power_law_1.2,0.09068160057067871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,8,power_law_1.2,0.09544960260391236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,8,power_law_1.2,0.02078080028295517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,8,power_law_1.2,0.09842560291290284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,8,power_law_1.2,0.10019199848175049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,8,power_law_1.2,0.09468799829483032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,8,power_law_1.2,0.02202879935503006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,8,power_law_1.2,0.12108800411224366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,8,power_law_1.2,0.02273920029401779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,8,power_law_1.2,0.09786880016326904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,8,power_law_1.2,0.13317760229110717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,8,power_law_1.2,0.023654399812221526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,8,power_law_1.2,0.09790080189704894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,8,power_law_1.2,0.16250879764556886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,8,power_law_1.2,0.024371199309825897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,8,power_law_1.2,0.09889280200004577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,8,power_law_1.2,0.19557759761810303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,8,power_law_1.2,0.025779199600219727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,8,power_law_1.2,0.02757120132446289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,8,power_law_1.2,0.2502079963684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,8,power_law_1.2,0.10399359464645386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,8,power_law_1.2,0.029625600576400755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,8,power_law_1.2,0.3499135971069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,8,power_law_1.2,0.10296319723129273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,8,power_law_1.2,0.03185920119285583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,8,power_law_1.2,0.43082242012023925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,8,power_law_1.2,0.11286400556564331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,8,power_law_1.2,0.03182719945907593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,8,power_law_1.2,0.591974401473999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,8,power_law_1.2,0.11549439430236816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,8,power_law_1.2,0.031488001346588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,8,power_law_1.2,0.12408959865570068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,8,power_law_1.2,0.03319680094718933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,8,power_law_1.2,0.8827520370483398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,8,power_law_1.2,0.13803520202636718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,8,power_law_1.2,0.03425920009613037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,8,power_law_1.2,1.1485695838928223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.2,0.1525823950767517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,8,power_law_1.2,0.03656319975852966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,8,power_law_1.2,1.3551487922668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.2,0.17630079984664918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,8,power_law_1.2,0.040217599272727965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,8,power_law_1.2,2.0638912200927733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.2,0.20375680923461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,8,power_law_1.2,0.04299519956111908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.2,0.26496000289916993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,8,power_law_1.2,4.639142227172852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,8,power_law_1.2,0.04739840030670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.2,0.312064003944397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,8,power_law_1.2,0.05792639851570129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.2,0.46312317848205564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.2,0.08502399921417236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.2,0.6058879852294922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.2,0.10017919540405273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.2,0.09941759705543518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.2,0.7953663825988769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.2,0.11932799816131592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.2,0.14945919513702394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.2,1.0050880432128906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.2,0.19673600196838378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,8,power_law_1.2,1.3209407806396485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.2,0.24368000030517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,8,power_law_1.2,2.109049606323242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.2,0.3511615991592407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.2,0.4400512218475342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,8,power_law_1.2,3.7606975555419924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,8,power_law_1.2,0.5396671772003174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,8,power_law_1.2,0.8591296195983886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,8,power_law_1.2,1.6845504760742187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,128,balanced,0.033402666449546814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,128,balanced,0.033813332517941795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,128,balanced,0.03342399994532267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,128,balanced,0.033626665671666466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,128,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,128,balanced,0.03347733368476232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,128,balanced,0.03575466573238373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,128,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,128,balanced,0.03334933271010717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,128,balanced,0.03773866593837738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,128,balanced,0.0354666660229365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,128,balanced,0.03363200028737386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,128,balanced,0.037471999724706016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,128,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,128,balanced,0.03719999889532725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,128,balanced,0.03739733248949051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,128,balanced,0.03733866661787033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,4,balanced,0.045706664522488914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,128,balanced,0.0360000009338061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,128,balanced,0.037765334049860634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,128,balanced,0.03709333389997482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,4,balanced,0.04781333108743032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,128,balanced,0.03777066618204117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,128,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,4,balanced,0.0487360010544459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,4,balanced,0.04402133325735728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,128,balanced,0.039834665755430855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,128,balanced,0.03786666691303253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,4,balanced,0.05540800094604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,4,balanced,0.04776533444722494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,128,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,128,balanced,0.039066667358080544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,4,balanced,0.06482133269309998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,4,balanced,0.048021331429481506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,128,balanced,0.04015466570854187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,128,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,4,balanced,0.05190399785836538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,4,balanced,0.09066667159398396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,128,balanced,0.0422026664018631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,4,balanced,0.05967999994754791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,128,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,4,power_law_1.01,0.04894079864025116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,4,balanced,0.09478400150934856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,128,balanced,0.04167999823888143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,4,balanced,0.07881600161393483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,128,balanced,0.043578664461771645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,4,power_law_1.01,0.05245440006256104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,4,balanced,0.09681600332260132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,128,balanced,0.0425546665986379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,4,balanced,0.07774933179219563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,128,balanced,0.041493333876132965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,4,power_law_1.01,0.05086719989776611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,4,balanced,0.0958666702111562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,128,balanced,0.042117332418759666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,4,balanced,0.08249066770076752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,128,balanced,0.043061330914497375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,4,power_law_1.01,0.055980801582336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,4,power_law_1.01,0.0494271993637085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,4,balanced,0.09751466910044353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,128,balanced,0.04574933151404063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,4,balanced,0.08294933537642162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,128,balanced,0.04398400088151296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,4,power_law_1.01,0.06435199975967407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,4,balanced,0.09853333234786987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,4,power_law_1.01,0.05310080051422119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,4,power_law_1.01,0.06972159743309021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,4,balanced,0.0846666693687439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,128,balanced,0.04587733248869578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,128,balanced,0.04693333307902018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,4,balanced,0.09989866614341736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,4,power_law_1.01,0.08297600150108338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,4,balanced,0.08256533245245616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,4,power_law_1.01,0.05110399723052979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,128,balanced,0.04762133459250132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,4,power_law_1.01,0.08880640268325805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,4,balanced,0.1011786659558614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,128,balanced,0.04791999856630961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,4,balanced,0.08591467142105103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,4,power_law_1.01,0.05259519815444946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,4,power_law_1.01,0.08967679738998413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,128,balanced,0.05077866713205973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,4,balanced,0.10494400064150493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,4,power_law_1.2,0.04940159916877747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,128,balanced,0.0517493337392807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,4,balanced,0.09030399719874065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,4,power_law_1.01,0.06039040088653565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,4,power_law_1.01,0.08893439769744874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,128,balanced,0.05421866476535797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,4,balanced,0.1095199982325236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,4,power_law_1.2,0.05231999754905701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,4,balanced,0.09514666597048442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,4,power_law_1.01,0.0656063973903656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,4,power_law_1.01,0.09271039962768554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,128,balanced,0.05809600154558817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,4,balanced,0.11155733466148376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,128,balanced,0.05995733539263407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,4,power_law_1.2,0.05071359872817993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,4,balanced,0.10156800349553426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,4,power_law_1.01,0.07397119998931885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,4,power_law_1.01,0.09594240188598632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,4,balanced,0.11705066760381062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,128,balanced,0.06598933537801106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,4,power_law_1.2,0.05633280277252197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,4,balanced,0.10610666871070862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,4,power_law_1.01,0.07791360020637512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,128,balanced,0.07492266595363617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,4,power_law_1.01,0.0987712025642395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,4,balanced,0.13156800468762717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,4,power_law_1.2,0.06048640012741089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,4,balanced,0.11637333035469055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,4,power_law_1.01,0.08030080199241638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,128,balanced,0.07886399825414021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,4,power_law_1.01,0.10053759813308716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,128,balanced,0.09708799918492635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,4,power_law_1.2,0.06710399985313416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,4,balanced,0.14050133029619852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,4,power_law_1.01,0.08023679852485657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,4,balanced,0.13216533263524374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,4,power_law_1.2,0.049465599656105044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,4,power_law_1.01,0.11004159450531006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,128,balanced,0.09303999940554301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,4,power_law_1.2,0.0791808009147644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,4,balanced,0.16476266582806906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,128,balanced,0.1338933308919271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,4,power_law_1.01,0.08542079925537109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,4,balanced,0.1446346640586853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,4,power_law_1.01,0.11600639820098876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,4,power_law_1.2,0.05382400155067444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,4,power_law_1.2,0.08920320272445678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,4,balanced,0.18126400311787924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,4,power_law_1.01,0.09118080139160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,4,power_law_1.01,0.1198848009109497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,4,balanced,0.17293866475423178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,4,power_law_1.2,0.08885759711265565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,4,power_law_1.2,0.04924800097942352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,4,balanced,0.24641066789627075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,128,balanced,0.16703466574350992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,128,balanced,0.10921600461006165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,4,power_law_1.01,0.09422720074653626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,4,power_law_1.01,0.1434048056602478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,4,balanced,0.20011732975641885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,4,power_law_1.2,0.08892160058021545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,4,power_law_1.2,0.05178880095481873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,4,balanced,0.28633066018422443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,4,power_law_1.2,0.09278079867362976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,4,power_law_1.01,0.15601279735565185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,4,power_law_1.01,0.10138880014419556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,4,power_law_1.2,0.05754240155220032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,4,balanced,0.28147733211517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,4,power_law_1.2,0.09709439873695373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,4,power_law_1.01,0.18760960102081298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,4,balanced,0.4097706476847331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,4,power_law_1.01,0.11200000047683716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,4,power_law_1.2,0.0631168007850647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,128,balanced,0.2002133329709371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,4,power_law_1.2,0.09907839894294738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,4,power_law_1.01,0.2213184118270874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,4,balanced,0.3309813340504964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,4,power_law_1.01,0.11830400228500366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,128,balanced,0.15588266650835672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,4,power_law_1.2,0.07268480062484742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,4,power_law_1.2,0.10410239696502685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,4,power_law_1.01,0.26740479469299316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,4,power_law_1.01,0.11980160474777221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,4,balanced,0.5060960054397583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,4,power_law_1.2,0.11203199625015259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,4,power_law_1.01,0.3392895936965942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,4,power_law_1.01,0.13808000087738037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,4,power_law_1.2,0.07608320116996765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,4,power_law_1.2,0.11676160097122193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,4,balanced,0.4824213186899821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,4,power_law_1.01,0.4664639949798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,4,power_law_1.2,0.07820799946784973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,4,power_law_1.01,0.1571328043937683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,4,power_law_1.2,0.12907520532608033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,4,balanced,0.7475679715474447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,4,power_law_1.01,0.5968512058258056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,4,power_law_1.2,0.08011519908905029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,4,power_law_1.01,0.2007744073867798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,4,balanced,0.607802669207255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,4,power_law_1.2,0.14820480346679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,128,balanced,0.2924373348553975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,4,power_law_1.01,0.861580753326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,4,power_law_1.2,0.08622080087661743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,4,power_law_1.01,0.2330496072769165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,4,power_law_1.2,0.16275839805603026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,4,balanced,0.9448693593343099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,4,power_law_1.01,1.109388828277588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,4,power_law_1.2,0.19304319620132446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,4,power_law_1.2,0.09103999733924865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,4,power_law_1.01,0.31843841075897217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,4,balanced,0.886853297551473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,4,power_law_1.2,0.2207808017730713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,4,power_law_1.01,1.332748794555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,4,power_law_1.2,0.09620479941368103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,4,power_law_1.01,0.3898943901062012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,4,power_law_1.2,0.275545597076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,4,power_law_1.01,2.063091278076172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,4,power_law_1.2,0.1019968032836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,128,balanced,0.271781325340271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,4,balanced,1.1568053563435872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,4,power_law_1.2,0.3603519916534424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,4,power_law_1.01,0.5532224178314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,4,power_law_1.01,4.19251823425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,4,balanced,1.162394682566325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,4,power_law_1.2,0.11020159721374512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,4,power_law_1.2,0.5025343894958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,4,power_law_1.01,0.6926464080810547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,4,power_law_1.2,0.6951168060302735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,4,power_law_1.2,0.1176576018333435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,4,power_law_1.01,1.0400383949279786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,4,power_law_1.2,0.9505599975585938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,128,balanced,0.5391733249028524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,4,power_law_1.2,0.12321280241012574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,4,power_law_1.01,1.321337604522705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,4,power_law_1.2,1.1818559646606446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,4,power_law_1.2,0.14536319971084594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,4,power_law_1.01,1.6540800094604493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,4,balanced,1.4370026588439941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,4,balanced,1.7616373697916667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,4,power_law_1.2,1.4747584342956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,4,power_law_1.2,0.16117119789123535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,4,power_law_1.01,2.490969657897949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,4,power_law_1.2,0.20373759269714356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,4,power_law_1.01,4.952684783935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,4,power_law_1.2,2.1412160873413084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,4,power_law_1.2,0.24495999813079833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,4,power_law_1.2,0.3234240055084229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,4,power_law_1.2,4.232627105712891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,4,power_law_1.2,0.39655680656433107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,4,power_law_1.2,0.56561279296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,4,balanced,2.2083306312561035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,4,power_law_1.2,0.7477888107299805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,4,power_law_1.2,1.0582655906677245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,4,power_law_1.2,1.4225600242614747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,4,power_law_1.2,1.8843263626098632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,4,power_law_1.2,2.686374473571777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,4,power_law_1.2,5.163059234619141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,4,balanced,3.476437250773112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,4,balanced,4.341792106628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,128,balanced,0.03380800038576126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,128,balanced,0.03389333436886469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,128,balanced,0.033759998778502144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,128,balanced,0.035887998839219414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,128,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,128,balanced,0.017765333255132038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,128,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,128,balanced,0.019146667172511418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,128,balanced,0.03749333322048187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,128,balanced,0.019178666174411774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,128,balanced,0.02128533273935318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,128,balanced,0.03612799942493439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,128,balanced,0.03159466634194056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,128,balanced,0.03755733370780945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,128,balanced,0.02716800073782603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,128,balanced,0.0377813329299291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,128,balanced,0.027322667340437572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,128,power_law_1.01,0.032518398761749265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,128,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,128,balanced,0.026917333404223125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,128,power_law_1.01,0.03278079926967621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,128,balanced,0.03947199881076813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,128,balanced,0.026949333647886913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,128,power_law_1.01,0.03255679905414581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,128,balanced,0.027658666173617046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,128,power_law_1.01,0.034508800506591795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,128,balanced,0.03786666691303253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,128,balanced,0.027984000742435455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,128,power_law_1.01,0.03482879996299744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,128,balanced,0.03841600070397059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,128,power_law_1.01,0.03594880104064942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,128,balanced,0.029370665550231934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,128,balanced,0.04200533529122671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,128,balanced,0.02934933453798294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,128,power_law_1.01,0.03587839901447296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,128,balanced,0.04196266829967499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,128,balanced,0.029279999434947968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,128,power_law_1.01,0.035622400045394895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,128,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,128,balanced,0.041834667325019836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,128,power_law_1.01,0.03619199991226196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,128,balanced,0.03044266750415166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,128,power_law_1.01,0.03596160113811493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,128,balanced,0.044026667873064675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,128,balanced,0.03145600110292435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,128,power_law_1.01,0.03658879995346069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,128,balanced,0.04608533283074697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,128,balanced,0.03339733431736628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,128,power_law_1.01,0.03706879913806915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,128,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,128,power_law_1.01,0.03726080060005188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,128,balanced,0.04607999821503957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,128,balanced,0.04004266609748205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,128,power_law_1.01,0.037567999958992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,128,balanced,0.05019199848175049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,128,balanced,0.0415040006240209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,128,power_law_1.01,0.04020479917526245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,128,balanced,0.05410666763782501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,128,power_law_1.01,0.0403903990983963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,128,balanced,0.04790933430194855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,128,power_law_1.01,0.04095999896526337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,128,balanced,0.0583840012550354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,128,power_law_1.01,0.042054399847984314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,128,power_law_1.01,0.043315199017524716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,128,balanced,0.05819733440876007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,128,power_law_1.01,0.04620800018310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,128,power_law_1.01,0.05027840137481689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,128,balanced,0.07036266724268596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,128,power_law_1.01,0.03158400058746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,128,power_law_1.01,0.032287999987602234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,128,power_law_1.01,0.05989760160446167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,128,balanced,0.0746506651242574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,128,power_law_1.01,0.03150720000267029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,128,balanced,0.07903466622034709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,128,power_law_1.01,0.03192319869995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,128,power_law_1.01,0.07271680235862732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,128,power_law_1.01,0.03188480138778686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,128,power_law_1.01,0.017177599668502807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,128,balanced,0.09504533807436626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,128,power_law_1.01,0.032518398761749265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,128,balanced,0.10377599795659383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,128,power_law_1.01,0.09234560132026673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,128,power_law_1.01,0.033004799485206605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,128,power_law_1.01,0.01714559942483902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,128,power_law_1.01,0.03407999873161316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,1,128,power_law_1.2,0.03213439881801605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,1,128,power_law_1.2,0.030745598673820495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,128,power_law_1.01,0.1094655990600586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,128,balanced,0.1300266683101654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,128,power_law_1.01,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,128,power_law_1.01,0.0177279993891716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,128,power_law_1.01,0.034534400701522826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,1,128,power_law_1.2,0.03198719918727875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,128,balanced,0.12498133381207784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,128,power_law_1.01,0.15002880096435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,1,128,power_law_1.2,0.0308351993560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,128,power_law_1.01,0.03491199910640717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,128,power_law_1.01,0.019283199310302736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,128,power_law_1.01,0.03561600148677826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,1,128,power_law_1.2,0.03221760094165802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,1,128,power_law_1.2,0.030739200115203858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,128,power_law_1.01,0.18958719968795776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,128,balanced,0.16434666514396667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,128,power_law_1.01,0.03514240086078644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,128,power_law_1.01,0.03107840120792389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,128,power_law_1.01,0.03532159924507141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,1,128,power_law_1.2,0.033395200967788696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,1,128,power_law_1.2,0.03308799862861633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,128,balanced,0.14713600277900696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,128,power_law_1.01,0.21920640468597413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,128,power_law_1.01,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,128,power_law_1.01,0.026169601082801818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,128,power_law_1.01,0.035724800825119016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,128,balanced,0.1982133388519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,1,128,power_law_1.2,0.03454079926013946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,1,128,power_law_1.2,0.03323520123958588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,128,power_law_1.01,0.3547391891479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,128,power_law_1.01,0.026035198569297792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,128,power_law_1.01,0.03581439852714539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,128,power_law_1.01,0.03578239977359772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,1,128,power_law_1.2,0.034969601035118106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,1,128,power_law_1.2,0.034720000624656674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,128,power_law_1.01,0.025791999697685242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,128,power_law_1.01,0.7378880023956299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,128,power_law_1.01,0.03592959940433502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,128,power_law_1.01,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,1,128,power_law_1.2,0.03553279936313629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,1,128,power_law_1.2,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,128,power_law_1.01,0.026208001375198364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,128,balanced,0.20582934220631918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,128,power_law_1.01,0.03628160059452057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,128,power_law_1.01,0.03650560081005096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,1,128,power_law_1.2,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,128,balanced,0.2993653416633606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,1,128,power_law_1.2,0.034560000896453856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,128,power_law_1.01,0.027276799082756042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,128,power_law_1.01,0.03670400083065033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,128,power_law_1.01,0.03686400055885315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,1,128,power_law_1.2,0.03581439852714539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,1,128,power_law_1.2,0.03500159978866577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,128,power_law_1.01,0.03702400028705597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,128,power_law_1.01,0.027059200406074523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,128,power_law_1.01,0.037273600697517395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,1,128,power_law_1.2,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,128,power_law_1.01,0.03768320083618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,1,128,power_law_1.2,0.03513599932193756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,128,power_law_1.01,0.02678399980068207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,128,power_law_1.01,0.037215998768806456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,1,128,power_law_1.2,0.03586559891700745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,128,power_law_1.01,0.041068801283836366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,1,128,power_law_1.2,0.03633280098438263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,128,power_law_1.01,0.027327999472618103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,128,power_law_1.01,0.04019840061664581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,1,128,power_law_1.2,0.036262398958206175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,128,power_law_1.01,0.04085760116577149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,128,balanced,0.3728213310241699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,1,128,power_law_1.2,0.036447998881340024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,128,power_law_1.01,0.027795198559761047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,128,power_law_1.01,0.0403903990983963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,128,balanced,0.5732160011927286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,1,128,power_law_1.2,0.03665919899940491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,128,power_law_1.01,0.041068801283836366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,1,128,power_law_1.2,0.03717760145664215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,128,power_law_1.01,0.028595200181007384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,128,power_law_1.01,0.04147840142250061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,1,128,power_law_1.2,0.03734399974346161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,128,power_law_1.01,0.043910399079322815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,1,128,power_law_1.2,0.037894400954246524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,128,power_law_1.01,0.02963840067386627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,1,128,power_law_1.2,0.04019840061664581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,128,power_law_1.01,0.043321600556373595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,128,power_law_1.01,0.046515199542045596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,1,128,power_law_1.2,0.03978239893913269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,128,power_law_1.01,0.02999039888381958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,1,128,power_law_1.2,0.03960959911346436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,128,power_law_1.01,0.045875200629234315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,128,power_law_1.01,0.05120639801025391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,1,128,power_law_1.2,0.04046719968318939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,128,power_law_1.01,0.032262399792671204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,1,128,power_law_1.2,0.04080640077590943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,128,power_law_1.01,0.051577597856521606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,128,power_law_1.01,0.05633280277252197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,1,128,power_law_1.2,0.0429504007101059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,128,power_law_1.01,0.03439359962940216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,128,power_law_1.01,0.05731840133666992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,1,128,power_law_1.2,0.042771199345588685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,128,power_law_1.01,0.06722559928894042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,1,128,power_law_1.2,0.04479359984397888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,128,power_law_1.01,0.037990400195121767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,128,power_law_1.01,0.06499199867248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,1,128,power_law_1.2,0.045772799849510194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,128,power_law_1.01,0.07893760204315185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,1,128,power_law_1.2,0.04832639992237091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,128,power_law_1.01,0.03956480026245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,128,power_law_1.01,0.07530239820480347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,1,128,power_law_1.2,0.05211520195007324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,128,power_law_1.01,0.04506239891052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,128,power_law_1.01,0.1171839952468872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,1,128,power_law_1.2,0.055257600545883176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,1,128,power_law_1.2,0.056831997632980344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,128,power_law_1.01,0.09994879961013795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,128,power_law_1.01,0.053363198041915895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,128,power_law_1.01,0.1456447958946228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,1,128,power_law_1.2,0.06629120111465454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,1,128,power_law_1.2,0.031814399361610415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,1,128,power_law_1.2,0.0692799985408783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,128,power_law_1.01,0.11984000205993653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,128,power_law_1.01,0.07066239714622498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,128,power_law_1.01,0.20149118900299073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,1,128,power_law_1.2,0.07912319898605347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,1,128,power_law_1.2,0.032076799869537355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,1,128,power_law_1.2,0.08700159788131714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,128,power_law_1.01,0.16582399606704712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,128,power_law_1.01,0.2599551916122437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,128,power_law_1.01,0.08730880022048951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,1,128,power_law_1.2,0.09941120147705078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,1,128,power_law_1.2,0.03214719891548157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,1,128,power_law_1.2,0.11773439645767211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,128,power_law_1.01,0.21274240016937257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,128,power_law_1.01,0.33164799213409424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,128,power_law_1.01,0.1225600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,1,128,power_law_1.2,0.14781440496444703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,1,128,power_law_1.2,0.033881598711013795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,1,128,power_law_1.2,0.15083520412445067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,128,power_law_1.01,0.1539903998374939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,1,128,power_law_1.2,0.034329599142074584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,1,128,power_law_1.2,0.19386240243911743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,128,power_law_1.01,0.4865471839904785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,1,128,power_law_1.2,0.1754047989845276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,128,power_law_1.01,0.25401599407196046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,128,power_law_1.01,0.1846336007118225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,1,128,power_law_1.2,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,1,128,power_law_1.2,0.26942079067230223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,128,power_law_1.01,0.9904767990112304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,1,128,power_law_1.2,0.2628671884536743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,128,power_law_1.01,0.2841984033584595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,128,power_law_1.01,0.3720383882522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,1,128,power_law_1.2,0.0350271999835968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,1,128,power_law_1.2,0.3339008092880249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,1,128,power_law_1.2,0.32880001068115233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,128,power_law_1.01,0.5445888042449951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,128,power_law_1.01,0.7999360084533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,1,128,power_law_1.2,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,1,128,power_law_1.2,0.5167424201965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,1,128,power_law_1.2,0.43230719566345216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,1,128,power_law_1.2,0.017113600671291352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,1,128,power_law_1.2,0.03561600148677826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,1,128,power_law_1.2,0.9985600471496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,1,128,power_law_1.2,0.6367231845855713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,1,128,power_law_1.2,0.017875200510025023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,1,128,power_law_1.2,0.03550719916820526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,1,128,power_law_1.2,1.3046784400939941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,1,128,power_law_1.2,0.01839359998703003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,1,128,power_law_1.2,0.03619840145111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,1,128,power_law_1.2,0.019782400131225585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,1,128,power_law_1.2,0.03675520122051239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,1,128,power_law_1.2,0.03136639893054962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,1,128,power_law_1.2,0.036697599291801455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,1,128,power_law_1.2,0.024915200471878052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,1,128,power_law_1.2,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,1,128,power_law_1.2,0.024639999866485594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,1,128,power_law_1.2,0.04010879993438721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,1,128,power_law_1.2,0.02542720139026642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,1,128,power_law_1.2,0.04005120098590851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,1,128,power_law_1.2,0.02571519911289215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,1,128,power_law_1.2,0.04188799858093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,1,128,power_law_1.2,0.02573440074920654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,1,128,power_law_1.2,0.044409599900245664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,1,128,power_law_1.2,0.026015999913215637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,1,128,power_law_1.2,0.04862079918384552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,1,128,power_law_1.2,0.028255999088287354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,1,128,power_law_1.2,0.055955201387405396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,1,128,power_law_1.2,0.02807680070400238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,1,128,power_law_1.2,0.06058239936828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,1,128,power_law_1.2,0.02797439992427826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,1,128,power_law_1.2,0.07549440264701843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,1,128,power_law_1.2,0.029151999950408937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,1,128,power_law_1.2,0.0866815984249115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,1,128,power_law_1.2,0.029606398940086365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,1,128,power_law_1.2,0.12108800411224366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,1,128,power_law_1.2,0.029875200986862183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,1,128,power_law_1.2,0.15352319478988646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,1,128,power_law_1.2,0.03288320004940033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,1,128,power_law_1.2,0.2077888011932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,1,128,power_law_1.2,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,1,128,power_law_1.2,0.2701632022857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,1,128,power_law_1.2,0.03813120126724243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,1,128,power_law_1.2,0.32930560111999513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,1,128,power_law_1.2,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,1,128,power_law_1.2,0.5207424163818359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,1,128,power_law_1.2,0.04604800045490265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,1,128,power_law_1.2,1.006662368774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,1,128,power_law_1.2,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,1,128,power_law_1.2,0.07178879976272583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,1,128,power_law_1.2,0.09123200178146362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,1,128,power_law_1.2,0.12777600288391114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,1,128,power_law_1.2,0.1610111951828003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,1,128,power_law_1.2,0.1956480026245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,1,128,power_law_1.2,0.29662721157073973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,1,128,power_law_1.2,0.5696000099182129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,64,balanced,0.02536533276240031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,64,balanced,0.025663999219735462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,64,balanced,0.027248000105222065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,64,balanced,0.025231999655564625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,64,balanced,0.025237334271272022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,64,balanced,0.027600000301996868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,64,balanced,0.02788266787926356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,64,balanced,0.02752000093460083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,64,balanced,0.027306665976842243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,64,balanced,0.02923733244339625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,64,balanced,0.02943466603755951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,64,balanced,0.027957332630952198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,64,balanced,0.029317334294319153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,64,balanced,0.032272001107533775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,64,balanced,0.03535466641187668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,64,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,64,balanced,0.03405333310365677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,64,balanced,0.035690667728583016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,64,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,64,balanced,0.035786665976047516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,64,balanced,0.037818667789300285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,64,balanced,0.040752001106739044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,64,balanced,0.041850666205088295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,64,balanced,0.04814933240413666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,64,balanced,0.05073600014050802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,64,balanced,0.06410666803518932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,64,balanced,0.07445333401362102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,64,balanced,0.08554133772850037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,64,balanced,0.11859200398127238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,64,power_law_1.01,0.02223999947309494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,64,power_law_1.01,0.02260479927062988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,64,power_law_1.01,0.0230335995554924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,64,power_law_1.01,0.02351360023021698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,64,power_law_1.01,0.024332800507545473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,64,power_law_1.01,0.025407999753952026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,64,power_law_1.01,0.025510400533676147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,64,power_law_1.01,0.026688000559806822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,64,power_law_1.01,0.027795198559761047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,64,power_law_1.01,0.02958720028400421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,64,balanced,0.01911466692884763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,64,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,64,power_law_1.01,0.031488001346588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,64,balanced,0.18515199422836304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,64,balanced,0.019493332753578823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,64,balanced,0.03366400053103765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,64,power_law_1.01,0.03078399896621704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,64,balanced,0.0191040001809597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,64,balanced,0.03329599897066752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,64,power_law_1.01,0.03159680068492889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,64,balanced,0.03138133386770884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,64,balanced,0.01899733394384384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,64,balanced,0.03377600014209747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,64,power_law_1.01,0.030508801341056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,64,balanced,0.031311998764673867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,64,balanced,0.018917333334684372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,64,balanced,0.03549333413441976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,64,power_law_1.01,0.03325439989566803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,64,balanced,0.029279999434947968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,64,power_law_1.01,0.03293440043926239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,64,balanced,0.021082667013009388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,64,balanced,0.03557866563399633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,64,power_law_1.01,0.03461120128631592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,64,balanced,0.029482667644818623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,64,power_law_1.01,0.018432000279426576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,64,power_law_1.01,0.032927998900413515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,64,balanced,0.02109866589307785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,64,balanced,0.036090667049090065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,64,power_law_1.01,0.03592320084571839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,64,balanced,0.03140799949566523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,64,power_law_1.01,0.018060800433158875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,64,power_law_1.01,0.03266560137271881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,64,balanced,0.035775999228159584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,64,balanced,0.027327999472618103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,64,power_law_1.01,0.037574398517608645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,64,power_law_1.01,0.01852159947156906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,64,balanced,0.03276800115903219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,64,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,64,power_law_1.01,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,64,balanced,0.027098665634791057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,64,power_law_1.01,0.04165120124816894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,64,power_law_1.01,0.016953599452972413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,64,balanced,0.03342933456103007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,64,balanced,0.03605333218971888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,64,power_law_1.01,0.035308799147605895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,64,balanced,0.02924266705910365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,64,power_law_1.01,0.046982398629188536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,64,power_law_1.01,0.017785599827766417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,64,power_law_1.01,0.03080959916114807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,64,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,64,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,64,power_law_1.01,0.03628160059452057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,64,balanced,0.027610667049884796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,64,power_law_1.01,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,64,power_law_1.01,0.020025600492954255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,64,power_law_1.01,0.029043200612068176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,64,balanced,0.03326933334271113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,64,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,64,power_law_1.01,0.0363072007894516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,64,balanced,0.029232000311215717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,64,power_law_1.01,0.05316479802131653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,64,power_law_1.01,0.0201664000749588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,64,power_law_1.01,0.028479999303817748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,64,balanced,0.03345066557327906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,64,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,64,power_law_1.01,0.03653759956359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,64,balanced,0.02922666569550832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,64,power_law_1.01,0.06041600108146668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,64,power_law_1.01,0.025363200902938844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,64,power_law_1.01,0.029203200340270997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,64,balanced,0.03376533339420954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,64,balanced,0.037952000896135964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,64,power_law_1.01,0.03650560081005096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,64,balanced,0.02924266705910365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,64,power_law_1.01,0.0813759982585907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,64,power_law_1.01,0.025791999697685242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,64,power_law_1.01,0.02991360127925873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,64,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,64,balanced,0.03976533313592275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,64,power_law_1.01,0.03691520094871521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,64,balanced,0.02961066613594691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,64,power_law_1.01,0.025971201062202454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,64,power_law_1.01,0.09434880018234253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,64,power_law_1.01,0.030604800581932066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,64,balanced,0.033802665770053864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,64,balanced,0.04177600145339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,64,power_law_1.01,0.03776000142097473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,64,balanced,0.029578665892283123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,64,power_law_1.01,0.026451200246810913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,64,power_law_1.01,0.13962240219116212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,64,power_law_1.01,0.03030399978160858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,64,balanced,0.03583999971548716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,64,balanced,0.041834667325019836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,64,power_law_1.01,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,64,balanced,0.03133333226044973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,64,power_law_1.01,0.026240000128746034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,64,power_law_1.01,0.1592512011528015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,64,power_law_1.01,0.03051519989967346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,64,balanced,0.03823466598987579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,64,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,64,power_law_1.01,0.0380160003900528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,64,power_law_1.01,0.02728320062160492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,64,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,64,power_law_1.01,0.21692800521850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,64,power_law_1.01,0.030822399258613586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,64,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,64,balanced,0.04377066592375437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,64,power_law_1.01,0.03845120072364807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,64,power_law_1.01,0.027340799570083618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,64,power_law_1.01,0.031667199730873105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,64,balanced,0.035536001125971474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,64,power_law_1.01,0.2813760042190552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,64,balanced,0.03948266555865606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,64,power_law_1.01,0.04172160029411316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,64,power_law_1.01,0.0282943993806839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,64,balanced,0.04420266548792521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,64,power_law_1.01,0.032627201080322264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,64,balanced,0.04192000130812327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,64,power_law_1.01,0.5208191871643066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,64,balanced,0.039834665755430855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,64,power_law_1.01,0.041580799221992495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,64,power_law_1.01,0.02917119860649109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,64,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,64,power_law_1.01,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,64,power_law_1.01,0.029894399642944335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,64,balanced,0.04452799757321676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,64,balanced,0.041663999358812966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,64,power_law_1.01,0.04297600090503693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,64,power_law_1.01,0.033817601203918454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,64,balanced,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,64,power_law_1.01,0.0319680005311966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,64,power_law_1.2,0.023155200481414794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,64,power_law_1.01,0.0462336003780365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,64,balanced,0.04197866717974345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,64,balanced,0.051818668842315674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,64,power_law_1.01,0.03552640080451965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,64,power_law_1.01,0.0338239997625351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,64,balanced,0.05380799869696299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,64,power_law_1.2,0.02412160038948059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,64,power_law_1.01,0.048895999789237976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,64,balanced,0.045834665497144066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,64,balanced,0.06252799928188324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,64,power_law_1.01,0.03813759982585907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,64,power_law_1.01,0.03847039937973022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,64,power_law_1.2,0.02417919933795929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,64,balanced,0.06442666550477345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,64,power_law_1.01,0.054630398750305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,64,balanced,0.04977599779764811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,64,balanced,0.07860800127188365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,64,power_law_1.01,0.03940480053424835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,64,power_law_1.01,0.04063999950885773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,64,power_law_1.2,0.025305598974227905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,64,balanced,0.0724533349275589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,64,power_law_1.01,0.05934720039367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,64,balanced,0.05231999854246775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,64,power_law_1.01,0.04037120044231415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,64,power_law_1.01,0.0477183997631073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,64,balanced,0.10063466429710388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,64,power_law_1.2,0.025721600651741026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,64,power_law_1.01,0.07008000016212464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,64,power_law_1.01,0.044095999002456664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,64,balanced,0.06805866460005443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,64,power_law_1.01,0.05441920161247253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,64,balanced,0.09269332885742188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,64,power_law_1.2,0.026572799682617186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,64,power_law_1.01,0.0807424008846283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,64,power_law_1.01,0.04503040015697479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,64,balanced,0.13614933689435324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,64,power_law_1.01,0.07432320117950439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,64,power_law_1.01,0.10936319828033447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,64,power_law_1.2,0.02682879865169525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,64,balanced,0.08892266949017842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,64,power_law_1.01,0.055103999376296994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,64,balanced,0.10915199915568034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,64,power_law_1.01,0.0925055980682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,64,power_law_1.01,0.13124480247497558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,64,power_law_1.2,0.02855679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,64,power_law_1.01,0.054995197057724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,64,power_law_1.01,0.12780159711837769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,64,power_law_1.01,0.1927616000175476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,64,power_law_1.2,0.029504001140594482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,64,balanced,0.17319466670354208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,64,power_law_1.01,0.06474239826202392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,64,balanced,0.12329600254694621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,64,power_law_1.01,0.15358079671859742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,64,power_law_1.01,0.2106175899505615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,64,power_law_1.2,0.030668801069259642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,64,balanced,0.12757866581281027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,64,power_law_1.01,0.07786239981651306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,64,power_law_1.01,0.19560960531234742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,64,power_law_1.01,0.2753920078277588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,64,balanced,0.20844266812006632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,64,power_law_1.2,0.031411200761795044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,64,balanced,0.15315199891726175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,64,power_law_1.01,0.11453440189361572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,64,power_law_1.01,0.28885760307312014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,64,power_law_1.01,0.4380864143371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,64,power_law_1.2,0.03162879943847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,64,power_law_1.01,0.1391487956047058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,64,power_law_1.01,0.5708223819732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,64,power_law_1.2,0.0320576012134552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,64,power_law_1.01,0.8093440055847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,64,power_law_1.01,0.20202240943908692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,64,power_law_1.2,0.03146879971027374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,64,balanced,0.18570667505264282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,64,balanced,0.1811093290646871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,64,power_law_1.01,0.2425407886505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,64,power_law_1.2,0.034355199337005614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,64,balanced,0.3168053428332011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,64,power_law_1.01,0.3109760046005249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,64,power_law_1.2,0.03563520014286041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,64,power_law_1.2,0.03650560081005096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,64,power_law_1.01,0.45505919456481936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,64,power_law_1.2,0.038047999143600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,64,power_law_1.2,0.047539201378822324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,64,power_law_1.01,0.928006362915039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,64,power_law_1.2,0.04869759976863861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,64,power_law_1.2,0.05629439949989319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,64,power_law_1.2,0.07000319957733155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,64,balanced,0.2637653350830078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,64,power_law_1.2,0.08254079818725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,64,power_law_1.2,0.11767040491104126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,64,power_law_1.2,0.1441151976585388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,64,power_law_1.2,0.2250432014465332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,64,power_law_1.2,0.3096832036972046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,64,power_law_1.2,0.29492480754852296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,64,power_law_1.2,0.02903040051460266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,64,power_law_1.2,0.548038387298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,64,power_law_1.2,0.02717440128326416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,64,power_law_1.2,0.02730880081653595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,64,power_law_1.2,1.2008064270019532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,64,power_law_1.2,0.02908799946308136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,64,power_law_1.2,0.029068800806999206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,64,balanced,0.3036373257637024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,64,power_law_1.2,0.031046399474143983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,64,power_law_1.2,0.031071999669075014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,64,balanced,0.6092640161514282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,64,power_law_1.2,0.03108479976654053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,64,power_law_1.2,0.031718400120735166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,64,power_law_1.2,0.0318336009979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,64,balanced,0.4679946502049764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,64,power_law_1.2,0.03287039995193482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,64,power_law_1.2,0.03306879997253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,64,power_law_1.2,0.03441919982433319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,64,power_law_1.2,0.03521279990673065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,64,power_law_1.2,0.040556800365447995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,64,power_law_1.2,0.0404992014169693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,64,power_law_1.2,0.040550398826599124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,64,power_law_1.2,0.04579200148582459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,64,power_law_1.2,0.05085440278053284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,64,power_law_1.2,0.05621119737625122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,64,power_law_1.2,0.06113920211791992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,64,power_law_1.2,0.07395840287208558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,64,power_law_1.2,0.09281920194625855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,64,power_law_1.2,0.1588736057281494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,64,power_law_1.2,0.17772799730300903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,64,power_law_1.2,0.28580479621887206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,64,power_law_1.2,0.3829823970794678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,64,power_law_1.2,0.45304322242736816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,64,power_law_1.2,0.8043392181396485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,64,power_law_1.2,1.6112127304077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,64,power_law_1.2,0.03312000036239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,64,power_law_1.2,0.03265919983386993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,64,power_law_1.2,0.033030399680137636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,64,power_law_1.2,0.017740799486637114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,64,power_law_1.2,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,64,power_law_1.2,0.035724800825119016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,64,power_law_1.2,0.01772159934043884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,64,power_law_1.2,0.036620798707008365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,64,power_law_1.2,0.01820800006389618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,64,power_law_1.2,0.036524799466133115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,64,power_law_1.2,0.01704320013523102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,64,power_law_1.2,0.03706879913806915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,64,power_law_1.2,0.017318400740623473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,64,power_law_1.2,0.03708159923553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,64,power_law_1.2,0.01969279944896698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,64,power_law_1.2,0.03713920116424561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,64,power_law_1.2,0.019897599518299103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,64,power_law_1.2,0.0377920001745224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,64,power_law_1.2,0.025798401236534117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,64,power_law_1.2,0.03779839873313904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,64,power_law_1.2,0.02587519884109497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,64,power_law_1.2,0.03856000006198883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,64,power_law_1.2,0.025830399990081788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,64,power_law_1.2,0.039366400241851805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,64,power_law_1.2,0.02630400061607361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,64,power_law_1.2,0.04226559996604919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,64,power_law_1.2,0.026719999313354493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,64,power_law_1.2,0.04186240136623383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,64,power_law_1.2,0.02728320062160492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,64,power_law_1.2,0.0438400000333786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,64,power_law_1.2,0.027526399493217467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,64,power_law_1.2,0.04776319861412048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,64,power_law_1.2,0.02815999984741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,64,power_law_1.2,0.05199360251426697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,64,power_law_1.2,0.029120001196861266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,64,power_law_1.2,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,64,power_law_1.2,0.02953599989414215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,64,power_law_1.2,0.06446719765663148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,64,power_law_1.2,0.03198080062866211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,64,power_law_1.2,0.08170239925384522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,64,power_law_1.2,0.034118399024009705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,64,power_law_1.2,0.09134079813957215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,64,power_law_1.2,0.039078399538993835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,64,power_law_1.2,0.12561919689178466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,64,power_law_1.2,0.04180479943752289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,64,power_law_1.2,0.1725823998451233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,64,power_law_1.2,0.04798080027103424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,64,power_law_1.2,0.22982399463653563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,64,power_law_1.2,0.05720319747924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,64,power_law_1.2,0.3648319959640503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,64,power_law_1.2,0.07587199807167053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,64,power_law_1.2,0.44766721725463865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,64,power_law_1.2,0.09683200120925903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,64,power_law_1.2,0.6659008026123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,64,power_law_1.2,0.13244160413742065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,64,power_law_1.2,1.529535961151123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,64,power_law_1.2,0.16424959897994995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,64,power_law_1.2,0.2030656099319458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,64,power_law_1.2,0.30995199680328367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,64,power_law_1.2,0.6046207904815674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,32,balanced,0.025221332907676697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,32,balanced,0.02309333284695943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,32,balanced,0.023354666928450268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,32,balanced,0.023728000621000927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,32,balanced,0.02309866746266683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,32,balanced,0.025114665428797405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,32,balanced,0.025434667865435284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,32,balanced,0.02533866713444392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,32,balanced,0.02510400116443634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,32,balanced,0.02573866645495097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,32,balanced,0.027456000447273254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,32,balanced,0.027349332968393963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,32,balanced,0.02735999971628189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,32,balanced,0.031290667752424874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,32,balanced,0.03355200091997782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,32,balanced,0.03348266581694285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,32,balanced,0.033530667424201965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,32,balanced,0.035605333745479584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,32,balanced,0.037717332442601524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,32,power_law_1.01,0.025209599733352663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,32,balanced,0.03734400123357773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,32,power_law_1.01,0.024422399699687958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,32,balanced,0.03962666789690653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,32,power_law_1.01,0.023097600042819976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,32,balanced,0.0313226655125618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,32,power_law_1.01,0.024742400646209715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,32,balanced,0.04204266766707102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,32,balanced,0.029232000311215717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,32,power_law_1.01,0.025465598702430724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,32,power_law_1.01,0.033228799700737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,32,power_law_1.2,0.025945600867271424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,32,balanced,0.029498666524887085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,32,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,32,power_law_1.01,0.025945600867271424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,32,power_law_1.01,0.030368000268936157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,32,power_law_1.2,0.023180800676345825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,32,balanced,0.031002665559450786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,32,power_law_1.01,0.026502400636672974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,32,balanced,0.052154665191968284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,32,power_law_1.01,0.029075199365615846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,32,power_law_1.2,0.02308479994535446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,32,balanced,0.031471999982992806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,32,power_law_1.01,0.026726400852203368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,32,power_law_1.01,0.029868799448013305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,32,power_law_1.2,0.024396799504756927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,32,balanced,0.033226666351159416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,32,balanced,0.05786666770776113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,32,power_law_1.01,0.02933120131492615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,32,power_law_1.01,0.030803200602531434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,32,power_law_1.2,0.025171199440956117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,32,power_law_1.2,0.03342719972133636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,32,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,32,power_law_1.01,0.030220800638198854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,32,power_law_1.01,0.03123840093612671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,32,balanced,0.07226133346557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,32,power_law_1.2,0.026163199543952943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,32,balanced,0.032399999598662056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,32,power_law_1.2,0.0286080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,32,power_law_1.01,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,32,power_law_1.01,0.0318015992641449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,32,power_law_1.2,0.02680320143699646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,32,balanced,0.0335359995563825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,32,balanced,0.09334933757781982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,32,power_law_1.2,0.029043200612068176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,32,power_law_1.01,0.03451519906520843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,32,power_law_1.01,0.032128000259399415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,32,power_law_1.2,0.028537601232528687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,32,balanced,0.03389333436886469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,32,power_law_1.2,0.02863360047340393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,32,power_law_1.01,0.03560959994792938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,32,power_law_1.01,0.03243519961833954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,32,power_law_1.2,0.02967039942741394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,32,balanced,0.10943999886512756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,32,balanced,0.033376000821590424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,32,power_law_1.2,0.029791998863220214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,32,power_law_1.01,0.03107840120792389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,32,power_law_1.01,0.03238399922847748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,32,power_law_1.2,0.030553600192070006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,32,balanced,0.035349334279696144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,32,power_law_1.2,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,32,power_law_1.01,0.033542400598526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,32,power_law_1.01,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,32,power_law_1.2,0.03454079926013946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,32,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,32,power_law_1.2,0.031007999181747438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,32,balanced,0.14426133036613464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,32,power_law_1.01,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,32,power_law_1.01,0.034118399024009705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,32,power_law_1.2,0.03583999872207642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,32,balanced,0.035690667728583016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,32,power_law_1.2,0.03187200129032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,32,power_law_1.01,0.03563520014286041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,32,power_law_1.01,0.034944000840187076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,32,power_law_1.2,0.036883199214935304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,32,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,32,power_law_1.2,0.03191039860248566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,32,power_law_1.01,0.03849599957466125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,32,power_law_1.01,0.036134400963783266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,32,power_law_1.2,0.03139840066432953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,32,balanced,0.03762666632731756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,32,power_law_1.2,0.03193599879741669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,32,power_law_1.01,0.040608000755310056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,32,power_law_1.01,0.03854719996452331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,32,balanced,0.23907732963562012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,32,power_law_1.2,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,32,balanced,0.039493332306543984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,32,power_law_1.2,0.0328575998544693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,32,power_law_1.01,0.044633600115776065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,32,power_law_1.01,0.04023680090904236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,32,power_law_1.2,0.034483200311660765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,32,balanced,0.039808000127474465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,32,power_law_1.2,0.034041601419448855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,32,power_law_1.01,0.049235200881958006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,32,power_law_1.01,0.04101119935512543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,32,power_law_1.2,0.03592959940433502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,32,balanced,0.04161066561937332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,32,power_law_1.2,0.03432320058345795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,32,power_law_1.01,0.0434688001871109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,32,power_law_1.01,0.0585919976234436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,32,power_law_1.2,0.0406464010477066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,32,balanced,0.04357333481311798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,32,power_law_1.2,0.03526400029659271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,32,power_law_1.01,0.04622719883918762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,32,power_law_1.01,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,32,power_law_1.2,0.04261119961738587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,32,balanced,0.04794133206208547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,32,power_law_1.01,0.05070719718933105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,32,power_law_1.2,0.03928320109844208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,32,power_law_1.01,0.09473919868469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,32,power_law_1.2,0.051571202278137204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,32,balanced,0.052058666944503784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,32,power_law_1.01,0.05999360084533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,32,power_law_1.2,0.040403199195861814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,32,power_law_1.01,0.099891197681427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,32,power_law_1.2,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,32,power_law_1.01,0.0690559983253479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,32,balanced,0.058149332801500954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,32,power_law_1.2,0.043007999658584595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,32,power_law_1.01,0.13957760334014893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,32,power_law_1.2,0.06269440054893494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,32,power_law_1.01,0.08480640053749085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,32,power_law_1.2,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,32,power_law_1.01,0.17614079713821412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,32,balanced,0.07317866881688435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,32,power_law_1.2,0.07720959782600403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,32,power_law_1.01,0.11685760021209717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,32,power_law_1.2,0.05011199712753296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,32,power_law_1.01,0.22901759147644044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,32,power_law_1.2,0.10963200330734253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,32,power_law_1.01,0.13987200260162352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,32,balanced,0.10357333223025005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,32,power_law_1.2,0.05665919780731201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,32,power_law_1.2,0.1322175979614258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,32,power_law_1.01,0.35068159103393554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,32,power_law_1.01,0.21926400661468506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,32,power_law_1.2,0.06309760212898255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,32,power_law_1.2,0.2549439907073975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,8,balanced,0.03127466638882955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,32,power_law_1.01,0.25727999210357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,32,power_law_1.01,0.5358848094940185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,8,balanced,0.03161066770553589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,32,balanced,0.14288533727327982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,32,power_law_1.2,0.08016639947891235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,32,power_law_1.2,0.2212928056716919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,8,balanced,0.03229333211978277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,32,power_law_1.01,0.34099841117858887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,8,balanced,0.031744000812371574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,32,power_law_1.2,0.09529600143432618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,32,power_law_1.2,0.31529600620269777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,8,balanced,0.03086400032043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,32,power_law_1.01,0.5091839790344238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,8,balanced,0.03134933362404505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,32,balanced,0.17127466201782227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,32,power_law_1.2,0.1347264051437378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,32,power_law_1.2,0.5482240200042725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,8,balanced,0.029578665892283123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,32,power_law_1.01,1.1303808212280273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,8,balanced,0.03009066730737686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,32,power_law_1.2,0.17463040351867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,32,power_law_1.2,1.1488896369934083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,8,balanced,0.031343999008337654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,8,balanced,0.03133866687615713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,32,power_law_1.2,0.3244800090789795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,8,balanced,0.03143999973932902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,32,balanced,0.2074293295542399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,8,balanced,0.0316746657093366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,32,power_law_1.2,0.3488895893096924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,8,balanced,0.031471999982992806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,8,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,32,power_law_1.2,0.4711359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,8,balanced,0.031583999594052635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,8,balanced,0.03175999969244003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,32,power_law_1.2,0.9029376029968261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,8,balanced,0.03145600110292435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,8,balanced,0.031701333820819855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,8,balanced,0.03156266609827677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,32,power_law_1.2,1.902592086791992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,8,balanced,0.033743999898433685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,8,balanced,0.03352533280849457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,32,balanced,0.2995199958483378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,8,balanced,0.03376533339420954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,8,balanced,0.03166399896144867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,8,balanced,0.03431999931732813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,8,balanced,0.03178133318821589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,8,balanced,0.03370666752258936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,8,balanced,0.03148799886306127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,8,balanced,0.03382933388153712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,8,balanced,0.035487999518712364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,8,balanced,0.03982399900754293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,8,balanced,0.03576533248027166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,8,balanced,0.037871999045213066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,8,balanced,0.037605332831541695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,8,balanced,0.039706667264302574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,8,balanced,0.038831998904546104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,8,balanced,0.04346133271853129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,8,balanced,0.04437333345413208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,8,balanced,0.03956799954175949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,8,balanced,0.05028266708056132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,8,balanced,0.04366933306058248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,8,balanced,0.054010664423306785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,8,balanced,0.04438933233420054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,32,balanced,0.5364053249359131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,8,balanced,0.0664160003264745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,8,balanced,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,8,balanced,0.09435733159383138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,8,balanced,0.05413866539796194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,8,balanced,0.13376532991727194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,8,balanced,0.07426133255163829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,8,balanced,0.15981333454449972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,8,balanced,0.08879466851552327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,8,balanced,0.2192266583442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,8,balanced,0.11944533387819926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,8,power_law_1.01,0.0293503999710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,8,balanced,0.26317334175109863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,8,balanced,0.13006400068600973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,8,power_law_1.01,0.030137598514556885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,8,power_law_1.01,0.026899200677871705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,8,balanced,0.3268853425979614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,8,balanced,0.16611199577649435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,8,power_law_1.01,0.029344001412391664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,8,power_law_1.01,0.028505599498748778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,8,power_law_1.01,0.02985599935054779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,8,balanced,0.48420798778533936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,8,power_law_1.01,0.029772800207138062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,8,balanced,0.21620800097783408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,8,power_law_1.01,0.030144000053405763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,8,power_law_1.01,0.031052801012992858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,8,balanced,0.9126933415730795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,8,power_law_1.01,0.03159680068492889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,8,balanced,0.38867199420928955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,8,power_law_1.01,0.031539198756217954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,8,power_law_1.01,0.03304960131645203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,8,power_law_1.01,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,8,power_law_1.01,0.03298560082912445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,8,power_law_1.01,0.03777279853820801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,8,power_law_1.01,0.03763839900493622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,8,power_law_1.01,0.041203200817108154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,8,power_law_1.01,0.050310397148132326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,8,power_law_1.01,0.05255039930343628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,8,power_law_1.01,0.06525440216064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,8,power_law_1.01,0.07123839855194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,8,power_law_1.01,0.09397119879722596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,8,power_law_1.01,0.10708479881286621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,8,power_law_1.01,0.14504319429397583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,8,power_law_1.01,0.17050880193710327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,8,power_law_1.01,0.2459264039993286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,8,power_law_1.01,0.33754239082336424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,8,power_law_1.01,0.41242241859436035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,8,power_law_1.01,0.565177583694458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,8,power_law_1.01,1.0058815956115723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,8,power_law_1.01,0.029094401001930236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,8,power_law_1.01,0.028601598739624024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,8,power_law_1.01,0.02800639867782593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,8,power_law_1.01,0.029811200499534608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,8,power_law_1.01,0.030272001028060914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,8,power_law_1.01,0.030585598945617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,8,power_law_1.01,0.03149439990520477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,8,power_law_1.01,0.0318336009979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,8,power_law_1.01,0.033107200264930726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,8,power_law_1.01,0.033478400111198424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,8,power_law_1.01,0.03383679986000061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,8,power_law_1.01,0.0367680013179779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,8,power_law_1.01,0.03699840009212494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,8,power_law_1.01,0.036262398958206175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,8,power_law_1.01,0.04095999896526337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,8,power_law_1.01,0.04220159947872162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,14336,2,8,32,8,power_law_1.2,0.031404799222946166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,8,power_law_1.01,0.04615040123462677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,8,power_law_1.01,0.05559679865837097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,14336,2,8,32,8,power_law_1.2,0.03131519854068756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,14336,2,8,32,8,power_law_1.2,0.029369598627090453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,8,power_law_1.01,0.055327999591827395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,14336,2,8,32,8,power_law_1.2,0.028352001309394838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,8,power_law_1.01,0.06741759777069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,14336,2,8,32,8,power_law_1.2,0.029497599601745604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,14336,2,8,32,8,power_law_1.2,0.02930560111999512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,8,power_law_1.01,0.07495679855346679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,14336,2,8,32,8,power_law_1.2,0.029497599601745604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,14336,2,8,32,8,power_law_1.2,0.028332799673080444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,8,power_law_1.01,0.09944959878921508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,14336,2,8,32,8,power_law_1.2,0.03054080009460449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,8,power_law_1.01,0.14684799909591675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,14336,2,8,32,8,power_law_1.2,0.030329599976539612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,14336,2,8,32,8,power_law_1.2,0.030483201146125793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,8,power_law_1.01,0.19682559967041016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,14336,2,8,32,8,power_law_1.2,0.030432000756263733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,14336,2,8,32,8,power_law_1.2,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,8,power_law_1.01,0.23671040534973145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,14336,2,8,32,8,power_law_1.2,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,14336,2,8,32,8,power_law_1.2,0.031001600623130798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,14336,2,8,32,8,power_law_1.2,0.031788799166679385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,8,power_law_1.01,0.34714879989624026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,14336,2,8,32,8,power_law_1.2,0.03178240060806274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,14336,2,8,32,8,power_law_1.2,0.032742398977279666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,14336,2,8,32,8,power_law_1.2,0.03134720027446747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,14336,2,8,32,8,power_law_1.2,0.033369600772857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,8,power_law_1.01,0.4567423820495605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,14336,2,8,32,8,power_law_1.2,0.033555200695991515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,14336,2,8,32,8,power_law_1.2,0.033504000306129454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,14336,2,8,32,8,power_law_1.2,0.033766400814056394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,8,power_law_1.01,0.518400001525879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,14336,2,8,32,8,power_law_1.2,0.03418880105018616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,14336,2,8,32,8,power_law_1.2,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,14336,2,8,32,8,power_law_1.2,0.03550080060958862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,8,power_law_1.01,0.8565823554992675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,14336,2,8,32,8,power_law_1.2,0.03703039884567261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,14336,2,8,32,8,power_law_1.2,0.03622399866580963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,8,power_law_1.01,1.651251220703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,14336,2,8,32,8,power_law_1.2,0.040652799606323245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,14336,2,8,32,8,power_law_1.2,0.03596799969673157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,14336,2,8,32,8,power_law_1.2,0.04095360040664673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,14336,2,8,32,8,power_law_1.2,0.04237439930438995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,14336,2,8,32,8,power_law_1.2,0.04848000109195709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,14336,2,8,32,8,power_law_1.2,0.04286080002784729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,14336,2,8,32,8,power_law_1.2,0.05373439788818359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,14336,2,8,32,8,power_law_1.2,0.04661119878292084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,14336,2,8,32,8,power_law_1.2,0.0621504008769989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,14336,2,8,32,8,power_law_1.2,0.05397760272026062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,14336,2,8,32,8,power_law_1.2,0.07575039863586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,14336,2,8,32,8,power_law_1.2,0.05729920268058777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,14336,2,8,32,8,power_law_1.2,0.06783360242843628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,14336,2,8,32,8,power_law_1.2,0.0942143976688385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,14336,2,8,32,8,power_law_1.2,0.07742720246315002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,14336,2,8,32,8,power_law_1.2,0.10896639823913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,14336,2,8,32,8,power_law_1.2,0.10729600191116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,14336,2,8,32,8,power_law_1.2,0.14595839977264405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,14336,2,8,32,8,power_law_1.2,0.15079679489135742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,14336,2,8,32,8,power_law_1.2,0.19631999731063843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,14336,2,8,32,8,power_law_1.2,0.16885759830474853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,14336,2,8,32,8,power_law_1.2,0.2429503917694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,14336,2,8,32,8,power_law_1.2,0.24873600006103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,14336,2,8,32,8,power_law_1.2,0.34628479480743407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,14336,2,8,32,8,power_law_1.2,0.3380095958709717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,14336,2,8,32,8,power_law_1.2,0.44407038688659667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,14336,2,8,32,8,power_law_1.2,0.41054720878601075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,14336,2,8,32,8,power_law_1.2,0.5631872177124023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,14336,2,8,32,8,power_law_1.2,0.5940288066864013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,14336,2,8,32,8,power_law_1.2,0.8195520401000976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,14336,2,8,32,8,power_law_1.2,1.0504447937011718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,14336,2,8,32,8,power_law_1.2,1.6549631118774415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,8,balanced,0.01929066702723503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,8,balanced,0.01903466631968816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,8,balanced,0.019194666296243668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,8,balanced,0.03661333272854487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,8,balanced,0.019253333409627277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,8,balanced,0.039813332259655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,8,balanced,0.01987733319401741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,8,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,8,balanced,0.021477334201335907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,8,balanced,0.035786665976047516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,8,balanced,0.021546666820844013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,8,balanced,0.03772266705830892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,8,balanced,0.02346666653951009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,8,balanced,0.03763733307520548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,8,balanced,0.037952000896135964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,8,balanced,0.02384000023206075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,8,balanced,0.03782399992148081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,8,balanced,0.02401600033044815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,8,balanced,0.03764266769091288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,8,balanced,0.03357866654793421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,8,balanced,0.04005866746107737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,8,balanced,0.03444266567627589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,8,balanced,0.037808001041412354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,8,balanced,0.03551999976237615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,8,balanced,0.03982399900754293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,8,balanced,0.03583466758330663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,8,balanced,0.03990933299064636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,8,balanced,0.03973866750796636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,8,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,8,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,8,balanced,0.03734933336575826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,8,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,8,balanced,0.040394666294256844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,8,balanced,0.046069333950678505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,8,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,8,balanced,0.048063998421033226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,8,balanced,0.04633066554864248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,8,balanced,0.052202666799227394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,8,balanced,0.05471999943256378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,8,balanced,0.0544053316116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,8,balanced,0.06457066535949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,8,balanced,0.062309334675470986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,8,balanced,0.08861333131790161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,8,balanced,0.08095466593901317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,8,balanced,0.10116799672444661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,8,balanced,0.08937066793441772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,8,balanced,0.13779200116793314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,8,balanced,0.11335999766985576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,8,balanced,0.1748159925142924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,8,balanced,0.13673599561055502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,8,balanced,0.24690133333206177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,8,balanced,0.18147200345993042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,8,balanced,0.2301973303159078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,8,balanced,0.3080693284670512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,8,balanced,0.27775466442108154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,8,balanced,0.38154133160909015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,8,balanced,0.40161065260569256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,8,balanced,0.589573343594869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,8,balanced,1.145690679550171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,8,balanced,0.7543413639068604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,8,balanced,0.03752533346414566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,8,balanced,0.038405333956082664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,8,balanced,0.033376000821590424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,8,balanced,0.033488000432650246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,8,balanced,0.031397332747777305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,8,balanced,0.033413333197434746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,8,balanced,0.033759998778502144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,8,balanced,0.03182400017976761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,8,balanced,0.031530665854612984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,8,balanced,0.03345066557327906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,8,balanced,0.03143466760714849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,8,balanced,0.033520000676314034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,8,balanced,0.03324799984693527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,8,balanced,0.03426666557788849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,8,balanced,0.0342399999499321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,8,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,8,balanced,0.03408000121514002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,8,balanced,0.035546667873859406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,8,balanced,0.03367999941110611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,8,balanced,0.03356799980004629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,8,balanced,0.03381866713364919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,8,power_law_1.01,0.03692800104618073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,8,balanced,0.035429333647092186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,8,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,8,balanced,0.035536001125971474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,8,power_law_1.01,0.03237760066986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,8,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,8,balanced,0.0378506655494372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,8,power_law_1.01,0.03155199885368347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,8,power_law_1.01,0.029440000653266907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,8,balanced,0.035573333501815796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,8,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,8,power_law_1.01,0.031065601110458373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,8,power_law_1.01,0.03671680092811584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,8,power_law_1.01,0.018796800076961516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,8,power_law_1.01,0.030803200602531434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,8,balanced,0.03551999976237615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,8,balanced,0.040896000961462654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,8,power_law_1.01,0.03198719918727875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,8,power_law_1.01,0.0361407995223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,8,balanced,0.03749333322048187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,8,power_law_1.01,0.029267200827598573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,8,power_law_1.01,0.01863040030002594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,8,balanced,0.0415786678592364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,8,power_law_1.01,0.03178240060806274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,8,power_law_1.01,0.036831998825073244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,8,balanced,0.04181866844495138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,8,power_law_1.01,0.031046399474143983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,8,power_law_1.01,0.01865600049495697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,8,balanced,0.0436160018046697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,8,power_law_1.01,0.03818880021572113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,8,power_law_1.01,0.03113600015640259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,8,power_law_1.01,0.03189760148525238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,8,balanced,0.03997866561015447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,8,power_law_1.01,0.019142399728298187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,8,power_law_1.01,0.03916800022125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,8,balanced,0.047781333327293396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,8,power_law_1.01,0.032313600182533264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,8,power_law_1.01,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,8,balanced,0.04363200068473816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,8,power_law_1.01,0.019200000166893005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,8,power_law_1.01,0.039136001467704774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,8,balanced,0.05186133086681366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,8,power_law_1.01,0.03331199884414673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,8,power_law_1.01,0.03276160061359405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,8,balanced,0.04576533536116282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,8,power_law_1.01,0.019014400243759156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,8,power_law_1.01,0.03906559944152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,8,power_law_1.01,0.03372800052165985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,8,balanced,0.0664160003264745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,8,power_law_1.01,0.03436160087585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,8,balanced,0.04985066751639048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,8,power_law_1.01,0.018886399269104005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,8,power_law_1.01,0.03914879858493805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,8,power_law_1.01,0.03487359881401062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,8,balanced,0.0730453332265218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,8,power_law_1.01,0.034745600819587705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,8,power_law_1.01,0.02222079932689667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,8,balanced,0.055786664287249245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,8,power_law_1.01,0.03924480080604553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,8,power_law_1.01,0.036294400691986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,8,balanced,0.09544000029563904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,8,power_law_1.01,0.035097599029541016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,8,power_law_1.01,0.022707200050354003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,8,power_law_1.01,0.039155200123786926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,8,balanced,0.06818133095900218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,8,power_law_1.01,0.03646079897880554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,8,power_law_1.01,0.022457599639892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,8,power_law_1.01,0.03576320111751556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,8,balanced,0.10263466835021973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,8,power_law_1.01,0.03741439878940582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,8,power_law_1.01,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,8,power_law_1.01,0.03694719970226288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,8,power_law_1.01,0.03375360071659088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,8,balanced,0.10611733794212341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,8,power_law_1.01,0.041920000314712526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,8,power_law_1.01,0.03686400055885315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,8,power_law_1.01,0.04154239892959595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,8,balanced,0.14362133542696634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,8,power_law_1.01,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,8,power_law_1.01,0.045228800177574156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,8,power_law_1.01,0.03924480080604553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,8,balanced,0.12175466616948445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,8,power_law_1.01,0.047014400362968445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,8,power_law_1.01,0.04371199905872345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,8,power_law_1.01,0.04128639996051788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,8,power_law_1.01,0.03403519988059998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,8,power_law_1.01,0.05560960173606873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,8,balanced,0.16538666685422262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,8,power_law_1.01,0.046649599075317384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,8,balanced,0.17661333084106445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,8,power_law_1.01,0.06096640229225159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,8,power_law_1.01,0.04257279932498932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,16384,2,8,32,8,power_law_1.2,0.03994239866733551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,16384,2,8,32,8,power_law_1.2,0.029497599601745604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,8,power_law_1.01,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,8,power_law_1.01,0.049318400025367734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,8,balanced,0.2028426726659139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,8,power_law_1.01,0.07873280048370361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,8,power_law_1.01,0.04657280147075653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,16384,2,8,32,8,power_law_1.2,0.03871999979019165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,8,balanced,0.2273013393084208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,8,power_law_1.01,0.03498879969120026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,8,power_law_1.01,0.05812479853630066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,16384,2,8,32,8,power_law_1.2,0.03150080144405365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,8,power_law_1.01,0.08773760199546814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,8,balanced,0.28550400336583454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,8,power_law_1.01,0.04753279983997345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,16384,2,8,32,8,power_law_1.2,0.030393600463867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,8,power_law_1.01,0.06555520296096802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,8,power_law_1.01,0.03628160059452057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,8,power_law_1.01,0.11508480310440064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,16384,2,8,32,8,power_law_1.2,0.029465600848197937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,8,balanced,0.2896053393681844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,8,power_law_1.01,0.05002239942550659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,16384,2,8,32,8,power_law_1.2,0.03195520043373108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,8,balanced,0.3633386691411336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,8,power_law_1.01,0.08115199804306031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,8,power_law_1.01,0.14279040098190307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,8,power_law_1.01,0.055283200740814206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,8,power_law_1.01,0.039526399970054624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,16384,2,8,32,8,power_law_1.2,0.031001600623130798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,16384,2,8,32,8,power_law_1.2,0.03226880133152008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,8,power_law_1.01,0.21234560012817383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,8,power_law_1.01,0.10490239858627319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.01,0.06079360246658325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,8,power_law_1.01,0.042796799540519716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,16384,2,8,32,8,power_law_1.2,0.030777600407600404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,8,balanced,0.4520159959793091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,16384,2,8,32,8,power_law_1.2,0.03253119885921478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,8,balanced,0.5236639976501465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,8,power_law_1.01,0.23624958992004394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,8,power_law_1.01,0.14526079893112182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.01,0.07263360023498536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.01,0.04563199877738953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,16384,2,8,32,8,power_law_1.2,0.03319680094718933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,8,power_law_1.01,0.3416575908660889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,8,balanced,0.6695573329925537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,16384,2,8,32,8,power_law_1.2,0.0332863986492157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,8,power_law_1.01,0.16993279457092286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.01,0.09047679901123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.01,0.05311359763145447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,8,power_law_1.01,0.42226557731628417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,16384,2,8,32,8,power_law_1.2,0.03284479975700379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,16384,2,8,32,8,power_law_1.2,0.03361920118331909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,8,power_law_1.01,0.25088000297546387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.01,0.11496319770812988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,8,power_law_1.01,0.5269184112548828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.01,0.0617792010307312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,16384,2,8,32,8,power_law_1.2,0.03420799970626831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,8,balanced,1.289247989654541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,16384,2,8,32,8,power_law_1.2,0.03461120128631592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.01,0.13118720054626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,8,power_law_1.01,0.29665920734405515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,8,power_law_1.01,0.778713607788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.01,0.08644480109214783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,16384,2,8,32,8,power_law_1.2,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,16384,2,8,32,8,power_law_1.2,0.03477759957313538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.01,0.18237440586090087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,8,power_law_1.01,1.5299776077270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,8,power_law_1.01,0.44572157859802247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.01,0.09948800206184387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,16384,2,8,32,8,power_law_1.2,0.035104000568389894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,16384,2,8,32,8,power_law_1.2,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.01,0.23023359775543212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.01,0.13508479595184325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,8,power_law_1.01,0.5716224193572998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,16384,2,8,32,8,power_law_1.2,0.03598720133304596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,16384,2,8,32,8,power_law_1.2,0.03705599904060364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.01,0.3425152063369751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.01,0.17242879867553712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,16384,2,8,32,8,power_law_1.2,0.03797119855880737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,8,power_law_1.01,0.6983295917510987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,16384,2,8,32,8,power_law_1.2,0.03706879913806915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.01,0.43935360908508303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,16384,2,8,32,8,power_law_1.2,0.03740800023078918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.01,0.2448960065841675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,8,power_law_1.01,1.1319871902465821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,8,power_law_1.01,0.5528639793395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,16384,2,8,32,8,power_law_1.2,0.039699199795722964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,16384,2,8,32,8,power_law_1.2,0.039136001467704774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.01,0.30531198978424073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,8,power_law_1.01,2.162175941467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,8,power_law_1.01,0.8720895767211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,16384,2,8,32,8,power_law_1.2,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,16384,2,8,32,8,power_law_1.2,0.044275200366973876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,8,power_law_1.01,0.37823359966278075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,8,power_law_1.01,1.703104019165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,16384,2,8,32,8,power_law_1.2,0.04564479887485504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,16384,2,8,32,8,power_law_1.2,0.04684799909591675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,8,power_law_1.01,0.5858943939208985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,16384,2,8,32,8,power_law_1.2,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,16384,2,8,32,8,power_law_1.2,0.04859519898891449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,8,power_law_1.01,1.1428031921386719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,16384,2,8,32,8,power_law_1.2,0.05914239883422852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,16384,2,8,32,8,power_law_1.2,0.05815039873123169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,16384,2,8,32,8,power_law_1.2,0.062457597255706786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,16384,2,8,32,8,power_law_1.2,0.06325119733810425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,16384,2,8,32,8,power_law_1.2,0.08072959780693054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,16384,2,8,32,8,power_law_1.2,0.08168960213661194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,16384,2,8,32,8,power_law_1.2,0.09300479888916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,16384,2,8,32,8,power_law_1.2,0.10751999616622925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,16384,2,8,32,8,power_law_1.2,0.12033280134201049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,16384,2,8,32,8,power_law_1.2,0.15016319751739501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,16384,2,8,32,8,power_law_1.2,0.15000959634780883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,16384,2,8,32,8,power_law_1.2,0.18240640163421631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,16384,2,8,32,8,power_law_1.2,0.2079551935195923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,16384,2,8,32,8,power_law_1.2,0.23848319053649902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,16384,2,8,32,8,power_law_1.2,0.2587199926376343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,16384,2,8,32,8,power_law_1.2,0.34920320510864256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,16384,2,8,32,8,power_law_1.2,0.317248010635376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,16384,2,8,32,8,power_law_1.2,0.4245567798614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,16384,2,8,32,8,power_law_1.2,0.4743040084838867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,16384,2,8,32,8,power_law_1.2,0.5522624015808105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,16384,2,8,32,8,power_law_1.2,0.6104191780090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,16384,2,8,32,8,power_law_1.2,0.7774975776672364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,16384,2,8,32,8,power_law_1.2,0.7452544212341309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,16384,2,8,32,8,power_law_1.2,1.5806207656860352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,16384,2,8,32,8,power_law_1.2,1.1041215896606444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,16384,2,8,32,8,power_law_1.2,2.226803207397461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,16384,2,8,32,8,power_law_1.2,0.03610239923000336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,16384,2,8,32,8,power_law_1.2,0.036294400691986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,16384,2,8,32,8,power_law_1.2,0.036364799737930296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,16384,2,8,32,8,power_law_1.2,0.038176000118255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,16384,2,8,32,8,power_law_1.2,0.018969599902629853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,16384,2,8,32,8,power_law_1.2,0.03825919926166534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,16384,2,8,32,8,power_law_1.2,0.018719999492168425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,16384,2,8,32,8,power_law_1.2,0.03864319920539856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,256,balanced,0.08059733112653096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,16384,2,8,32,8,power_law_1.2,0.01839359998703003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,16384,2,8,32,8,power_law_1.2,0.03860479891300202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,16384,2,8,32,8,power_law_1.2,0.018508799374103546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,256,balanced,0.08226666847864787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,16384,2,8,32,8,power_law_1.2,0.03920640051364899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,16384,2,8,32,8,power_law_1.2,0.018681600689888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,256,balanced,0.08265066643555959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,16384,2,8,32,8,power_law_1.2,0.03900800049304962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,16384,2,8,32,8,power_law_1.2,0.019174399971961974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,256,balanced,0.08226133386294048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,16384,2,8,32,8,power_law_1.2,0.038867199420928956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,16384,2,8,32,8,power_law_1.2,0.01926400065422058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,256,balanced,0.08429333567619324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,16384,2,8,32,8,power_law_1.2,0.0390720009803772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,16384,2,8,32,8,power_law_1.2,0.021695999801158904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,256,balanced,0.08653333783149719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,16384,2,8,32,8,power_law_1.2,0.040729600191116336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,16384,2,8,32,8,power_law_1.2,0.022540800273418427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,16384,2,8,32,8,power_law_1.2,0.022771200537681578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,256,balanced,0.08589866757392883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,16384,2,8,32,8,power_law_1.2,0.04192639887332916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,16384,2,8,32,8,power_law_1.2,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,256,balanced,0.08763200044631958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,16384,2,8,32,8,power_law_1.2,0.04209280014038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,16384,2,8,32,8,power_law_1.2,0.03470079898834229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,256,balanced,0.0906826655069987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,16384,2,8,32,8,power_law_1.2,0.04708479940891266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,16384,2,8,32,8,power_law_1.2,0.034643200039863584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,256,balanced,0.08884800473848979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,16384,2,8,32,8,power_law_1.2,0.04647040069103241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,16384,2,8,32,8,power_law_1.2,0.034790399670600894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,256,balanced,0.08835732936859131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,16384,2,8,32,8,power_law_1.2,0.05013120174407959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,16384,2,8,32,8,power_law_1.2,0.03552640080451965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,256,balanced,0.08967999617258708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,16384,2,8,32,8,power_law_1.2,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,16384,2,8,32,8,power_law_1.2,0.03619199991226196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,256,balanced,0.09173867106437683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.2,0.061273598670959474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,16384,2,8,32,8,power_law_1.2,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,256,balanced,0.09046399593353271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.2,0.07609599828720093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,16384,2,8,32,8,power_law_1.2,0.04200960099697113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,256,balanced,0.09893332918485005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.2,0.08988159894943237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.2,0.045926401019096376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,256,balanced,0.09562666217486064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.2,0.11148159503936768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.2,0.052883201837539674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,256,balanced,0.09715732932090759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.2,0.14067200422286988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.2,0.06154239773750305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.2,0.19479680061340332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.2,0.08590720295906067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,256,balanced,0.10147200028101604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.2,0.24662399291992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.2,0.09936000108718872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.2,0.34099199771881106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,256,balanced,0.10532800356547038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.2,0.45554561614990235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.2,0.13451520204544068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,16384,2,8,32,8,power_law_1.2,0.5288576126098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,256,balanced,0.11546666423479716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.2,0.17123839855194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,16384,2,8,32,8,power_law_1.2,0.8639552116394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.2,0.24324479103088378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,256,balanced,0.12369599938392639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,16384,2,8,32,8,power_law_1.2,1.674336051940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.2,0.305184006690979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,256,balanced,0.06053866446018219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,16384,2,8,32,8,power_law_1.2,0.377619194984436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,256,balanced,0.14484799901644388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,256,balanced,0.062309334675470986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,16384,2,8,32,8,power_law_1.2,0.5854144096374512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,256,balanced,0.060565332571665444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,16384,2,8,32,8,power_law_1.2,1.141862392425537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,256,balanced,0.06222933530807495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,256,balanced,0.16647467017173767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,256,balanced,0.06434666613737743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,256,balanced,0.0642133355140686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,256,balanced,0.06651199857393901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,256,balanced,0.06783466537793477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,256,balanced,0.20541866620381674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,256,balanced,0.06731200218200684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,256,balanced,0.06833066542943318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,256,balanced,0.0689333329598109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,256,balanced,0.0708426684141159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,256,balanced,0.07021866738796234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,256,balanced,0.2465440034866333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,256,balanced,0.07246933380762736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,256,balanced,0.07587733368078868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,256,balanced,0.07738133271535237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,256,balanced,0.08096533517042796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,256,balanced,0.08810133735338847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,256,balanced,0.023386667172114056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,256,balanced,0.060085331400235496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,256,balanced,0.09248000383377075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,256,balanced,0.06030400097370148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,256,balanced,0.0230880007147789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,256,balanced,0.32957865794499713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,256,balanced,0.059978668888409935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,256,balanced,0.11345066626866658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,256,balanced,0.025087999800841015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,256,balanced,0.062080000837643944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,256,balanced,0.06454400221506755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,256,balanced,0.14017066359519958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,256,power_law_1.01,0.07618560194969178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,256,balanced,0.029146666328112285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,256,balanced,0.06439466774463654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,256,power_law_1.01,0.06228479743003845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,256,balanced,0.05209066470464071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,256,balanced,0.06620799998442332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,256,balanced,0.06635199983914693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,256,balanced,0.042837331692377724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,256,power_law_1.01,0.06225919723510742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,256,power_law_1.01,0.07896320223808288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,256,balanced,0.17621866861979166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,256,balanced,0.06675733129183452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,256,balanced,0.043391997615496315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,256,power_law_1.01,0.06313599944114685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,256,power_law_1.01,0.07715839743614197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,256,balanced,0.406335989634196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,256,balanced,0.06605333089828491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,256,balanced,0.21234667301177979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,256,balanced,0.04241600135962168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,256,power_law_1.01,0.06636160016059875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,256,power_law_1.01,0.07930240035057068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,256,balanced,0.06868266562620799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,256,balanced,0.042549331982930504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,256,power_law_1.01,0.06446719765663148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,256,balanced,0.06833066542943318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,256,power_law_1.01,0.08017280101776122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,256,balanced,0.28010666370391846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,256,balanced,0.0697920024394989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,256,balanced,0.04377066592375437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,256,power_law_1.01,0.06621440052986145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,256,power_law_1.01,0.08339200019836426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,256,balanced,0.070783997575442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,256,balanced,0.5015466610590616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,256,balanced,0.04390933116277059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,256,balanced,0.07452799876530965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,256,power_law_1.01,0.08119680285453797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,256,power_law_1.01,0.06719359755516052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,256,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,256,balanced,0.07549866537253062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,256,balanced,0.3477120002110799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,256,power_law_1.01,0.08144639730453491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,256,power_law_1.01,0.06689919829368592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,256,balanced,0.04195733368396759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,256,balanced,0.07860266665617625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,256,power_law_1.01,0.08186879754066467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,256,power_law_1.01,0.06730239987373351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,256,balanced,0.04174399872620901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,256,balanced,0.08482666810353597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,256,power_law_1.01,0.0833407998085022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,256,balanced,0.48895466327667236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,256,power_law_1.01,0.06791679859161377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,256,balanced,0.7416959603627523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,256,balanced,0.050106664498647056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,256,balanced,0.08877333005269368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,256,power_law_1.01,0.08312320113182067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,256,power_law_1.01,0.06776319742202759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,256,balanced,0.04751466711362203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,256,balanced,0.10159466663996379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,256,power_law_1.01,0.08233600258827209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,256,power_law_1.01,0.06723200082778931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,256,balanced,0.04586666822433472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,256,balanced,0.11141332983970642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,256,power_law_1.01,0.08190720081329346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,256,power_law_1.01,0.06824319958686828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,256,balanced,0.6206560134887695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,256,balanced,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,256,power_law_1.01,0.07921280264854431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,256,power_law_1.01,0.06848639845848084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,256,balanced,0.14050666491190592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,256,power_law_1.01,0.08618239760398864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,256,power_law_1.01,0.07069439888000488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,256,balanced,0.05609600245952606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,256,power_law_1.01,0.08165119886398316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,256,power_law_1.01,0.07241600155830383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,256,balanced,0.0790719985961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,256,power_law_1.01,0.08684160113334656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,256,balanced,0.16752000649770102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,256,power_law_1.01,0.0749888002872467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,256,power_law_1.01,0.10654079914093018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,256,balanced,0.08860799670219421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,256,power_law_1.01,0.09162880182266235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,256,power_law_1.01,0.10380159616470337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,256,balanced,0.21794666846593222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,256,power_law_1.01,0.10038399696350098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,256,power_law_1.01,0.12538880109786987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,256,balanced,0.7646293640136719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,256,balanced,0.12165866295496623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,256,power_law_1.01,0.12755199670791625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,256,power_law_1.01,0.14135680198669434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,256,power_law_1.01,0.14972800016403198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,256,balanced,0.1548853317896525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,256,balanced,0.2657173275947571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,256,power_law_1.01,0.18599679470062255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,256,power_law_1.01,0.20157439708709718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,256,balanced,0.21780800819396973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,256,power_law_1.01,0.22352640628814696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,256,balanced,1.4462025960286458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,256,power_law_1.01,0.2505919933319092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,256,power_law_1.01,0.3115135908126831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,256,power_law_1.01,0.3336064100265503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,256,power_law_1.01,0.38735361099243165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,256,power_law_1.01,0.42196478843688967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,256,balanced,0.36742933591206867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,256,power_law_1.01,0.5456511974334717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,256,power_law_1.01,0.6139328002929687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,256,balanced,0.2825973431269328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,256,power_law_1.01,0.7330495834350585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,256,balanced,1.173690636952718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,256,power_law_1.01,0.8143679618835449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,256,power_law_1.01,0.8682880401611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,256,power_law_1.01,0.9858943939208984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,256,power_law_1.01,1.434438419342041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,256,power_law_1.01,1.5864128112792968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,256,balanced,0.4090133508046468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,256,balanced,0.46169598897298175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,256,power_law_1.01,2.876448059082031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,256,power_law_1.01,3.191628837585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,256,balanced,0.5362826585769653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,256,balanced,0.5622719923655192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,256,balanced,0.6609546740849813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,256,balanced,0.8580693403879801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,256,balanced,2.3063947359720864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,256,power_law_1.01,0.060787200927734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,256,power_law_1.01,0.021516799926757812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,256,power_law_1.01,0.0616703987121582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,256,power_law_1.01,0.02082560062408447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,256,power_law_1.01,0.05859839916229248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,256,power_law_1.01,0.022617599368095397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,256,balanced,1.0419466495513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,256,power_law_1.01,0.06167680025100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,256,power_law_1.01,0.025676798820495606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,256,power_law_1.01,0.06314880251884461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,256,power_law_1.01,0.052825599908828735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,256,power_law_1.01,0.06252800226211548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,256,power_law_1.01,0.041465601325035094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,256,power_law_1.01,0.06328319907188415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,256,power_law_1.01,0.06491519808769226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,256,power_law_1.01,0.04145280122756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,256,power_law_1.01,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,256,power_law_1.01,0.041606399416923526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,256,power_law_1.01,0.06438400149345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,256,power_law_1.01,0.04407039880752563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,256,power_law_1.01,0.06553599834442139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,256,power_law_1.01,0.04311679899692535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,256,power_law_1.01,0.06631680130958557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,256,power_law_1.01,0.04222719967365265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,256,power_law_1.01,0.0656000018119812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,256,power_law_1.01,0.042393600940704344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,256,power_law_1.01,0.0679744005203247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,256,power_law_1.01,0.041305598616600034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,256,power_law_1.01,0.07043200135231018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,256,power_law_1.01,0.03914879858493805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,256,power_law_1.01,0.041555199027061465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,256,power_law_1.01,0.0674560010433197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,256,power_law_1.01,0.04499199986457825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,256,power_law_1.01,0.0702015995979309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,256,balanced,1.6768852869669597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,256,power_law_1.01,0.049497601389884946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,256,power_law_1.01,0.0869055986404419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,256,power_law_1.01,0.05110399723052979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.01,0.09054080247879029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.01,0.054636800289154054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.01,0.11185920238494873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.01,0.12962559461593628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.01,0.0642304003238678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.01,0.1693567991256714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.01,0.08412160277366638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.01,0.20015358924865723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.01,0.27155840396881104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.01,0.11326080560684204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.01,0.33902080059051515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,256,balanced,2.0500319798787436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.01,0.1267840027809143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.01,0.48755202293395994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.01,0.18979840278625487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.01,0.6190783977508545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.01,0.2281791925430298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.01,0.3289792060852051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,256,power_law_1.01,0.7777344226837158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.01,0.43543038368225095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,256,power_law_1.01,1.2033920288085938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,256,power_law_1.01,0.5396224021911621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,256,power_law_1.01,0.8623744010925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,256,power_law_1.01,2.4144384384155275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,256,power_law_1.01,1.695756721496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,1,256,power_law_1.2,0.08139520287513732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,1,256,power_law_1.2,0.08412799835205079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,1,256,power_law_1.2,0.08143360018730164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,1,256,power_law_1.2,0.08505600094795226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,1,256,power_law_1.2,0.08373759984970093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,1,256,power_law_1.2,0.08533759713172913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,1,256,power_law_1.2,0.08532480001449586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,1,256,power_law_1.2,0.08526080250740051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,1,256,power_law_1.2,0.08502399921417236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,1,256,power_law_1.2,0.08516479730606079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,1,256,power_law_1.2,0.08339840173721313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,1,256,power_law_1.2,0.08389760255813598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,1,256,power_law_1.2,0.08054400086402894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,1,256,power_law_1.2,0.05994240045547485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,1,256,power_law_1.2,0.0791808009147644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,1,256,power_law_1.2,0.0837440013885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,1,256,power_law_1.2,0.06037120223045349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,1,256,power_law_1.2,0.08416640162467956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,1,256,power_law_1.2,0.06259840130805969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,1,256,power_law_1.2,0.06108800172805786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,1,256,power_law_1.2,0.09849600195884704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,1,256,power_law_1.2,0.06277120113372803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,1,256,power_law_1.2,0.06042240262031555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,1,256,power_law_1.2,0.063673597574234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,1,256,power_law_1.2,0.10167039632797241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,1,256,power_law_1.2,0.06490880250930786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,1,256,power_law_1.2,0.06238719820976257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,1,256,power_law_1.2,0.11395200490951538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,1,256,power_law_1.2,0.06451839804649354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,1,256,power_law_1.2,0.06414719820022582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,1,256,power_law_1.2,0.1406272053718567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,1,256,power_law_1.2,0.06590080261230469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,1,256,power_law_1.2,0.0652671992778778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,1,256,power_law_1.2,0.17541120052337647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,1,256,power_law_1.2,0.06416000127792358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,1,256,power_law_1.2,0.06626560091972351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,1,256,power_law_1.2,0.2436288118362427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,1,256,power_law_1.2,0.06419839859008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,1,256,power_law_1.2,0.021465599536895752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,1,256,power_law_1.2,0.06599680185317994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,1,256,power_law_1.2,0.0654528021812439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,1,256,power_law_1.2,0.2733439922332764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,1,256,power_law_1.2,0.021190400421619415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,1,256,power_law_1.2,0.06547840237617493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,1,256,power_law_1.2,0.06640639901161194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,1,256,power_law_1.2,0.02207999974489212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,1,256,power_law_1.2,0.06540799736976624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,1,256,power_law_1.2,0.432857608795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,1,256,power_law_1.2,0.06653439998626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,1,256,power_law_1.2,0.026035198569297792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,1,256,power_law_1.2,0.0658623993396759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,1,256,power_law_1.2,0.5419968128204345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,1,256,power_law_1.2,0.06696320176124573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,1,256,power_law_1.2,0.052102398872375486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,1,256,power_law_1.2,0.066457599401474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,1,256,power_law_1.2,0.8415231704711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,1,256,power_law_1.2,0.06630399823188782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,1,256,power_law_1.2,0.039955198764801025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,1,256,power_law_1.2,0.06600959897041321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,1,256,power_law_1.2,1.1003071784973144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,1,256,power_law_1.2,0.06565759778022766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,1,256,power_law_1.2,0.040761598944664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,1,256,power_law_1.2,0.06892160177230836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,1,256,power_law_1.2,1.45797758102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,1,256,power_law_1.2,0.06930559873580933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,1,256,power_law_1.2,0.04030719995498657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,1,256,power_law_1.2,0.07020800113677979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,1,256,power_law_1.2,2.493657684326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,1,256,power_law_1.2,0.0679423987865448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,1,256,power_law_1.2,0.04134399890899658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,1,256,power_law_1.2,0.08331519961357117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,1,256,power_law_1.2,5.085715103149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,1,256,power_law_1.2,0.07884160280227662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,1,256,power_law_1.2,0.04120959937572479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,1,256,power_law_1.2,0.08668799996376038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,1,256,power_law_1.2,0.0839743971824646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,1,256,power_law_1.2,0.04108160138130188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,1,256,power_law_1.2,0.11048320531845093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,1,256,power_law_1.2,0.0424703985452652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.2,0.10113919973373413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,1,256,power_law_1.2,0.13814400434494017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,1,256,power_law_1.2,0.041254401206970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,1,256,power_law_1.2,0.1663424015045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.2,0.1189568042755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,1,256,power_law_1.2,0.04077439904212952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,1,256,power_law_1.2,0.2212735891342163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.2,0.13877760171890258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,1,256,power_law_1.2,0.041119998693466185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,1,256,power_law_1.2,0.27674241065979005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.2,0.19041279554367066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,1,256,power_law_1.2,0.04635519981384277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.2,0.23441920280456544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,1,256,power_law_1.2,0.38662400245666506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,1,256,power_law_1.2,0.05002880096435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.2,0.3188479900360107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,1,256,power_law_1.2,0.5151040077209472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,1,256,power_law_1.2,0.05377280116081238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.2,0.4424191951751709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,1,256,power_law_1.2,0.7475903987884521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.2,0.056467199325561525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.2,0.6516223907470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,1,256,power_law_1.2,1.06179838180542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.2,0.07244160175323486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.2,0.8423999786376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,1,256,power_law_1.2,1.2911487579345704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.2,0.0862272024154663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,1,256,power_law_1.2,1.0810175895690919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.2,0.11463680267333984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,1,256,power_law_1.2,2.2935232162475585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.2,0.13685760498046876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,1,256,power_law_1.2,1.8726207733154296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,1,256,power_law_1.2,4.471577453613281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.2,0.2002432107925415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,1,256,power_law_1.2,3.6255615234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.2,0.25905919075012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,128,balanced,0.06433066725730896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.2,0.3638144016265869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,128,balanced,0.06454400221506755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.2,0.49587202072143555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,1,256,power_law_1.2,0.6072192192077637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,128,balanced,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,1,256,power_law_1.2,1.0000063896179199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,128,balanced,0.06639466683069865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,1,256,power_law_1.2,2.0028032302856444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,128,balanced,0.06637866795063019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,128,balanced,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,128,balanced,0.07316266496976216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,128,balanced,0.0749013324578603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,128,balanced,0.08063999811808269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,128,balanced,0.07565866907437642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,128,balanced,0.07876266539096832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,128,balanced,0.0781173308690389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,128,balanced,0.0775733341773351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,128,balanced,0.07866666714350383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,128,balanced,0.08462933699289958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,128,balanced,0.08524800340334575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,128,balanced,0.08675199747085571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,128,balanced,0.09731733798980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,128,balanced,0.09894399841626485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,128,balanced,0.05438933273156484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,128,balanced,0.11372799674669902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,128,balanced,0.05411200225353241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,128,balanced,0.054133335749308266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,128,balanced,0.12237333257993062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,128,balanced,0.05595199763774872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,128,balanced,0.05961066484451294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,128,balanced,0.14404799540837607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,128,balanced,0.06482133269309998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,128,balanced,0.06228800117969513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,128,balanced,0.16496533155441284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,128,balanced,0.06442666550477345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,128,balanced,0.06445866823196411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,128,balanced,0.0646613339583079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,128,balanced,0.20769067605336508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,128,balanced,0.06618133187294006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,128,balanced,0.0664106657107671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,128,balanced,0.06850666801134746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,128,balanced,0.2485919992129008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,128,balanced,0.0684746652841568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,128,balanced,0.07261866827805837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,128,balanced,0.07600533465544383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,128,balanced,0.07667733232180278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,128,balanced,0.0845973292986552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,128,balanced,0.3311093250910441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,128,balanced,0.09037866195042928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,128,balanced,0.11525332927703857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,128,balanced,0.139765332142512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,128,balanced,0.175546665986379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,128,balanced,0.4089599847793579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,128,balanced,0.054101333022117615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,128,balanced,0.2114293376604716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,128,balanced,0.05564799904823303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,128,balanced,0.05504000186920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,128,balanced,0.056549335519472756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,128,balanced,0.281061331431071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,128,balanced,0.05795733133951823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,128,balanced,0.0643039991458257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,128,balanced,0.06608533362547557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,128,balanced,0.06443200012048085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,128,balanced,0.35093867778778076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,128,balanced,0.06711466610431671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,128,balanced,0.5087519884109497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,128,balanced,0.021274665991465252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,128,balanced,0.0640533318122228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,128,balanced,0.06650133430957794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,128,balanced,0.02187199890613556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,128,balanced,0.06650666892528534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,128,balanced,0.023205332458019257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,128,balanced,0.06615466872851054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,128,balanced,0.023237332701683044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,128,balanced,0.06835733354091644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,128,balanced,0.027232001225153606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,128,balanced,0.0724373310804367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,128,balanced,0.49058131376902264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,128,balanced,0.05264000097910563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,128,balanced,0.042378668983777366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,128,balanced,0.07250666618347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,128,balanced,0.042026668787002563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,128,balanced,0.07678933441638947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,128,balanced,0.04193066557248434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,128,balanced,0.08703466256459554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,128,balanced,0.042122667034467064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,128,balanced,0.042319998145103455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,128,balanced,0.08684800068537395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,128,balanced,0.042810668547948204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,128,balanced,0.04160533348719279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,128,balanced,0.09924266735712688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,128,balanced,0.04171733558177948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,128,balanced,0.6253226598103842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,128,balanced,0.11513066291809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,128,balanced,0.05219733218352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,128,balanced,0.7471413612365723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,128,balanced,0.04942933221658071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,128,balanced,0.1421440045038859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,128,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,128,balanced,0.05160533388455709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,128,balanced,0.16873600085576376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,128,balanced,0.05798399945100149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,128,balanced,0.08384000261624654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,128,balanced,0.21860265731811523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,128,balanced,0.7710826396942139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,128,balanced,0.09305066863695781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,128,balanced,0.12808000048001608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,128,balanced,0.26690133412679035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,128,balanced,0.16247466206550598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,128,balanced,0.22871466477711996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,128,balanced,0.3694719870885213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,128,balanced,0.2887786626815796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,128,balanced,1.1836106777191162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,128,balanced,0.46650667985280353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,128,balanced,0.4261600176493327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,128,balanced,1.4594720204671223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,128,balanced,0.5602399905522665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,128,balanced,0.5677706797917684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,128,balanced,0.6938347021738688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,128,power_law_1.01,0.07051519751548767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,128,power_law_1.01,0.07356799840927124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,128,power_law_1.01,0.06934400200843811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,128,balanced,2.3262826601664224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,128,power_law_1.01,0.0716543972492218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,128,balanced,0.8679359753926595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,128,power_law_1.01,0.07319679856300354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,128,power_law_1.01,0.07633280158042907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,128,power_law_1.01,0.07491840124130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,128,power_law_1.01,0.07713919878005981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,128,power_law_1.01,0.07648000121116638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,128,power_law_1.01,0.07617279887199402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,128,power_law_1.01,0.07617279887199402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,128,balanced,1.0911093552907307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,128,power_law_1.01,0.07774720191955567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,128,power_law_1.01,0.08013439774513245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,128,power_law_1.01,0.07917439937591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,128,power_law_1.01,0.0837440013885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,128,power_law_1.01,0.0850048005580902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,128,power_law_1.01,0.08855040073394775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,128,power_law_1.01,0.101254403591156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,128,power_law_1.01,0.10623359680175781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,128,power_law_1.01,0.11854079961776734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,128,power_law_1.01,0.13292800188064574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,128,power_law_1.01,0.16942720413208007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,128,power_law_1.01,0.20417280197143556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,128,power_law_1.01,0.26606719493865966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,128,power_law_1.01,0.3454400062561035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,128,power_law_1.01,0.4956480026245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,128,power_law_1.01,0.6975488185882568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,128,power_law_1.01,0.05963519811630249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,128,power_law_1.01,0.8018048286437989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,128,power_law_1.01,0.06490240097045899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,128,power_law_1.01,0.05865600109100342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,128,balanced,1.7036693890889485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,128,power_law_1.01,1.2338751792907714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,128,power_law_1.01,0.06416640281677247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,128,power_law_1.01,0.05729920268058777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,128,power_law_1.01,2.556723213195801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,128,power_law_1.01,0.06174719929695129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,128,power_law_1.01,0.05978879928588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,128,power_law_1.01,0.06095359921455383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,128,power_law_1.01,0.06034560203552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,128,balanced,2.150767962137858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,128,power_law_1.01,0.062483197450637816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,128,power_law_1.01,0.06344959735870362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,128,power_law_1.01,0.021030400693416596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,128,power_law_1.01,0.06558079719543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,128,power_law_1.01,0.06289280056953431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,128,power_law_1.01,0.02139520049095154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,128,power_law_1.01,0.06664320230484008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,128,power_law_1.01,0.06407039761543273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,128,power_law_1.01,0.02173440009355545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,128,power_law_1.01,0.06273279786109924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,128,power_law_1.01,0.06321920156478882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,128,power_law_1.01,0.021657599508762358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,128,power_law_1.01,0.06257280111312866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,128,power_law_1.01,0.06421120166778564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,128,power_law_1.01,0.026425600051879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,128,power_law_1.01,0.06494719982147217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,128,power_law_1.01,0.06364799737930298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,128,power_law_1.01,0.051692801713943484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,128,power_law_1.01,0.06449919939041138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,128,power_law_1.01,0.0646399974822998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,128,power_law_1.01,0.04089600145816803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,128,power_law_1.01,0.06691200137138367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,128,power_law_1.01,0.06528639793395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,128,power_law_1.01,0.04038400053977966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,128,power_law_1.01,0.06534399986267089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,128,power_law_1.01,0.06721280217170715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,128,power_law_1.01,0.04079360067844391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,128,power_law_1.01,0.06632959842681885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,128,power_law_1.01,0.0703935980796814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,128,power_law_1.01,0.04090240001678467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,128,power_law_1.01,0.07119359970092773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,128,power_law_1.01,0.07233920097351074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,128,power_law_1.01,0.04159359931945801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,128,power_law_1.01,0.07207679748535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,128,power_law_1.01,0.07671679854393006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,128,power_law_1.01,0.041331198811531064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,128,power_law_1.01,0.07441920042037964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,128,power_law_1.01,0.08777599930763244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,128,power_law_1.01,0.04038400053977966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,128,power_law_1.01,0.08503680229187012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,128,power_law_1.01,0.10140160322189332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,128,power_law_1.01,0.0394239991903305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.01,0.09042559862136841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,128,power_law_1.01,0.12614400386810304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,128,power_law_1.01,0.04074879884719849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.01,0.11302399635314941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,128,power_law_1.01,0.1491328001022339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,128,power_law_1.01,0.042182400822639465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.01,0.12908799648284913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,128,power_law_1.01,0.1965824007987976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,128,power_law_1.01,0.049779200553894044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.01,0.16155519485473632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,128,power_law_1.01,0.2361664056777954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,128,power_law_1.01,0.0484607994556427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.01,0.1994815945625305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,128,power_law_1.01,0.33092479705810546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.01,0.054611200094223024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.01,0.2597759962081909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,128,power_law_1.01,0.41264638900756834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.01,0.06272640228271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.01,0.32629759311676027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,128,power_law_1.01,0.595033597946167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.01,0.08421120047569275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.01,0.4782847881317139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,128,power_law_1.01,0.7968895912170411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.01,0.10684800148010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.01,0.613862419128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.01,0.13191039562225343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,128,power_law_1.01,0.9513407707214355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.01,0.1830016016960144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,128,power_law_1.01,0.7312960147857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,128,power_law_1.01,1.4633472442626954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.01,0.22619519233703614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,128,power_law_1.01,1.1502400398254395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.01,0.3495232105255127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,128,power_law_1.01,3.0731199264526365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,128,power_law_1.01,2.347078323364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.01,0.4243008136749268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,128,power_law_1.01,0.5593664169311523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,128,power_law_1.01,0.8881407737731933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,128,power_law_1.01,1.732102394104004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,2,128,power_law_1.2,0.0679040014743805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,2,128,power_law_1.2,0.06926079988479614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,2,128,power_law_1.2,0.06729599833488464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,2,128,power_law_1.2,0.071424001455307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,2,128,power_law_1.2,0.07214080095291138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,2,128,power_law_1.2,0.05921919941902161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,2,128,power_law_1.2,0.07238399982452393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,2,128,power_law_1.2,0.058220797777175905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,2,128,power_law_1.2,0.07313920259475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,2,128,power_law_1.2,0.05470719933509827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,2,128,power_law_1.2,0.05989760160446167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,2,128,power_law_1.2,0.07587199807167053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,2,128,power_law_1.2,0.05861120223999024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,2,128,power_law_1.2,0.07490559816360473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,2,128,power_law_1.2,0.061612802743911746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,2,128,power_law_1.2,0.07664639949798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,2,128,power_law_1.2,0.06327679753303528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,2,128,power_law_1.2,0.0773311972618103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,2,128,power_law_1.2,0.06261119842529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,2,128,power_law_1.2,0.07708160281181335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,2,128,power_law_1.2,0.06368640065193176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,2,128,power_law_1.2,0.07846400141716003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,2,128,power_law_1.2,0.062463998794555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,2,128,power_law_1.2,0.07856000065803528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,2,128,power_law_1.2,0.06368640065193176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,2,128,power_law_1.2,0.08439679741859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,2,128,power_law_1.2,0.06515840291976929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,2,128,power_law_1.2,0.08406400084495544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,2,128,power_law_1.2,0.06477439999580384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,2,128,power_law_1.2,0.09377279877662659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,2,128,power_law_1.2,0.06686080098152161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,2,128,power_law_1.2,0.07021440267562866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,2,128,power_law_1.2,0.1017024040222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,2,128,power_law_1.2,0.07134079933166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,2,128,power_law_1.2,0.1092352032661438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,2,128,power_law_1.2,0.07878400087356567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,2,128,power_law_1.2,0.12906880378723146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,2,128,power_law_1.2,0.0899392008781433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,2,128,power_law_1.2,0.14669439792633057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,2,128,power_law_1.2,0.10462720394134521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,2,128,power_law_1.2,0.19718400239944459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,2,128,power_law_1.2,0.13075200319290162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,2,128,power_law_1.2,0.23834879398345948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,2,128,power_law_1.2,0.15534720420837403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,2,128,power_law_1.2,0.32718720436096194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,2,128,power_law_1.2,0.2093503952026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,2,128,power_law_1.2,0.4323008060455322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,2,128,power_law_1.2,0.2514303922653198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,2,128,power_law_1.2,0.6031936168670654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,2,128,power_law_1.2,0.3589184045791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,2,128,power_law_1.2,0.8353471755981445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,2,128,power_law_1.2,0.4668416023254395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,2,128,power_law_1.2,0.07180160284042358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,2,128,power_law_1.2,1.1044672012329102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,2,128,power_law_1.2,0.6902783870697021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,2,128,power_law_1.2,1.8359167098999023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,2,128,power_law_1.2,0.9460479736328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,2,128,power_law_1.2,0.06228479743003845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,2,128,power_law_1.2,1.1862719535827637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,2,128,power_law_1.2,0.057715201377868654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,2,128,power_law_1.2,3.778879928588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,2,128,power_law_1.2,0.020838400721549986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,2,128,power_law_1.2,0.06098560094833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,2,128,power_law_1.2,1.845088005065918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,2,128,power_law_1.2,0.02142080068588257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,2,128,power_law_1.2,0.06124160289764404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,2,128,power_law_1.2,0.020640000700950623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,2,128,power_law_1.2,4.118598556518554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,2,128,power_law_1.2,0.06125440001487732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,2,128,power_law_1.2,0.021881599724292756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,2,128,power_law_1.2,0.06291199922561645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,2,128,power_law_1.2,0.026873600482940675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,2,128,power_law_1.2,0.06344959735870362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,2,128,power_law_1.2,0.05125759840011597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,2,128,power_law_1.2,0.041529598832130435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,2,128,power_law_1.2,0.06149119734764099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,2,128,power_law_1.2,0.04167680144309997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,2,128,power_law_1.2,0.06183680295944214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,2,128,power_law_1.2,0.042259201407432556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,2,128,power_law_1.2,0.06406400203704835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,2,128,power_law_1.2,0.041657599806785586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,2,128,power_law_1.2,0.06420480012893677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,2,128,power_law_1.2,0.042099198698997496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,2,128,power_law_1.2,0.06405760049819946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,2,128,power_law_1.2,0.042342400550842284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,2,128,power_law_1.2,0.06614400148391723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,2,128,power_law_1.2,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,2,128,power_law_1.2,0.03920640051364899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,2,128,power_law_1.2,0.07026559710502625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,2,128,power_law_1.2,0.04147840142250061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,2,128,power_law_1.2,0.07204480171203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,2,128,power_law_1.2,0.04344319999217987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,2,128,power_law_1.2,0.07638400197029113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,2,128,power_law_1.2,0.04853120148181915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,2,128,power_law_1.2,0.04842880070209503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,2,128,power_law_1.2,0.08440319895744323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.2,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.2,0.09238399863243103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.2,0.06508799791336059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.2,0.11758079528808593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.2,0.08416640162467956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.2,0.13781119585037233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.2,0.10649600028991699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.2,0.1779584050178528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.2,0.13772159814834595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.2,0.21834239959716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.2,0.19451520442962647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.2,0.29424641132354734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.2,0.237772798538208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.2,0.3867392063140869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.2,0.36538240909576414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.2,0.5393407821655274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.2,0.4770559787750244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.2,0.7219776153564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,2,128,power_law_1.2,0.6012095928192138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,2,128,power_law_1.2,0.8713791847229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,2,128,power_law_1.2,0.9661696434020997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,2,128,power_law_1.2,1.471008014678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,2,128,power_law_1.2,1.8899200439453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,2,128,power_law_1.2,3.489574432373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,64,balanced,0.023247999449570973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,64,balanced,0.025173333783944447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,64,balanced,0.021216000119845074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,64,balanced,0.02272533377011617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,64,balanced,0.02248000105222066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,64,balanced,0.023557332654794056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,64,balanced,0.025568000972270966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,64,balanced,0.02613866577545802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,64,balanced,0.03293866664171219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,64,balanced,0.031541332602500916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,64,balanced,0.06020266811052958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,64,balanced,0.04201599955558777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,64,balanced,0.052111998200416565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,64,balanced,0.04306666553020477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,64,balanced,0.049882665276527405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,64,balanced,0.04219200213750204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,64,balanced,0.05179200073083242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,64,balanced,0.041936000188191734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,64,balanced,0.05209066470464071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,64,balanced,0.0602453351020813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,64,balanced,0.05279466509819031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,64,balanced,0.0569706658522288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,64,balanced,0.05576533575852712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,64,balanced,0.05165866514046987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,64,balanced,0.055733333031336464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,64,balanced,0.05709333221117655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,64,balanced,0.05604266623655955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,64,balanced,0.05610666672388712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,64,balanced,0.0661653329928716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,64,balanced,0.05806399881839752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,64,balanced,0.09212266405423482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,64,balanced,0.058037335673967995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,64,balanced,0.059232001503308616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,64,balanced,0.10367467006047566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,64,balanced,0.060175999999046326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,64,balanced,0.06432533264160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,64,balanced,0.1441920002301534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,64,balanced,0.06427733103434245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,64,balanced,0.06799999872843425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,64,balanced,0.1805973251660665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,64,balanced,0.07449600100517273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,64,balanced,0.07715733349323273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,64,balanced,0.07431999842325847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,64,balanced,0.0888213316599528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,64,balanced,0.25861867268880206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,64,balanced,0.045791998505592346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,64,balanced,0.10230400164922078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,64,balanced,0.04460800190766653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,64,balanced,0.045968001087506614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,64,balanced,0.04586133360862732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,64,balanced,0.12787200013796488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,64,balanced,0.05046399931112925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,64,balanced,0.3264906605084737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,64,balanced,0.05217599868774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,64,balanced,0.15230400363604227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,64,balanced,0.052095999320348106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,64,balanced,0.052069331208864846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,64,balanced,0.05380799869696299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,64,balanced,0.053743998209635414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,64,balanced,0.1939786672592163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,64,balanced,0.48441068331400555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,64,balanced,0.053930665055910744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,64,balanced,0.055770665407180786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,64,balanced,0.058133333921432495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,64,balanced,0.06437333424886067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,64,balanced,0.23678932587305704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,64,balanced,0.06342933575312297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,64,balanced,0.06842666864395142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,64,balanced,0.6402986844380697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,64,balanced,0.0765066643555959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,64,balanced,0.07929599781831105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,64,balanced,0.09102933605511983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,64,balanced,0.3213333288828532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,64,balanced,0.09919466574986775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,64,balanced,0.11656000216801961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,64,balanced,0.13778666655222574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,64,balanced,0.8056480089823405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,64,balanced,0.4039466778437297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,64,balanced,0.06215466558933258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,64,balanced,0.17086400588353476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,64,balanced,0.04612799982229868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,64,balanced,0.04563733438650767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,64,balanced,0.045610666275024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,64,balanced,0.04771199822425842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,64,balanced,0.04795200129350027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,64,balanced,0.050069332122802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,64,balanced,0.20526933670043945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,64,balanced,0.05106133222579956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,64,balanced,0.0517546683549881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,64,balanced,0.05194666484991709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,64,balanced,0.49396268526713055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,64,balanced,0.05375466744105021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,64,balanced,0.05351466437180837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,64,balanced,0.056074668963750206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,64,balanced,0.2855199972788493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,64,balanced,0.058058664202690125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,64,balanced,0.06019733349482218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,64,balanced,0.062234664956728615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,64,balanced,1.2625280221303303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,64,balanced,0.06628799935181935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,64,balanced,0.07216533521811168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,64,balanced,0.33952534198760986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,64,balanced,0.07860266665617625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,64,balanced,0.10149332880973816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,64,balanced,0.7447466850280762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,64,balanced,0.12388799587885539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,64,balanced,0.1586079994837443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,64,balanced,0.18900267283121744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,64,power_law_1.01,0.06470400094985962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,64,power_law_1.01,0.12158080339431762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,64,balanced,0.42095998922983807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,64,power_law_1.01,0.04615040123462677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,64,power_law_1.01,0.04731520116329193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,64,balanced,0.25276799996693927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,64,power_law_1.01,0.04418559968471527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,64,power_law_1.01,0.04635519981384277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,64,power_law_1.01,0.04636160135269165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,64,power_law_1.01,0.05242879986763001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,64,power_law_1.01,0.047014400362968445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,64,power_law_1.01,0.05229439735412598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,64,balanced,0.32443199555079144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,64,power_law_1.01,0.048511999845504764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,64,power_law_1.01,0.0526528000831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,64,power_law_1.01,0.04954879879951477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,64,power_law_1.01,0.05358080267906189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,64,power_law_1.01,0.05095040202140808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,64,power_law_1.01,0.05186560153961182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,64,power_law_1.01,0.051052802801132204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,64,power_law_1.01,0.05443199872970581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,64,balanced,2.4975412686665854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,64,power_law_1.01,0.05127040147781372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,64,balanced,0.4489440123240153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,64,power_law_1.01,0.05648639798164368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,64,balanced,0.6155360142389933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,64,power_law_1.01,0.05241600275039673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,64,balanced,1.4509600003560383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,64,power_law_1.01,0.06402559876441956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,64,power_law_1.01,0.02319999933242798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,64,power_law_1.01,0.05702400207519531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,64,power_law_1.01,0.05351679921150208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,64,power_law_1.01,0.06225919723510742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,64,power_law_1.01,0.024172799289226533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,64,power_law_1.01,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,64,power_law_1.01,0.056595200300216676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,64,balanced,0.5642240047454834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,64,power_law_1.01,0.05434240102767944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,64,power_law_1.01,0.02033279985189438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,64,power_law_1.01,0.05583360195159912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,64,power_law_1.01,0.05707520246505737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,64,power_law_1.01,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,64,power_law_1.01,0.020454399287700653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,64,power_law_1.01,0.05986559987068176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,64,power_law_1.01,0.06181120276451111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,64,power_law_1.01,0.06275200247764587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,64,power_law_1.01,0.0621504008769989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,64,power_law_1.01,0.02136960029602051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,64,power_law_1.01,0.06806399822235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,64,power_law_1.01,0.061817598342895505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,64,power_law_1.01,0.0658623993396759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,64,power_law_1.01,0.021612800657749176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,64,power_law_1.01,0.0642687976360321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,64,power_law_1.01,0.06969599723815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,64,power_law_1.01,0.06365439891815186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,64,power_law_1.01,0.07254400253295898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,64,power_law_1.01,0.023097600042819976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,64,power_law_1.01,0.06527360081672669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,64,power_law_1.01,0.07502080202102661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,64,power_law_1.01,0.08393599987030029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,64,power_law_1.01,0.0652351975440979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,64,balanced,0.7003520329793295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,64,power_law_1.01,0.02707839906215668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,64,power_law_1.01,0.08184319734573364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,64,power_law_1.01,0.06586880087852479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,64,power_law_1.01,0.09474560022354125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,64,power_law_1.01,0.02590720057487488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,64,power_law_1.01,0.08878719806671143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,64,power_law_1.01,0.06615039706230164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,64,power_law_1.01,0.11847039461135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,64,power_law_1.01,0.026188799738883974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,64,power_law_1.01,0.10664960145950317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,64,power_law_1.01,0.06863999962806702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,64,power_law_1.01,0.04428800046443939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,64,power_law_1.01,0.07276160120964051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,64,power_law_1.01,0.14073599576950074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,64,power_law_1.01,0.11938560009002686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,64,balanced,1.1781493028004963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,64,power_law_1.01,0.043270400166511534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,64,power_law_1.01,0.07281919717788696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,64,power_law_1.01,0.18377599716186524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,64,power_law_1.01,0.04318720102310181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,64,balanced,1.0623520215352376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,64,power_law_1.01,0.1480064034461975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,64,power_law_1.01,0.0760320007801056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,64,power_law_1.01,0.22167038917541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,64,power_law_1.01,0.042752000689506534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,64,power_law_1.01,0.08485119938850402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,64,power_law_1.01,0.18253439664840698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,64,power_law_1.01,0.30410239696502683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,64,power_law_1.01,0.0451200008392334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.01,0.09151359796524047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,64,power_law_1.01,0.2501120090484619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,64,power_law_1.01,0.38816640377044676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,64,power_law_1.01,0.04752640128135681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.01,0.11327999830245972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,64,power_law_1.01,0.32403199672698973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,64,power_law_1.01,0.5578303813934327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,64,power_law_1.01,0.04894079864025116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.01,0.13156479597091675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,64,power_law_1.01,0.4390528202056885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,64,power_law_1.01,0.7225535869598388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,64,power_law_1.01,0.05060480237007141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.01,0.16128000020980834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,64,power_law_1.01,0.5127359867095947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.01,0.05553920269012451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.01,0.20093441009521484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,64,power_law_1.01,0.8708864212036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.01,0.06639360189437866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,64,power_law_1.01,0.6064127922058106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.01,0.2659264087677002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,64,power_law_1.01,1.397107219696045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.01,0.07582079768180847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.01,0.34501760005950927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,64,power_law_1.01,0.9241408348083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.01,0.10422400236129761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.01,0.49649920463562014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,64,power_law_1.01,2.7021631240844726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,64,power_law_1.01,1.7847103118896483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.01,0.12814719676971437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.01,0.5959167957305909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,64,balanced,2.0779147148132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.01,0.17509759664535524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,64,power_law_1.01,0.7632575988769531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.01,0.21731200218200683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,64,power_law_1.01,1.1464256286621093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.01,0.3077568054199219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,64,power_law_1.01,2.3267391204833983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.01,0.4160192012786865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,64,power_law_1.01,0.4999551773071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,64,power_law_1.01,0.7459968090057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,64,power_law_1.01,1.4766655921936036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,4,64,power_law_1.2,0.06488320231437683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,4,64,power_law_1.2,0.12183040380477905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,4,64,power_law_1.2,0.04652799963951111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,4,64,power_law_1.2,0.04714879989624023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,4,64,power_law_1.2,0.04371840059757233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,4,64,power_law_1.2,0.045407998561859134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,4,64,power_law_1.2,0.04645760059356689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,4,64,power_law_1.2,0.048204800486564635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,4,64,power_law_1.2,0.04681600034236908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,4,64,power_law_1.2,0.05133439898490906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,4,64,power_law_1.2,0.04856959879398346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,4,64,power_law_1.2,0.05173119902610779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,4,64,power_law_1.2,0.05047680139541626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,4,64,power_law_1.2,0.0542527973651886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,4,64,power_law_1.2,0.05063040256500244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,4,64,power_law_1.2,0.05144320130348205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,4,64,power_law_1.2,0.050963199138641356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,4,64,power_law_1.2,0.05294079780578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,4,64,power_law_1.2,0.0513152003288269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,4,64,power_law_1.2,0.05294719934463501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,4,64,power_law_1.2,0.05294719934463501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,4,64,power_law_1.2,0.05598719716072083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,4,64,power_law_1.2,0.05414400100708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,4,64,power_law_1.2,0.05491840243339539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,4,64,power_law_1.2,0.05492479801177978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,4,64,power_law_1.2,0.05826560258865356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,4,64,power_law_1.2,0.057683199644088745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,4,64,power_law_1.2,0.06017919778823853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,4,64,power_law_1.2,0.06376960277557372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,4,64,power_law_1.2,0.06883839964866638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,4,64,power_law_1.2,0.06558079719543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,4,64,power_law_1.2,0.07061759829521179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,4,64,power_law_1.2,0.07164160013198853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,4,64,power_law_1.2,0.07667199969291687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,4,64,power_law_1.2,0.08862079977989197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,4,64,power_law_1.2,0.08736000061035157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,4,64,power_law_1.2,0.09694079756736755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,4,64,power_law_1.2,0.09733759760856628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,4,64,power_law_1.2,0.12275840044021606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,4,64,power_law_1.2,0.11112320423126221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,4,64,power_law_1.2,0.1460479974746704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,4,64,power_law_1.2,0.13194880485534669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,4,64,power_law_1.2,0.1974720001220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,4,64,power_law_1.2,0.16631040573120118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,4,64,power_law_1.2,0.24343678951263428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,4,64,power_law_1.2,0.06398720145225525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,4,64,power_law_1.2,0.201964807510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,4,64,power_law_1.2,0.3322432041168213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,4,64,power_law_1.2,0.2987263917922974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,4,64,power_law_1.2,0.061952000856399535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,4,64,power_law_1.2,0.4273280143737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,4,64,power_law_1.2,0.3953855991363525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,4,64,power_law_1.2,0.05470079779624939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,4,64,power_law_1.2,0.6212992191314697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,4,64,power_law_1.2,0.5572927951812744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,4,64,power_law_1.2,0.058156800270080564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,4,64,power_law_1.2,0.8936511993408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,4,64,power_law_1.2,0.6700416088104248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,4,64,power_law_1.2,0.05914239883422852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,4,64,power_law_1.2,1.1415936470031738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,4,64,power_law_1.2,0.8341888427734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,4,64,power_law_1.2,0.06321280002593994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,4,64,power_law_1.2,1.6702655792236327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,4,64,power_law_1.2,1.2437567710876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,4,64,power_law_1.2,0.06172800064086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,4,64,power_law_1.2,3.3740673065185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,4,64,power_law_1.2,0.06211199760437012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,4,64,power_law_1.2,2.5662208557128907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,4,64,power_law_1.2,0.06534399986267089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,4,64,power_law_1.2,0.06458879709243774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,4,64,power_law_1.2,0.06491519808769226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,4,64,power_law_1.2,0.06666240096092224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,4,64,power_law_1.2,0.06610559821128845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,4,64,power_law_1.2,0.06826879978179931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,4,64,power_law_1.2,0.07356799840927124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,4,64,power_law_1.2,0.07308160066604615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,4,64,power_law_1.2,0.07856000065803528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,4,64,power_law_1.2,0.08775039911270141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.2,0.09482240080833435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.2,0.1192255973815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.2,0.13761919736862183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.2,0.18031359910964967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.2,0.21330559253692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.2,0.28387200832366943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.2,0.36401278972625734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.2,0.5336383819580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.2,0.6991360187530518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,4,64,power_law_1.2,0.9310463905334473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,4,64,power_law_1.2,1.4772543907165527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,4,64,power_law_1.2,3.065452766418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,4,64,power_law_1.2,0.023129600286483764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,4,64,power_law_1.2,0.024294400215148927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,4,64,power_law_1.2,0.020275199413299562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,4,64,power_law_1.2,0.02040960043668747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,4,64,power_law_1.2,0.02134400010108948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,4,64,power_law_1.2,0.021715199947357176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,4,64,power_law_1.2,0.02276480048894882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,4,64,power_law_1.2,0.02654080092906952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,4,64,power_law_1.2,0.02597759962081909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,4,64,power_law_1.2,0.026655998826026917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,4,64,power_law_1.2,0.04300160109996796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,4,64,power_law_1.2,0.042752000689506534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,4,64,power_law_1.2,0.042822399735450746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,4,64,power_law_1.2,0.042233601212501526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,4,64,power_law_1.2,0.044249600172042845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,4,64,power_law_1.2,0.046188798546791074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,4,64,power_law_1.2,0.04906879961490631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,4,64,power_law_1.2,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.2,0.0547327995300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.2,0.06728320121765137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.2,0.07669119834899903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.2,0.10814080238342286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.2,0.1304255962371826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.2,0.18500479459762573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.2,0.24618880748748778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.2,0.33410561084747314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.2,0.44133758544921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,32,balanced,0.027482666075229645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,4,64,power_law_1.2,0.5711103916168213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,32,balanced,0.022874665757020313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,4,64,power_law_1.2,0.8609536170959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,32,balanced,0.02109333376089732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,32,balanced,0.021381333470344543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,4,64,power_law_1.2,1.8295295715332032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,32,balanced,0.022453332940737408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,32,balanced,0.024821333587169647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,32,balanced,0.025226667523384094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,32,balanced,0.025413334369659424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,32,balanced,0.02722666660944621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,32,balanced,0.049733335773150124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,32,balanced,0.027301333844661713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,32,balanced,0.04974400003751119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,32,balanced,0.030970667799313862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,32,balanced,0.04974933465321859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,32,balanced,0.03365333378314972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,32,balanced,0.05192000170548757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,32,balanced,0.033285332222779594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,32,balanced,0.05412800113360087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,32,balanced,0.04609066744645437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,32,balanced,0.055685331424077354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,32,balanced,0.05786666770776113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,32,balanced,0.05454400181770325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,32,balanced,0.058880001306533813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,32,balanced,0.052426666021347046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,32,balanced,0.05797333518664042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,32,balanced,0.05277333160241445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,32,balanced,0.05793066819508871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,32,balanced,0.06002666552861532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,32,balanced,0.05987200140953064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,32,balanced,0.060122668743133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,32,balanced,0.06895466645558675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,32,balanced,0.06131199995676676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,32,balanced,0.062314664324124656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,32,balanced,0.08257600168387096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,32,balanced,0.0664106657107671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,32,balanced,0.09672000010808308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,32,balanced,0.06674133241176605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,32,balanced,0.07027733325958252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,32,balanced,0.13235732913017273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,32,balanced,0.07669866581757863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,32,balanced,0.0807360013326009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,32,balanced,0.1584106683731079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,32,balanced,0.09483200311660767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,32,balanced,0.10703999797503154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,32,balanced,0.22032533089319864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,32,balanced,0.13984533150990805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,32,balanced,0.16267200311024985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,32,balanced,0.2834186752637227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,32,balanced,0.2091040015220642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,32,balanced,0.4121760129928589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,32,power_law_1.01,0.07032319903373718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,32,balanced,0.2629493276278178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,32,power_law_1.01,0.06520320177078247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,32,power_law_1.01,0.05626239776611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,32,power_law_1.01,0.05687040090560913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,32,power_law_1.01,0.06170240044593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,32,power_law_1.01,0.06312959790229797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,32,power_law_1.01,0.0644864022731781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,32,power_law_1.01,0.0670144021511078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,32,power_law_1.01,0.06654080152511596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,32,balanced,0.5437013308207194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,32,power_law_1.01,0.06716160178184509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,32,balanced,0.3551146586736043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,32,power_law_1.01,0.06883199810981751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,32,power_law_1.01,0.06858239769935608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,32,power_law_1.01,0.0692031979560852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,32,power_law_1.01,0.07252479791641235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,32,power_law_1.01,0.07594879865646362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,32,power_law_1.01,0.07671679854393006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,32,power_law_1.01,0.07978240251541138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,32,power_law_1.01,0.08833280205726624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.01,0.0995519995689392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.01,0.11856640577316284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,32,balanced,0.44684267044067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.01,0.13348480463027954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.01,0.17594239711761475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,32,balanced,0.6865599950154623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.01,0.20388479232788087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.01,0.27991681098937987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.01,0.3601279973983765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.01,0.48497920036315917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.01,0.6588160037994385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,32,power_law_1.01,0.8057472229003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,32,power_law_1.01,1.2853759765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,32,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,32,balanced,0.06001600126425425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,32,balanced,0.549343983332316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,32,power_law_1.01,2.425772857666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,32,balanced,0.04553066690762838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,32,balanced,0.046037331223487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,32,balanced,0.045552000403404236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,32,balanced,0.04772266745567322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,32,balanced,0.0459146648645401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,32,balanced,0.04775466521581014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,32,balanced,0.0476746658484141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,32,balanced,0.05087466537952423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,32,balanced,0.04991999765237173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,32,balanced,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,32,balanced,0.05208000044027964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,32,balanced,0.053957333167394005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,32,balanced,0.05376000205675761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,32,balanced,0.05497066676616669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,32,balanced,0.054010664423306785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,32,balanced,0.056128000219662987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,32,balanced,0.0537120004494985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,32,balanced,0.05593599875768026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,32,balanced,0.05595199763774872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,32,balanced,1.0712160269419353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,32,balanced,0.05795200169086456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,32,balanced,0.05585599939028422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,32,balanced,0.05885866781075796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,32,balanced,0.05402666827042898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,32,balanced,0.06028266747792562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,32,balanced,0.058143998185793556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,32,balanced,0.06217066446940104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,32,balanced,0.06073066592216492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,32,balanced,0.0643146683772405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,32,balanced,0.0682666649421056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,32,balanced,0.06860266625881195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,32,balanced,0.07445333401362102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,32,balanced,0.8212266763051351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,32,balanced,0.06644799808661143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,32,balanced,0.08092799782752991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,32,balanced,0.07148266832033794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,32,balanced,0.1092639962832133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,32,balanced,0.07897066573301952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,32,balanced,0.08438400427500407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,32,balanced,0.12601066629091898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,32,balanced,0.09471467137336731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,32,balanced,0.16542399923006693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,32,balanced,0.10532800356547038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,32,balanced,0.19883733987808228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,32,balanced,0.12780800461769104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,32,power_law_1.01,0.03118720054626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,32,balanced,0.2670666575431824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,32,balanced,0.1441386640071869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,32,power_law_1.01,0.0226623997092247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,32,power_law_1.01,0.020787200331687926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,32,power_law_1.01,0.02021760046482086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,32,power_law_1.01,0.02131839990615845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,32,power_law_1.01,0.0225600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,32,balanced,0.333621342976888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,32,balanced,0.18414932489395142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,32,power_law_1.01,0.023846399784088135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,32,power_law_1.01,0.024383999407291412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,32,power_law_1.01,0.024377599358558655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,32,power_law_1.01,0.02502399981021881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,32,power_law_1.01,0.02884480059146881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,32,balanced,0.22644267479578653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,32,power_law_1.01,0.029766398668289184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,32,balanced,0.47274665037790936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,32,power_law_1.01,0.030092799663543703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,32,balanced,2.1139626502990723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,32,power_law_1.01,0.04521600008010864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,32,power_law_1.01,0.04629760086536407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,32,power_law_1.01,0.04708479940891266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,32,power_law_1.01,0.07466880083084107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,32,power_law_1.01,0.05082240104675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,32,power_law_1.01,0.049702399969100954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,32,power_law_1.01,0.05123839974403381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,32,power_law_1.01,0.05127040147781372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,32,balanced,0.3158559997876485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,32,power_law_1.01,0.053855997323989865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,32,power_law_1.01,0.05021439790725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,32,power_law_1.01,0.04792320132255554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,32,power_law_1.01,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.01,0.05898879766464234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,32,balanced,0.605951984723409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,32,power_law_1.01,0.05178880095481873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,32,power_law_1.01,0.045612800121307376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,32,power_law_1.01,0.04686720073223114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,32,power_law_1.01,0.0542464017868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,32,balanced,1.6128746668497722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.01,0.06842240095138549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,32,power_law_1.01,0.04814079999923706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,32,power_law_1.01,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,32,power_law_1.01,0.05061119794845581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.01,0.07750399708747864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,32,power_law_1.01,0.05716480016708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,32,power_law_1.01,0.05182719826698303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,32,power_law_1.01,0.05716480016708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,32,power_law_1.01,0.05194240212440491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.01,0.09838719964027405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,32,balanced,0.3826933304468791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,32,power_law_1.01,0.052172797918319705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,32,power_law_1.01,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,32,power_law_1.01,0.05367040038108826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.01,0.11307519674301147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,32,power_law_1.01,0.05919359922409058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,32,power_law_1.01,0.05469440221786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,32,power_law_1.01,0.05676800012588501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,32,power_law_1.01,0.05985919833183288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.01,0.16736639738082887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,32,power_law_1.01,0.05916799902915955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,32,balanced,0.747221310933431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,32,power_law_1.01,0.05978239774703979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.01,0.1905087947845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,32,power_law_1.01,0.06525440216064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,32,power_law_1.01,0.06328319907188415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.01,0.2873791933059692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,32,power_law_1.01,0.06755840182304382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,32,balanced,0.47016533215840656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,32,power_law_1.01,0.07025279998779296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.01,0.35075840950012205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,32,power_law_1.01,0.07227519750595093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,32,power_law_1.01,0.07400959730148315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,32,power_law_1.01,0.08736640214920044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,32,power_law_1.01,0.45516161918640136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,32,power_law_1.01,0.07912319898605347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,32,power_law_1.01,0.0959231972694397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,32,power_law_1.01,0.08675199747085571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,32,power_law_1.01,0.12291200160980224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,32,power_law_1.01,0.70414719581604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,32,power_law_1.01,0.09045119881629944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,32,power_law_1.01,0.14581120014190674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,32,power_law_1.01,0.18443520069122316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,32,power_law_1.01,0.11087360382080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,32,power_law_1.01,1.3628543853759765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,32,balanced,1.1335039933522542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,32,power_law_1.01,0.23251841068267823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,32,power_law_1.01,0.12568960189819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,32,power_law_1.01,0.3133375883102417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,32,power_law_1.01,0.15712000131607057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,32,power_law_1.01,0.41208319664001464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,32,power_law_1.01,0.6277503967285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,32,power_law_1.01,0.1816831946372986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,32,power_law_1.01,0.26784000396728513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,32,power_law_1.01,0.7659647941589356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,32,power_law_1.01,0.32079999446868895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,32,power_law_1.01,0.931334400177002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,32,power_law_1.01,0.47728638648986815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,32,balanced,0.6797066529591879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,32,power_law_1.01,1.447417640686035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,32,power_law_1.01,0.5651008129119873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,32,power_law_1.01,2.951628875732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,32,power_law_1.01,0.6485695838928223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,32,power_law_1.01,1.0304256439208985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,32,power_law_1.01,1.9695295333862304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,8,32,power_law_1.2,0.07470080256462097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,8,32,power_law_1.2,0.048582398891448976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,8,32,power_law_1.2,0.0440064013004303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,8,32,power_law_1.2,0.047712001204490664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,8,32,power_law_1.2,0.048390400409698484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,8,32,power_law_1.2,0.05238400101661682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,8,32,power_law_1.2,0.05475199818611145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,8,32,power_law_1.2,0.055961602926254274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,8,32,power_law_1.2,0.05729920268058777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,8,32,power_law_1.2,0.05546240210533142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,8,32,power_law_1.2,0.0562175989151001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,8,32,power_law_1.2,0.05751039981842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,8,32,power_law_1.2,0.060844802856445314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,8,32,power_law_1.2,0.06348159909248352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,8,32,power_law_1.2,0.07070080041885377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,8,32,power_law_1.2,0.07070080041885377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,8,32,power_law_1.2,0.07594239711761475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,8,32,power_law_1.2,0.08679680228233337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,8,32,power_law_1.2,0.09537280201911927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,32,balanced,2.233802636464437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,8,32,power_law_1.2,0.11949440240859985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,8,32,power_law_1.2,0.1333567976951599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,8,32,power_law_1.2,0.17278079986572265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,8,32,power_law_1.2,0.20245120525360108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,8,32,power_law_1.2,0.28501119613647463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,8,32,power_law_1.2,0.35589120388031004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,8,32,power_law_1.2,0.5725247859954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,8,32,power_law_1.2,0.651910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,8,32,power_law_1.2,0.8411456108093261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,32,balanced,1.3217706680297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,8,32,power_law_1.2,1.0927807807922363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,8,32,power_law_1.2,2.914681625366211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,8,32,power_law_1.2,0.04715520143508911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,8,32,power_law_1.2,0.047244799137115476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,8,32,power_law_1.2,0.043808001279830935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,8,32,power_law_1.2,0.04676479995250702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,8,32,power_law_1.2,0.047539201378822324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,8,32,power_law_1.2,0.04961279928684235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,8,32,power_law_1.2,0.05098239779472351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,8,32,power_law_1.2,0.05198720097541809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,8,32,power_law_1.2,0.07092480063438415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,8,32,power_law_1.2,0.05208960175514221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,8,32,power_law_1.2,0.05273600220680237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,8,32,power_law_1.2,0.06481919884681701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,8,32,power_law_1.2,0.054041600227355956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,8,32,power_law_1.2,0.05355520248413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,8,32,power_law_1.2,0.055929601192474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,8,32,power_law_1.2,0.05589119791984558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,8,32,power_law_1.2,0.058483201265335086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,8,32,power_law_1.2,0.0576960027217865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,8,32,power_law_1.2,0.06223359704017639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,8,32,power_law_1.2,0.06069120168685913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,8,32,power_law_1.2,0.06584960222244263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,8,32,power_law_1.2,0.0660863995552063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,8,32,power_law_1.2,0.06740480065345764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,8,32,power_law_1.2,0.06775040030479432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,8,32,power_law_1.2,0.07109760046005249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,8,32,power_law_1.2,0.06730239987373351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,8,32,power_law_1.2,0.07364479899406433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,8,32,power_law_1.2,0.06802560091018676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,8,32,power_law_1.2,0.08700799942016602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,8,32,power_law_1.2,0.07033600211143494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,8,32,power_law_1.2,0.06825600266456604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,8,32,power_law_1.2,0.0985472023487091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,8,32,power_law_1.2,0.07056639790534973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,8,32,power_law_1.2,0.1274623990058899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,8,32,power_law_1.2,0.0767359972000122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,8,32,power_law_1.2,0.07690240144729614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,8,32,power_law_1.2,0.1513856053352356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,8,32,power_law_1.2,0.08147199749946595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,8,32,power_law_1.2,0.21643519401550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,8,32,power_law_1.2,0.09063040018081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,8,32,power_law_1.2,0.2503999948501587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.2,0.10197759866714477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,8,32,power_law_1.2,0.34318718910217283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.2,0.1230463981628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,8,32,power_law_1.2,0.44467840194702146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.2,0.14116480350494384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,8,32,power_law_1.2,0.6701824188232421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.2,0.1804800033569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.2,0.21821439266204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,8,32,power_law_1.2,0.8211071968078614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.2,0.3108992099761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,8,32,power_law_1.2,1.0903807640075684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.2,0.40280961990356445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.2,0.5849215984344482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,8,32,power_law_1.2,1.8209983825683593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.2,0.6871871948242188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,8,32,power_law_1.2,3.4629184722900392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,8,32,power_law_1.2,0.948038387298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,8,32,power_law_1.2,1.448300838470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,8,32,power_law_1.2,2.9359359741210938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,8,32,power_law_1.2,0.030899199843406677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,8,32,power_law_1.2,0.02237440049648285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,8,32,power_law_1.2,0.01998720020055771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,8,32,power_law_1.2,0.02022400051355362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,8,32,power_law_1.2,0.021529600024223328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,8,32,power_law_1.2,0.022355200350284578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,8,32,power_law_1.2,0.023929600417613984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,8,32,power_law_1.2,0.02426239997148514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,8,32,power_law_1.2,0.02457599937915802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,8,32,power_law_1.2,0.025235199928283693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,8,32,power_law_1.2,0.028275200724601747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,8,32,power_law_1.2,0.029158401489257812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,8,32,power_law_1.2,0.029548799991607665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,8,32,power_law_1.2,0.04465279877185822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,8,32,power_law_1.2,0.04688000082969666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,8,32,power_law_1.2,0.04614399969577789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,8,32,power_law_1.2,0.04997119903564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,8,32,power_law_1.2,0.051712000370025636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.2,0.05861120223999024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.2,0.06886399984359741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.2,0.07642239928245545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.2,0.09910399913787842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.2,0.12373119592666626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.2,0.17339520454406737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.2,0.22059519290924073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.2,0.32216320037841795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.2,0.3963200092315674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,8,32,power_law_1.2,0.4952511787414551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,8,32,power_law_1.2,0.7799551963806153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,8,32,power_law_1.2,1.5385472297668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,16,balanced,0.05017599960168203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,16,balanced,0.052069331208864846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,16,balanced,0.052015999952952065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,16,balanced,0.054133335749308266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,16,balanced,0.05781333148479462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,16,balanced,0.06262399752934773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,16,balanced,0.06464000046253204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,16,balanced,0.0659093310435613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,16,balanced,0.06492800017197926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,16,balanced,0.06433066725730896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,16,balanced,0.0670666644970576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,16,balanced,0.0664106657107671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,16,balanced,0.06810133159160614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,16,balanced,0.06968000034491222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,16,balanced,0.07453866799672444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,16,balanced,0.021045332153638203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,16,balanced,0.0745119998852412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,16,balanced,0.021349333226680756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,16,balanced,0.07851733267307281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,16,balanced,0.023071999351183575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,16,balanced,0.08680533369382222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,16,balanced,0.022965334355831146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,16,balanced,0.0232640008131663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,16,balanced,0.09282132983207703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,16,balanced,0.025381334125995636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,16,balanced,0.10899200042088826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,16,balanced,0.02643733223279317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,16,balanced,0.027232001225153606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,16,balanced,0.1237546702226003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,16,balanced,0.028181334336598713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,16,balanced,0.029178666571776073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,16,balanced,0.15568000078201294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,16,balanced,0.030805334448814392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,16,balanced,0.03194133440653483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,16,balanced,0.03370666752258936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,16,balanced,0.19031999508539835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,16,balanced,0.03797333439191183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,16,balanced,0.041349334021409355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,16,balanced,0.04346133271853129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,16,balanced,0.25085333983103436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,16,balanced,0.05965333183606466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,16,balanced,0.06772799789905548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,16,balanced,0.0772213339805603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,16,balanced,0.3053599993387858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,16,balanced,0.08925333619117737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,16,balanced,0.10291199882825215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,16,balanced,0.1297760009765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,16,balanced,0.41833066940307617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,16,balanced,0.16362667083740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,16,balanced,0.21521600087483725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,16,balanced,0.5331413348515829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,16,balanced,0.26577599843343097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,16,balanced,0.056277334690093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,16,balanced,0.3895039955774943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,16,balanced,0.04460266729195913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,16,balanced,0.04789866507053375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,16,balanced,0.6554506619771322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,16,balanced,0.04817600051561991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,16,balanced,0.05172266562779745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,16,balanced,0.05747200051943461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,16,balanced,0.059978668888409935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,16,balanced,0.05864533285299937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,16,balanced,0.04491733511288961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,16,balanced,0.5034506718317667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,16,balanced,0.06029333174228668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,16,balanced,0.045968001087506614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,16,balanced,0.060592000683148704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,16,balanced,0.04651733239491781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,16,balanced,0.06229866544405619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,16,balanced,0.047839999198913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,16,balanced,0.06234133243560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,16,balanced,0.04844800134499868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,16,balanced,0.06533866624037425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,16,balanced,0.054272000988324486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,16,balanced,1.0031253496805828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,16,balanced,0.06463466584682465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,16,balanced,0.05518933137257894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,16,balanced,0.6345973412195841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,16,balanced,0.07090133428573608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,16,balanced,0.05630399783452352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,16,balanced,0.07082666456699371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,16,balanced,0.05677333474159241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,16,balanced,0.07670933504899342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,16,balanced,0.0584746648867925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,16,balanced,0.08624000350634257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,16,balanced,0.05761066575845083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,16,balanced,0.08995733658472697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,16,balanced,0.06031466523806254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,16,balanced,0.10340799887975057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,16,balanced,0.06018666426340739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,16,balanced,0.06222933530807495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,16,balanced,0.11328533291816711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,16,balanced,0.06659733255704244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,16,balanced,0.06855466465155284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,16,balanced,0.1405119995276133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,16,balanced,1.0031466484069824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,16,balanced,0.0726506660381953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,16,balanced,0.1660160024960836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,16,balanced,0.08282133440176646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,16,balanced,0.08706667025883992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,16,balanced,0.21334399779637656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,16,balanced,0.11750933527946472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,16,power_law_1.01,0.05739520192146301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,16,balanced,0.13620266318321228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,16,power_law_1.01,0.0542464017868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,16,balanced,1.965461254119873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,16,balanced,0.2656586567560832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,16,power_law_1.01,0.047366398572921756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,16,balanced,0.1798080007235209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,16,power_law_1.01,0.047916799783706665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,16,power_law_1.01,0.05103999972343445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,16,balanced,0.37322131792704266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,16,balanced,0.2173866629600525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,16,power_law_1.01,0.05438079833984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,16,power_law_1.01,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,16,power_law_1.01,0.05877760052680969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,16,balanced,1.9920159975687664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,16,balanced,0.29552533229192096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,16,power_law_1.01,0.05921279788017273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,16,balanced,0.4533226490020752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,16,power_law_1.01,0.05930240154266357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,16,power_law_1.01,0.06004480123519897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,16,power_law_1.01,0.06305279731750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,16,balanced,0.37564265727996826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,16,power_law_1.01,0.0630079984664917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,16,power_law_1.01,0.0662335991859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,16,power_law_1.01,0.0726207971572876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,16,power_law_1.01,0.07406079769134521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,16,balanced,0.5632586479187012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,16,balanced,0.5267626841862997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,16,power_law_1.01,0.08283519744873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,16,power_law_1.01,0.09644799828529357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,16,power_law_1.01,0.09623680114746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,16,power_law_1.01,0.11259520053863525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,16,power_law_1.01,0.13343360424041747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,16,power_law_1.01,0.16933120489120485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,16,power_law_1.01,0.2006848096847534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,16,balanced,0.6806879838307699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,16,power_law_1.01,0.2866431951522827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,16,power_law_1.01,0.34225280284881593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,16,power_law_1.01,0.48755841255187987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,16,power_law_1.01,0.6412735939025879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,16,balanced,0.8359039624532064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,16,power_law_1.01,0.7332543849945068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,16,power_law_1.01,0.054054397344589236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,16,power_law_1.01,0.04855040013790131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,16,power_law_1.01,1.1249855995178222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,16,power_law_1.01,0.06508799791336059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,16,balanced,0.840831995010376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,16,power_law_1.01,0.06438400149345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,16,power_law_1.01,0.05454720258712768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,16,power_law_1.01,2.3382591247558593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,16,power_law_1.01,0.055340802669525145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,16,power_law_1.01,0.04772480130195618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,16,power_law_1.01,0.05545600056648255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,16,power_law_1.01,0.04776960015296936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,16,power_law_1.01,0.05854079723358154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,16,power_law_1.01,0.048921599984169006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,16,power_law_1.01,0.060838401317596436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,16,power_law_1.01,0.0510591983795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,16,power_law_1.01,0.06136959791183472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,16,power_law_1.01,0.05289599895477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,16,power_law_1.01,0.06204800009727478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,16,power_law_1.01,0.054451197385787964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,16,power_law_1.01,0.06302719712257385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,16,power_law_1.01,0.05445759892463684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,16,power_law_1.01,0.06447359919548035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,16,power_law_1.01,0.055302399396896365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,16,power_law_1.01,0.06607360243797303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,16,power_law_1.01,0.05704960227012634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,16,power_law_1.01,0.06687359809875489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,16,power_law_1.01,0.058905601501464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,16,balanced,1.2870453198750813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,16,power_law_1.01,0.0675711989402771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,16,power_law_1.01,0.0606656014919281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,16,power_law_1.01,0.07322880029678344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,16,power_law_1.01,0.06369280219078063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,16,power_law_1.01,0.07715839743614197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,16,power_law_1.01,0.0690559983253479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,16,power_law_1.01,0.08206080198287964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,16,power_law_1.01,0.07323520183563233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,16,power_law_1.01,0.02220800071954727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,16,balanced,1.6325920422871907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,16,power_law_1.01,0.0932864010334015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,16,power_law_1.01,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,16,power_law_1.01,0.022899200022220612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.01,0.10316799879074097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,16,power_law_1.01,0.022899200022220612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,16,power_law_1.01,0.0945792019367218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.01,0.1264448046684265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,16,power_law_1.01,0.022566400468349457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,16,power_law_1.01,0.09976320266723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.01,0.1452288031578064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,16,power_law_1.01,0.02457599937915802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,16,power_law_1.01,0.13051520586013793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.01,0.18338559865951537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,16,power_law_1.01,0.023878400027751923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,16,power_law_1.01,0.15382399559020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.01,0.21923840045928955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,16,power_law_1.01,0.024243199825286867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,16,power_law_1.01,0.21105918884277344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.01,0.31259520053863527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,16,power_law_1.01,0.2499392032623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,16,power_law_1.01,0.024851199984550477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.01,0.3936064004898071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,16,power_law_1.01,0.34323840141296386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,16,power_law_1.01,0.030163198709487915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.01,0.5356160163879394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,16,power_law_1.01,0.46596479415893555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,16,power_law_1.01,0.030687999725341798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.01,0.6690432071685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,16,power_law_1.01,0.6627071857452392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,16,power_law_1.01,0.03126400113105774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,16,power_law_1.01,0.8548543930053711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,16,power_law_1.01,0.8186559677124023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,16,balanced,2.548367977142334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,16,power_law_1.01,0.05560960173606873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,16,power_law_1.01,1.3019840240478515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,16,power_law_1.01,1.0127679824829101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,16,power_law_1.01,0.05669119954109192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,16,power_law_1.01,2.6401472091674805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,16,power_law_1.01,1.624857521057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,16,power_law_1.01,0.056831997632980344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,16,power_law_1.01,3.2060928344726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,16,power_law_1.01,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,16,power_law_1.01,0.05027840137481689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,16,power_law_1.01,0.053350400924682614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,16,power_law_1.01,0.05794559717178345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.01,0.06490240097045899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.01,0.07357439994812012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.01,0.0809984028339386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.01,0.10224640369415283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.01,0.12297600507736206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.01,0.1608191967010498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.01,0.20863358974456786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.01,0.2920703887939453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.01,0.3665152072906494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,16,power_law_1.01,0.4582528114318848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,16,power_law_1.01,0.6984000205993652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,16,power_law_1.01,1.4234496116638184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,16,16,power_law_1.2,0.05570560097694397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,16,16,power_law_1.2,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,16,16,power_law_1.2,0.04438399970531463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,16,16,power_law_1.2,0.0477183997631073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,16,16,power_law_1.2,0.050425601005554196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,16,16,power_law_1.2,0.052127999067306516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,16,16,power_law_1.2,0.05742719769477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,16,16,power_law_1.2,0.058483201265335086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,16,16,power_law_1.2,0.05912960171699524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,16,16,power_law_1.2,0.04853120148181915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,16,16,power_law_1.2,0.05983359813690185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,16,16,power_law_1.2,0.06120960116386413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,16,16,power_law_1.2,0.05991680026054382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,16,16,power_law_1.2,0.06258559823036194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,16,16,power_law_1.2,0.044460800290107724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,16,16,power_law_1.2,0.06609280109405517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,16,16,power_law_1.2,0.048691201210021975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,16,16,power_law_1.2,0.0680191993713379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,16,16,power_law_1.2,0.048844799399375916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,16,16,power_law_1.2,0.0748799979686737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,16,16,power_law_1.2,0.051046401262283325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,16,16,power_law_1.2,0.07770879864692688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,16,16,power_law_1.2,0.053414398431777955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,16,16,power_law_1.2,0.08475520014762879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,16,16,power_law_1.2,0.054579198360443115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,16,16,power_law_1.2,0.09799039959907532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,16,16,power_law_1.2,0.05546240210533142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,16,16,power_law_1.2,0.09932159781455993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,16,16,power_law_1.2,0.05532159805297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,16,16,power_law_1.2,0.1224128007888794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,16,16,power_law_1.2,0.058259201049804685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,16,16,power_law_1.2,0.13836159706115722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,16,16,power_law_1.2,0.060211199522018435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,16,16,power_law_1.2,0.17508480548858643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,16,16,power_law_1.2,0.061375999450683595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,16,16,power_law_1.2,0.23208959102630616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,16,16,power_law_1.2,0.06553599834442139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,16,16,power_law_1.2,0.3031167984008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,16,16,power_law_1.2,0.06976640224456787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,16,16,power_law_1.2,0.3765631914138794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,16,16,power_law_1.2,0.07592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,16,16,power_law_1.2,0.5604544162750245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,16,16,power_law_1.2,0.08348159790039063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,16,16,power_law_1.2,0.7245696067810059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,16,16,power_law_1.2,0.09771519899368286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,16,16,power_law_1.2,0.8515968322753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,256,16,16,power_law_1.2,0.05425919890403748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,16,16,power_law_1.2,0.10607999563217163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,16,16,power_law_1.2,1.3124671936035157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,256,16,16,power_law_1.2,0.06541439890861511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,16,16,power_law_1.2,0.13430399894714357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,16,16,power_law_1.2,2.7840576171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,256,16,16,power_law_1.2,0.049779200553894044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,16,16,power_law_1.2,0.1598528027534485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,256,16,16,power_law_1.2,0.0543936014175415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,16,16,power_law_1.2,0.2073728084564209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,256,16,16,power_law_1.2,0.05589119791984558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,16,16,power_law_1.2,0.28383359909057615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,256,16,16,power_law_1.2,0.05768960118293762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,16,16,power_law_1.2,0.387007999420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,256,16,16,power_law_1.2,0.060678398609161376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,16,16,power_law_1.2,0.47940478324890134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,256,16,16,power_law_1.2,0.061791998147964475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,16,16,power_law_1.2,0.7036159992218017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,256,16,16,power_law_1.2,0.06302080154418946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,16,16,power_law_1.2,0.96693115234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,256,16,16,power_law_1.2,0.06297600269317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,256,16,16,power_law_1.2,0.0641152024269104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,16,16,power_law_1.2,1.2953984260559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,256,16,16,power_law_1.2,0.06591359972953796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,16,16,power_law_1.2,1.7772607803344727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,256,16,16,power_law_1.2,0.06717439889907836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,256,16,16,power_law_1.2,0.06809599995613098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,7168,2048,8,256,16,16,power_law_1.2,0.02168319970369339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,16,16,power_law_1.2,3.6375999450683594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,256,16,16,power_law_1.2,0.07605119943618774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,256,16,16,power_law_1.2,0.07916160225868225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,7168,2048,8,256,16,16,power_law_1.2,0.022860799729824067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,256,16,16,power_law_1.2,0.0839680016040802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,7168,2048,8,256,16,16,power_law_1.2,0.022387200593948366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,256,16,16,power_law_1.2,0.0986624002456665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,7168,2048,8,256,16,16,power_law_1.2,0.0225600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.2,0.10522240400314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,7168,2048,8,256,16,16,power_law_1.2,0.02439039945602417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.2,0.14074239730834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,7168,2048,8,256,16,16,power_law_1.2,0.02378239929676056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.2,0.15311360359191895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,7168,2048,8,256,16,16,power_law_1.2,0.024243199825286867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.2,0.19311360120773316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,7168,2048,8,256,16,16,power_law_1.2,0.02489600032567978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.2,0.24563200473785402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,7168,2048,8,256,16,16,power_law_1.2,0.02917119860649109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.2,0.3398272037506104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,7168,2048,8,256,16,16,power_law_1.2,0.029824000597000123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.2,0.42507519721984866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,7168,2048,8,256,16,16,power_law_1.2,0.029843199253082275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.2,0.6030335903167725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,7168,2048,8,256,16,16,power_law_1.2,0.055936002731323244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.2,0.8710783958435059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,7168,2048,8,256,16,16,power_law_1.2,0.056627202033996585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,256,16,16,power_law_1.2,1.0101632118225097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,7168,2048,8,256,16,16,power_law_1.2,0.05722879767417908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,256,16,16,power_law_1.2,1.6137472152709962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,7168,2048,8,256,16,16,power_law_1.2,0.05081599950790405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,256,16,16,power_law_1.2,3.5014976501464843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,7168,2048,8,256,16,16,power_law_1.2,0.05062400102615357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,7168,2048,8,256,16,16,power_law_1.2,0.0532800018787384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,7168,2048,8,256,16,16,power_law_1.2,0.05798400044441223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.2,0.06262400150299072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.2,0.07587839961051941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.2,0.0850816011428833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.2,0.10805120468139648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.2,0.12223360538482667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.2,0.17532800436019896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.2,0.22805120944976806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.2,0.31067519187927245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.2,0.3885440111160278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,7168,2048,8,256,16,16,power_law_1.2,0.5067840099334717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,7168,2048,8,256,16,16,power_law_1.2,0.7748864173889161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,7168,2048,8,256,16,16,power_law_1.2,1.6060735702514648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,128,balanced,0.03161599983771642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,128,balanced,0.03161066770553589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,128,balanced,0.03198933353026708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,128,balanced,0.03331733246644338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,128,balanced,0.03367999941110611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,128,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,128,balanced,0.03555733213822047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,128,balanced,0.03669333209594091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,128,balanced,0.03473066786924998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,128,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,128,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,128,balanced,0.03555733213822047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,128,balanced,0.03746666759252548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,128,balanced,0.03733866661787033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,128,balanced,0.03940266619126002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,128,balanced,0.039834665755430855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,128,balanced,0.03956266740957896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,128,balanced,0.04178133110205332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,128,balanced,0.041450666884581246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,128,balanced,0.04379733403523763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,128,balanced,0.04557866851488749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,128,balanced,0.045909335215886436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,8,balanced,0.04370133578777313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,128,balanced,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,8,balanced,0.045706664522488914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,8,balanced,0.04595733185609182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,8,balanced,0.04757866760094961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,8,balanced,0.05379199981689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,8,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,128,balanced,0.056101332108179726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,128,balanced,0.03209066639343897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,8,power_law_1.01,0.04856959879398346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,8,balanced,0.06630399823188782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,128,balanced,0.03182933231194814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,8,power_law_1.01,0.05251200199127197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,128,balanced,0.03317866722742716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,8,power_law_1.01,0.05048959851264954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,8,balanced,0.043978666265805565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,8,balanced,0.06708799799283345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,128,balanced,0.06300266583760579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,128,balanced,0.03382933388153712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,8,power_law_1.01,0.05004799962043762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,8,balanced,0.06844266752401988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,8,balanced,0.045941332976023354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,128,balanced,0.033573334415753685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,8,power_law_1.01,0.0541055977344513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,8,balanced,0.04761599997679392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,8,balanced,0.06923200190067291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,128,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,8,power_law_1.01,0.058719998598098753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,128,balanced,0.075914666056633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,8,balanced,0.04808533191680908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,8,balanced,0.07018133501211803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,128,balanced,0.035749333600203194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,8,power_law_1.01,0.04694400131702423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,8,power_law_1.01,0.06264320015907288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,8,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,128,balanced,0.035487999518712364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,8,balanced,0.06672533353169759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,8,power_law_1.01,0.05257599949836731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,8,balanced,0.06032533446947733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,8,power_law_1.01,0.06392319798469544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,128,balanced,0.03586133321126302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,8,balanced,0.0702400008837382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,8,power_law_1.01,0.047654399275779726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,8,balanced,0.060959999759991966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,8,power_law_1.01,0.06479359865188598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,128,balanced,0.03684266656637192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,128,balanced,0.09036800265312195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,8,power_law_1.01,0.04863359928131104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,8,balanced,0.06423999865849812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,8,balanced,0.07218666871388753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,8,power_law_1.01,0.06630399823188782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,128,balanced,0.037589333951473236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,8,power_law_1.01,0.05095679759979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,8,balanced,0.06434666613737743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,128,balanced,0.03789866715669632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,8,power_law_1.01,0.06885759830474854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,8,balanced,0.08046933511892955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,8,power_law_1.01,0.05432959794998169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,8,power_law_1.01,0.070796799659729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,128,balanced,0.036831999818483986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,8,balanced,0.06434666613737743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,8,balanced,0.08243733147780101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,8,power_law_1.01,0.05856000185012818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,8,power_law_1.01,0.07175040245056152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,8,power_law_1.01,0.058432000875473025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,128,balanced,0.03958933303753535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,8,balanced,0.06635199983914693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,8,balanced,0.08673600355784099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,128,balanced,0.1051093339920044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,8,power_law_1.01,0.07575680017471313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,8,power_law_1.01,0.060159999132156375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,128,balanced,0.04194133480389913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,8,balanced,0.06816533207893372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,8,balanced,0.09892266988754272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,8,power_law_1.01,0.08280959725379944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,8,power_law_1.01,0.061286401748657224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,128,balanced,0.04152533411979675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,8,power_law_1.01,0.08499199748039246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,8,balanced,0.06850133339564006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,8,power_law_1.01,0.062636798620224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,8,balanced,0.10194666186968486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,128,balanced,0.041877334316571556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,8,power_law_1.01,0.09218559861183166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,8,balanced,0.07009600102901459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,8,power_law_1.01,0.06500480175018311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,256,32,8,power_law_1.2,0.048665601015090945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,8,balanced,0.11733333269755046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,128,balanced,0.04363200068473816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,8,power_law_1.01,0.10543999671936036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,8,balanced,0.0763733337322871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,8,power_law_1.01,0.06919040083885193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,256,32,8,power_law_1.2,0.052960002422332765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,8,balanced,0.13195199767748514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,128,balanced,0.04576533536116282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,8,power_law_1.01,0.108787202835083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,128,balanced,0.14615466197331747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,8,balanced,0.07845866680145264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,8,power_law_1.01,0.07391999959945679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,256,32,8,power_law_1.2,0.04705919921398163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,8,balanced,0.1726613243420919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,128,balanced,0.04607999821503957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,8,power_law_1.01,0.13320319652557372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,8,balanced,0.08251200119654338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,8,power_law_1.01,0.08052480220794678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,256,32,8,power_law_1.2,0.050470399856567386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,8,power_law_1.01,0.1688704013824463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,128,balanced,0.050250664353370667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,8,balanced,0.20707199970881143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,8,balanced,0.09482666850090027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,8,power_law_1.01,0.08632320165634155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,256,32,8,power_law_1.2,0.05492479801177978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,8,power_law_1.01,0.19927040338516236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,128,balanced,0.05395199855168661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,8,power_law_1.01,0.08721920251846313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,8,balanced,0.10319466392199199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,256,32,8,power_law_1.2,0.05578240156173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,8,balanced,0.28150399525960285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,256,32,8,power_law_1.2,0.04824959933757782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,8,power_law_1.01,0.2674175977706909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,8,power_law_1.01,0.10474879741668701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,256,32,8,power_law_1.2,0.06338559985160827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,128,balanced,0.06028800209363302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,8,balanced,0.13499200344085693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,256,32,8,power_law_1.2,0.05294079780578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,8,power_law_1.01,0.3340735912322998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,8,balanced,0.34219733874003094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,8,power_law_1.01,0.11946239471435546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,256,32,8,power_law_1.2,0.06497920155525208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,128,balanced,0.07468266785144806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,8,power_law_1.01,0.15014400482177734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,8,balanced,0.15744533141454062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,256,32,8,power_law_1.2,0.047417598962783816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,8,power_law_1.01,0.4216000080108643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,128,balanced,0.25922133525212604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,256,32,8,power_law_1.2,0.06607360243797303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,8,power_law_1.01,0.18021119832992555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,8,balanced,0.21018133560816446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,8,balanced,0.48795731862386066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,256,32,8,power_law_1.2,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,8,power_law_1.01,0.6302847862243652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,256,32,8,power_law_1.2,0.06577280163764954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,128,balanced,0.09501333038012187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,8,power_law_1.01,0.2382591962814331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,8,balanced,0.2537173430124919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,8,power_law_1.01,0.30804479122161865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,256,32,8,power_law_1.2,0.05089920163154602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,8,power_law_1.01,0.7545407772064209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,256,32,8,power_law_1.2,0.06842880249023438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,8,power_law_1.01,0.41710720062255857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,8,balanced,0.6103200117746989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,256,32,8,power_law_1.2,0.053855997323989865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,8,balanced,0.3534880081812541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,8,power_law_1.01,1.0196800231933594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,128,balanced,0.13331733147303262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,256,32,8,power_law_1.2,0.07026559710502625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,8,power_law_1.01,0.5321152210235596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,256,32,8,power_law_1.2,0.05926399827003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,8,power_law_1.01,1.5942848205566407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,256,32,8,power_law_1.2,0.07230719923973083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,8,balanced,0.44933334986368817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,8,power_law_1.01,0.7773248195648194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,256,32,8,power_law_1.2,0.059494400024414064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,8,balanced,0.7605546315511068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,8,power_law_1.01,2.881990432739258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,256,32,8,power_law_1.2,0.07605119943618774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,8,power_law_1.01,0.9973952293395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,128,balanced,0.16484799981117249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,256,32,8,power_law_1.2,0.06110720038414001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,256,32,8,power_law_1.2,0.08321920037269592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,8,balanced,0.6489439805348715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,8,power_law_1.01,1.226252841949463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,256,32,8,power_law_1.2,0.061337602138519284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,256,32,8,power_law_1.2,0.08943359851837158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,256,32,8,power_law_1.2,0.06416640281677247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,256,32,8,power_law_1.2,0.09244800209999085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,8,power_law_1.01,1.8554431915283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,256,32,8,power_law_1.2,0.06707839965820313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,256,32,8,power_law_1.2,0.10455679893493652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,256,32,8,power_law_1.2,0.07012479901313781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,8,power_law_1.01,3.815769577026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,256,32,8,power_law_1.2,0.11953279972076417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,256,32,8,power_law_1.2,0.07507839798927307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,128,balanced,0.1973759929339091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,256,32,8,power_law_1.2,0.1464319944381714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,8,balanced,0.8352159659067789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,256,32,8,power_law_1.2,0.08177919983863831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,8,balanced,1.1426293055216472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,256,32,8,power_law_1.2,0.16637439727783204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,256,32,8,power_law_1.2,0.0887167990207672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,256,32,8,power_law_1.2,0.22295041084289552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,256,32,8,power_law_1.2,0.09262080192565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,256,32,8,power_law_1.2,0.2666879892349243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,256,32,8,power_law_1.2,0.10782079696655274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,256,32,8,power_law_1.2,0.3630016088485718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,256,32,8,power_law_1.2,0.12021119594573974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,256,32,8,power_law_1.2,0.48000001907348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,256,32,8,power_law_1.2,0.1611456036567688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,256,32,8,power_law_1.2,0.783955192565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,256,32,8,power_law_1.2,0.18511359691619872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,256,32,8,power_law_1.2,0.8832832336425781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,8,balanced,1.0316853523254395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,256,32,8,power_law_1.2,0.2550719976425171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,256,32,8,power_law_1.2,1.0651968002319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,256,32,8,power_law_1.2,0.3129535913467407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,256,32,8,power_law_1.2,1.8100799560546874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,256,32,8,power_law_1.2,0.4412543773651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,128,balanced,0.2860693335533142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,8,power_law_1.2,3.534841537475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,256,32,8,power_law_1.2,0.5779007911682129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,256,32,8,power_law_1.2,0.9696512222290039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,256,32,8,power_law_1.2,1.1049216270446778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,256,32,8,power_law_1.2,1.370310401916504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,8,power_law_1.2,2.0371007919311523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,8,power_law_1.2,4.176121520996094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,256,32,8,balanced,1.5930453936258953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,256,32,8,balanced,2.2590506871541343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,128,balanced,0.5306453307469686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,256,32,8,balanced,3.1551733016967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,128,balanced,0.03178133318821589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,128,balanced,0.0335359995563825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,128,balanced,0.033002667129039764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,128,balanced,0.03533333291610082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,128,balanced,0.033402666449546814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,128,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,128,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,128,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,128,balanced,0.035775999228159584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,128,balanced,0.03690666705369949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,128,balanced,0.037317333122094475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,128,balanced,0.037690666814645134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,128,balanced,0.03629333277543386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,128,balanced,0.03752533346414566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,128,balanced,0.04147200038035711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,128,balanced,0.03988266736268997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,128,balanced,0.04154666761557261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,128,balanced,0.043365334471066795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,128,balanced,0.04560000201066335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,128,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,128,balanced,0.0496373325586319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,128,balanced,0.017397332936525345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,128,balanced,0.052005335688591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,128,balanced,0.018816000471512478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,128,balanced,0.01907733331123988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,128,balanced,0.05796800057093302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,128,balanced,0.019952000429232914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,128,balanced,0.031632001201311745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,128,balanced,0.025221332907676697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,128,balanced,0.06806399921576183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,128,balanced,0.025424001117547352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,128,balanced,0.025600001215934753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,128,balanced,0.025221332907676697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,128,balanced,0.026554666459560394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,128,balanced,0.0786186655362447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,128,balanced,0.027306665976842243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,128,balanced,0.02759466568628947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,128,balanced,0.027845333019892376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,128,balanced,0.02934933453798294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,128,power_law_1.01,0.031404799222946166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,128,balanced,0.029738667110602062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,128,balanced,0.10265599687894185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,128,power_law_1.01,0.03155840039253235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,128,balanced,0.029658667743206024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,128,power_law_1.01,0.0315775990486145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,128,balanced,0.03143999973932902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,128,power_law_1.01,0.03323520123958588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,128,balanced,0.03349866718053818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,128,power_law_1.01,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,128,balanced,0.0354720006386439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,128,power_law_1.01,0.034745600819587705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,128,power_law_1.01,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,128,balanced,0.12168000141779582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,128,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,128,power_law_1.01,0.03436160087585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,128,balanced,0.04053333401679993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,128,power_law_1.01,0.035104000568389894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,128,power_law_1.01,0.03446399867534637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,128,balanced,0.046682665745417275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,128,power_law_1.01,0.03527039885520935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,128,power_law_1.01,0.031219199299812317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,128,balanced,0.14470400412877402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,128,balanced,0.05264533559481303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,128,power_law_1.01,0.03607040047645569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,128,power_law_1.01,0.03054719865322113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,128,power_law_1.01,0.03591679930686951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,128,power_law_1.01,0.03113600015640259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,128,power_law_1.01,0.03666560053825378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,128,balanced,0.0664106657107671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,128,power_law_1.01,0.03255040049552917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,128,power_law_1.01,0.038841599225997926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,128,power_law_1.01,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,128,power_law_1.01,0.03911679983139038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,128,power_law_1.01,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,128,power_law_1.01,0.0401856005191803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,128,balanced,0.08488532900810242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,128,power_law_1.01,0.03461120128631592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,128,power_law_1.01,0.041152000427246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,128,power_law_1.01,0.035283198952674864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,128,power_law_1.01,0.04181120097637177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,128,balanced,0.2051466703414917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,128,power_law_1.01,0.035071998834609985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,128,power_law_1.01,0.04467839896678925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,128,power_law_1.01,0.03223040103912354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,128,power_law_1.01,0.016415999829769136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,128,power_law_1.01,0.035129600763320924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,128,balanced,0.11404800415039062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,128,power_law_1.01,0.04793600142002106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,128,power_law_1.01,0.0322816014289856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,128,power_law_1.01,0.016172799468040466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,128,power_law_1.01,0.03594880104064942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,128,power_law_1.01,0.0543615996837616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,128,power_law_1.01,0.0320576012134552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,128,power_law_1.01,0.016652800142765045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,2,128,power_law_1.2,0.03193599879741669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,128,power_law_1.01,0.03649280071258545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,128,power_law_1.01,0.059494400024414064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,128,balanced,0.1431946655114492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,128,power_law_1.01,0.03386879861354828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,128,power_law_1.01,0.018592000007629395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,2,128,power_law_1.2,0.031353598833084105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,128,power_law_1.01,0.03681919872760773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,128,power_law_1.01,0.08131200075149536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,128,power_law_1.01,0.034534400701522826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,128,power_law_1.01,0.030713599920272828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,2,128,power_law_1.2,0.031353598833084105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,128,power_law_1.01,0.03758080005645752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,128,power_law_1.01,0.09709439873695373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,2,128,power_law_1.2,0.030585598945617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,128,power_law_1.01,0.03555200099945068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,128,power_law_1.01,0.024166400730609893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,2,128,power_law_1.2,0.033164799213409424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,128,balanced,0.36804266770680744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,128,power_law_1.01,0.04047360122203827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,128,balanced,0.1722453236579895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,128,power_law_1.01,0.12581759691238403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,2,128,power_law_1.2,0.030579200387001036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,128,power_law_1.01,0.03527039885520935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,128,power_law_1.01,0.024409599602222443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,2,128,power_law_1.2,0.03341439962387085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,128,power_law_1.01,0.03994239866733551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,128,power_law_1.01,0.16456320285797119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,2,128,power_law_1.2,0.03136639893054962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,128,power_law_1.01,0.0357120007276535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,128,power_law_1.01,0.02489600032567978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,2,128,power_law_1.2,0.035366401076316833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,128,power_law_1.01,0.04131200015544891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,128,power_law_1.01,0.2050879955291748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,2,128,power_law_1.2,0.03322240114212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,128,power_law_1.01,0.03531520068645477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,128,balanced,0.25860800345738727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,128,power_law_1.01,0.02457599937915802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,2,128,power_law_1.2,0.034534400701522826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,128,power_law_1.01,0.04366079866886139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,128,power_law_1.01,0.31746559143066405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,2,128,power_law_1.2,0.033478400111198424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,128,power_law_1.01,0.0357120007276535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,128,power_law_1.01,0.024864000082015992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,2,128,power_law_1.2,0.03430399894714355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,128,power_law_1.01,0.04599680006504059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,128,power_law_1.01,0.5551296234130859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,2,128,power_law_1.2,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,128,power_law_1.01,0.03632639944553375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,128,power_law_1.01,0.02536959946155548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,2,128,power_law_1.2,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,128,power_law_1.01,0.04940800070762634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,2,128,power_law_1.2,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,128,power_law_1.01,0.03664000034332275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,2,128,power_law_1.2,0.034815999865531924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,128,power_law_1.01,0.026745599508285523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,128,power_law_1.01,0.054016000032424925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,2,128,power_law_1.2,0.035071998834609985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,2,128,power_law_1.2,0.03608959913253784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,128,power_law_1.01,0.03758080005645752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,128,power_law_1.01,0.02622080147266388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,128,power_law_1.01,0.06348800063133239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,2,128,power_law_1.2,0.03516800105571747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,2,128,power_law_1.2,0.035897600650787356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,128,power_law_1.01,0.037529599666595456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,128,power_law_1.01,0.027692800760269164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,2,128,power_law_1.2,0.035123199224472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,128,power_law_1.01,0.07432320117950439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,2,128,power_law_1.2,0.036032000184059144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,128,power_law_1.01,0.04009599983692169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,128,power_law_1.01,0.028281599283218384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,2,128,power_law_1.2,0.03625600039958954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,2,128,power_law_1.2,0.037062400579452516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,128,power_law_1.01,0.10784000158309937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,128,power_law_1.01,0.040633600950241086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,128,balanced,0.4898826678593953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,128,power_law_1.01,0.028435200452804565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,2,128,power_law_1.2,0.036550399661064145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,2,128,power_law_1.2,0.03899520039558411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,128,power_law_1.01,0.14072959423065184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,128,power_law_1.01,0.04108799993991852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,128,power_law_1.01,0.030003198981285097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,2,128,power_law_1.2,0.036883199214935304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,2,128,power_law_1.2,0.03917439877986908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,128,power_law_1.01,0.18528640270233154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,128,power_law_1.01,0.04318079948425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,128,power_law_1.01,0.031718400120735166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,2,128,power_law_1.2,0.04103040099143982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,2,128,power_law_1.2,0.03752320110797882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,128,power_law_1.01,0.24800639152526854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,128,power_law_1.01,0.046060800552368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,2,128,power_law_1.2,0.042361599206924436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,128,power_law_1.01,0.03314560055732727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,2,128,power_law_1.2,0.04058879911899567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,128,power_law_1.01,0.2989248037338257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,2,128,power_law_1.2,0.04371840059757233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,128,power_law_1.01,0.050732797384262084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,128,power_law_1.01,0.03711360096931458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,2,128,power_law_1.2,0.04023039937019348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,128,power_law_1.01,0.45679359436035155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,2,128,power_law_1.2,0.04716159999370575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,128,power_law_1.01,0.05459840297698974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,128,power_law_1.01,0.038380798697471616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,2,128,power_law_1.2,0.040838399529457094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,2,128,power_law_1.2,0.05329279899597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,128,power_law_1.01,0.9726143836975097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,128,power_law_1.01,0.06357759833335877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,128,power_law_1.01,0.04456959962844849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,2,128,power_law_1.2,0.04442239999771118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,2,128,power_law_1.2,0.06123520135879516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,128,power_law_1.01,0.07208319902420043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,128,power_law_1.01,0.05055999755859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,2,128,power_law_1.2,0.04699519872665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,2,128,power_law_1.2,0.07438719868659974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,128,power_law_1.01,0.09937919974327088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,128,power_law_1.01,0.06421759724617004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,2,128,power_law_1.2,0.05330560207366943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,2,128,power_law_1.2,0.09783040285110474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,128,power_law_1.01,0.11112960577011108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,128,power_law_1.01,0.0818943977355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,2,128,power_law_1.2,0.12188800573348998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,2,128,power_law_1.2,0.06049280166625977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,128,power_law_1.01,0.15842560529708863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,128,power_law_1.01,0.1094208002090454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,2,128,power_law_1.2,0.1683135986328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,2,128,power_law_1.2,0.07272319793701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,128,power_law_1.01,0.20078721046447753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,128,power_law_1.01,0.13939199447631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,2,128,power_law_1.2,0.21285760402679443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,2,128,power_law_1.2,0.09095680117607116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,128,power_law_1.01,0.2385472059249878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,128,power_law_1.01,0.16386560201644898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,2,128,power_law_1.2,0.27425920963287354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,2,128,power_law_1.2,0.1263167977333069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,128,power_law_1.01,0.352454400062561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,2,128,power_law_1.2,0.4169792175292969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,128,power_law_1.01,0.2500096082687378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,2,128,power_law_1.2,0.16496000289916993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,128,power_law_1.01,0.6874815940856933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,2,128,power_law_1.2,0.7930880069732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,128,power_law_1.01,0.46211838722229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,2,128,power_law_1.2,0.23644800186157228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,2,128,power_law_1.2,0.3083008050918579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,2,128,power_law_1.2,0.3741055965423584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,2,128,power_law_1.2,0.5990272045135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,2,128,power_law_1.2,1.173036766052246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,2048,768,8,128,2,128,power_law_1.2,0.031276801228523256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,2048,768,8,128,2,128,power_law_1.2,0.031065601110458373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,2048,768,8,128,2,128,power_law_1.2,0.03116160035133362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,2048,768,8,128,2,128,power_law_1.2,0.03342719972133636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,2048,768,8,128,2,128,power_law_1.2,0.03298560082912445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,2048,768,8,128,2,128,power_law_1.2,0.0347135990858078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,2048,768,8,128,2,128,power_law_1.2,0.034668800234794614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,2048,768,8,128,2,128,power_law_1.2,0.0349375993013382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,2048,768,8,128,2,128,power_law_1.2,0.034822401404380796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,2048,768,8,128,2,128,power_law_1.2,0.03506560027599335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,2048,768,8,128,2,128,power_law_1.2,0.03581439852714539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,2048,768,8,128,2,128,power_law_1.2,0.03586559891700745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,2048,768,8,128,2,128,power_law_1.2,0.016524800658226015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,2048,768,8,128,2,128,power_law_1.2,0.037190398573875426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,2048,768,8,128,2,128,power_law_1.2,0.017183999717235564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,2048,768,8,128,2,128,power_law_1.2,0.03716480135917664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,2048,768,8,128,2,128,power_law_1.2,0.017907199263572694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,2048,768,8,128,2,128,power_law_1.2,0.039955198764801025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,2048,768,8,128,2,128,power_law_1.2,0.01880960017442703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,2048,768,8,128,2,128,power_law_1.2,0.039801600575447085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,2048,768,8,128,2,128,power_law_1.2,0.03073279857635498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,2048,768,8,128,2,128,power_law_1.2,0.041247999668121337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,2048,768,8,128,2,128,power_law_1.2,0.024883200228214265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,2048,768,8,128,2,128,power_law_1.2,0.044607999920845035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,2048,768,8,128,2,128,power_law_1.2,0.024345600605010988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,2048,768,8,128,2,128,power_law_1.2,0.048089599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,2048,768,8,128,2,128,power_law_1.2,0.02508159875869751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,2048,768,8,128,2,128,power_law_1.2,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,2048,768,8,128,2,128,power_law_1.2,0.024953599274158477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,2048,768,8,128,2,128,power_law_1.2,0.06149759888648987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,2048,768,8,128,2,128,power_law_1.2,0.025190401077270507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,2048,768,8,128,2,128,power_law_1.2,0.0720192015171051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,2048,768,8,128,2,128,power_law_1.2,0.02537600100040436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,2048,768,8,128,2,128,power_law_1.2,0.0863103985786438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,2048,768,8,128,2,128,power_law_1.2,0.026399999856948853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,2048,768,8,128,2,128,power_law_1.2,0.11460479497909545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,2048,768,8,128,2,128,power_law_1.2,0.02680320143699646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,2048,768,8,128,2,128,power_law_1.2,0.027641600370407103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,2048,768,8,128,2,128,power_law_1.2,0.13779200315475465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,2048,768,8,128,2,128,power_law_1.2,0.028352001309394838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,2048,768,8,128,2,128,power_law_1.2,0.1964352011680603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,2048,768,8,128,2,128,power_law_1.2,0.02879360020160675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,2048,768,8,128,2,128,power_law_1.2,0.0296640008687973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,2048,768,8,128,2,128,power_law_1.2,0.26229760646820066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,2048,768,8,128,2,128,power_law_1.2,0.03215999901294708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,2048,768,8,128,2,128,power_law_1.2,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,2048,768,8,128,2,128,power_law_1.2,0.3164031982421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,2048,768,8,128,2,128,power_law_1.2,0.03696640133857727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,2048,768,8,128,2,128,power_law_1.2,0.4824704170227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,2048,768,8,128,2,128,power_law_1.2,0.03896960020065308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,2048,768,8,128,2,128,power_law_1.2,0.044460800290107724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,2048,768,8,128,2,128,power_law_1.2,0.9468031883239746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,2048,768,8,128,2,128,power_law_1.2,0.05086719989776611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,2048,768,8,128,2,128,power_law_1.2,0.06403840184211732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,2048,768,8,128,2,128,power_law_1.2,0.08407679796218873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,2048,768,8,128,2,128,power_law_1.2,0.11182719469070435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,2048,768,8,128,2,128,power_law_1.2,0.1402624011039734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,2048,768,8,128,2,128,power_law_1.2,0.16952320337295532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,2048,768,8,128,2,128,power_law_1.2,0.25518720149993895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,2048,768,8,128,2,128,power_law_1.2,0.4875967979431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,64,balanced,0.023775999744733173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,64,balanced,0.025797332326571148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,64,balanced,0.025610665480295818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,64,balanced,0.025061334172884624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,64,balanced,0.025424001117547352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,64,balanced,0.027098665634791057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,64,balanced,0.027461332579453785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,64,balanced,0.027744000156720478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,64,balanced,0.027653334041436512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,64,balanced,0.027722666660944622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,64,balanced,0.027914665639400482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,64,balanced,0.027274665733178455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,64,balanced,0.027845333019892376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,64,balanced,0.03161599983771642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,64,balanced,0.033610666791598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,64,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,64,balanced,0.03336533407370249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,64,balanced,0.03499199946721395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,64,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,64,balanced,0.03578133384386698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,64,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,64,balanced,0.03939199944337209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,64,balanced,0.04164266586303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,64,balanced,0.04666133224964142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,64,balanced,0.050373335679372154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,64,balanced,0.03133333226044973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,64,balanced,0.030058667063713074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,64,balanced,0.029391999046007793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,64,balanced,0.02926933268706004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,64,balanced,0.06030400097370148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,64,balanced,0.03125333289305369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,64,balanced,0.03180266668399175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,64,balanced,0.03369600077470144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,64,balanced,0.03153600047032038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,64,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,64,balanced,0.03363200028737386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,64,balanced,0.03367999941110611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,64,balanced,0.07457600037256877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,64,balanced,0.03473066786924998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,64,balanced,0.034186666210492454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,64,balanced,0.035445332527160645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,64,balanced,0.03828266759713491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,64,balanced,0.03775466730197271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,64,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,64,balanced,0.03974399964014689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,64,balanced,0.04170133173465729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,64,balanced,0.08981866637865703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,64,balanced,0.041802664597829185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,64,power_law_1.2,0.025279998779296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,2048,768,8,128,4,64,power_law_1.01,0.023219199478626253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,64,balanced,0.04561600089073181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,64,power_law_1.2,0.02433920055627823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,2048,768,8,128,4,64,power_law_1.01,0.024025599658489227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,64,power_law_1.2,0.024716800451278685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,64,power_law_1.2,0.03132160007953644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,2048,768,8,128,4,64,power_law_1.01,0.02401919960975647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,2048,768,8,128,4,64,power_law_1.01,0.03080959916114807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,64,balanced,0.04950400193532308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,64,power_law_1.2,0.025183999538421632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,64,power_law_1.2,0.029139199852943422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,2048,768,8,128,4,64,power_law_1.01,0.024006399512290954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,2048,768,8,128,4,64,power_law_1.01,0.0286655992269516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,64,power_law_1.2,0.025459200143814087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,64,balanced,0.05202133456865946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,64,power_law_1.2,0.02908160090446472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,64,balanced,0.11176533500353496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,2048,768,8,128,4,64,power_law_1.01,0.024582399427890776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,2048,768,8,128,4,64,power_law_1.01,0.028799998760223388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,64,power_law_1.2,0.026310399174690247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,64,power_law_1.2,0.029260799288749695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,2048,768,8,128,4,64,power_law_1.01,0.025241601467132568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,64,balanced,0.06444799900054932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,2048,768,8,128,4,64,power_law_1.01,0.028384000062942505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,64,power_law_1.2,0.026579201221466064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,2048,768,8,128,4,64,power_law_1.01,0.025791999697685242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,64,power_law_1.2,0.030144000053405763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,2048,768,8,128,4,64,power_law_1.01,0.028921601176261903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,64,power_law_1.2,0.027091199159622194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,2048,768,8,128,4,64,power_law_1.01,0.026547199487686156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,64,balanced,0.08667733271916707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,64,power_law_1.2,0.03128319978713989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,64,power_law_1.2,0.029600000381469725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,2048,768,8,128,4,64,power_law_1.01,0.030374398827552794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,2048,768,8,128,4,64,power_law_1.01,0.0289792001247406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,64,power_law_1.2,0.03136639893054962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,64,power_law_1.2,0.03000960052013397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,2048,768,8,128,4,64,power_law_1.01,0.02961919903755188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,2048,768,8,128,4,64,power_law_1.01,0.030099201202392577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,64,power_law_1.2,0.031513598561286923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,64,power_law_1.2,0.033657601475715636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,2048,768,8,128,4,64,power_law_1.01,0.032128000259399415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,64,balanced,0.12179199854532878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,64,power_law_1.2,0.03566080033779144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,64,power_law_1.2,0.03207040131092072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,2048,768,8,128,4,64,power_law_1.01,0.030559998750686646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,2048,768,8,128,4,64,power_law_1.01,0.03475199937820435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,64,power_law_1.2,0.03619840145111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,64,power_law_1.2,0.032287999987602234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,64,balanced,0.17612266540527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,2048,768,8,128,4,64,power_law_1.01,0.035046398639678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,2048,768,8,128,4,64,power_law_1.01,0.03125759959220886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,64,power_law_1.2,0.030796799063682555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,64,balanced,0.15032000343004862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,64,power_law_1.2,0.03273600041866302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,2048,768,8,128,4,64,power_law_1.01,0.030643200874328612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,2048,768,8,128,4,64,power_law_1.01,0.031302401423454286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,64,power_law_1.2,0.034220799803733826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,2048,768,8,128,4,64,power_law_1.01,0.03308799862861633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,64,power_law_1.2,0.03309440016746521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,2048,768,8,128,4,64,power_law_1.01,0.032652801275253295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,64,power_law_1.2,0.03477759957313538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,2048,768,8,128,4,64,power_law_1.01,0.03297280073165894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,64,power_law_1.2,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,64,power_law_1.2,0.035020801424980166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,64,balanced,0.18315200010935465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,2048,768,8,128,4,64,power_law_1.01,0.03280639946460724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,2048,768,8,128,4,64,power_law_1.01,0.03472639918327332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,64,power_law_1.2,0.03606399893760681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,64,power_law_1.2,0.03870719969272614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,2048,768,8,128,4,64,power_law_1.01,0.03573760092258453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,2048,768,8,128,4,64,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,64,power_law_1.2,0.03903999924659729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,64,power_law_1.2,0.04017919898033142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,2048,768,8,128,4,64,power_law_1.01,0.038431999087333676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,2048,768,8,128,4,64,power_law_1.01,0.035180801153182985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,64,power_law_1.2,0.039238399267196654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,64,balanced,0.2593119939168294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,64,power_law_1.01,0.0441536009311676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,2048,768,8,128,4,64,power_law_1.2,0.04688000082969666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,2048,768,8,128,4,64,power_law_1.01,0.03816959857940674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,64,power_law_1.2,0.0411327987909317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,64,power_law_1.01,0.044377601146698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,2048,768,8,128,4,64,power_law_1.2,0.051577597856521606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,2048,768,8,128,4,64,power_law_1.01,0.038815999031066896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,64,power_law_1.2,0.04650880098342895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,64,power_law_1.01,0.052767997980117796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,2048,768,8,128,4,64,power_law_1.2,0.06023039817810059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,2048,768,8,128,4,64,power_law_1.01,0.04009599983692169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,64,power_law_1.2,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,64,power_law_1.01,0.05683839917182922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,2048,768,8,128,4,64,power_law_1.2,0.07310720086097718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,64,power_law_1.2,0.057036799192428586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,2048,768,8,128,4,64,power_law_1.01,0.042854401469230655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,64,power_law_1.01,0.06990079879760742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,2048,768,8,128,4,64,power_law_1.2,0.10193920135498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,64,power_law_1.2,0.05937920212745666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,2048,768,8,128,4,64,power_law_1.01,0.04451839923858643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,64,power_law_1.01,0.10119680166244507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,2048,768,8,128,4,64,power_law_1.2,0.10310399532318115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,64,power_law_1.2,0.07098879814147949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,2048,768,8,128,4,64,power_law_1.01,0.050835198163986205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,64,power_law_1.01,0.11907199621200562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,2048,768,8,128,4,64,power_law_1.2,0.18432639837265014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,64,power_law_1.2,0.08825600147247314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,2048,768,8,128,4,64,power_law_1.01,0.05310720205307007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,64,balanced,0.463589350382487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,64,power_law_1.01,0.16397440433502197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,2048,768,8,128,4,64,power_law_1.2,0.20851199626922606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,64,power_law_1.2,0.12874239683151245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,2048,768,8,128,4,64,power_law_1.01,0.06275839805603027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,64,power_law_1.01,0.16040960550308228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,2048,768,8,128,4,64,power_law_1.2,0.32410240173339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,64,power_law_1.2,0.16986240148544313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,2048,768,8,128,4,64,power_law_1.01,0.06998400092124939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,64,power_law_1.01,0.2569792032241821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,2048,768,8,128,4,64,power_law_1.2,0.4483391761779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,64,power_law_1.2,0.2751296043395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,2048,768,8,128,4,64,power_law_1.01,0.10890239477157593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,64,power_law_1.01,0.46874241828918456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,2048,768,8,128,4,64,power_law_1.2,1.0316160202026368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,64,power_law_1.2,0.35576961040496824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,2048,768,8,128,4,64,power_law_1.01,0.1279744029045105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,64,power_law_1.2,0.4348288059234619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,64,power_law_1.2,0.7199808120727539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,2048,768,8,128,4,64,power_law_1.01,0.18316160440444945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,64,power_law_1.2,1.685036849975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,2048,768,8,128,4,64,power_law_1.01,0.23574399948120117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,2048,768,8,128,4,64,power_law_1.01,0.30427520275115966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,2048,768,8,128,4,64,power_law_1.01,0.4459199905395508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,2048,768,8,128,4,64,power_law_1.01,0.9689536094665527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,1,balanced,0.05397333204746246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,1,balanced,0.054010664423306785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,1,balanced,0.07714133461316426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,1,balanced,0.02926933268706004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,1,balanced,0.07293866574764252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,1,power_law_1.01,0.053574401140213015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,1,power_law_1.01,0.07429119944572449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,1,balanced,0.07663999994595845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,1,power_law_1.01,0.0532800018787384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,1,power_law_1.01,0.02715519964694977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,1,balanced,0.1186186671257019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,1,balanced,0.037045332292715706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,1,balanced,0.10124267141024272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,1,power_law_1.01,0.059059202671051025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,1,balanced,0.1202826698621114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,1,power_law_1.01,0.06232320070266724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,1,power_law_1.01,0.08394880294799804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,1,balanced,0.051130667328834534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,1,balanced,0.20228266716003418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,1,power_law_1.01,0.029542401432991028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,1,balanced,0.16237866878509521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,1,power_law_1.01,0.10488320589065551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,1,balanced,0.20719999074935913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,1,power_law_1.01,0.08626559972763062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,1,balanced,0.07295466462771098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,1,power_law_1.01,0.16187520027160646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,1,balanced,0.37270931402842206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,1,power_law_1.01,0.04891520142555237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,1,balanced,0.27190933624903363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,1,power_law_1.01,0.14426239728927612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,1,balanced,0.37546666463216144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,1,power_law_1.01,0.12043520212173461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,1,balanced,0.12020267049471538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,1,power_law_1.01,0.23690879344940186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,1,balanced,0.7062239646911621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,1,power_law_1.01,0.06991360187530518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,1,balanced,0.2767573396364848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,1,balanced,0.3774346510569255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,1,power_law_1.01,0.18814719915390016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,1,power_law_1.01,0.15527679920196533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,1,balanced,0.7147946357727051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,1,balanced,0.12504000465075174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,1,power_law_1.01,0.31994240283966063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,1,power_law_1.01,0.10212479829788208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,1,power_law_1.01,0.23744640350341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,1,balanced,0.38285334904988605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,1,balanced,0.28014399607976276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,1,power_law_1.01,0.3147455930709839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,1,balanced,0.721232016881307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,1,power_law_1.01,0.16047999858856202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,1,balanced,0.13004799683888754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,1,power_law_1.01,0.5813632011413574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,1,power_law_1.01,0.2474560022354126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,1,balanced,0.3843253453572591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,1,balanced,0.2805440028508504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,1,power_law_1.01,0.17222399711608888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,1,power_law_1.01,0.3310015916824341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,1,power_law_1.01,0.618611192703247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,1,balanced,0.7204319636027018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,1,balanced,0.13620799779891968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,1,power_law_1.01,0.25243520736694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,1,balanced,0.3866080045700073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,1,power_law_1.01,0.18021759986877442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,1,balanced,0.2812266747156779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,1,power_law_1.01,0.34939520359039306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,1,power_law_1.01,0.6432767868041992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,1,balanced,0.14230400323867798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,1,balanced,0.7245759963989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,1,power_law_1.01,0.2591423988342285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,1,balanced,0.39158932367960614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,1,power_law_1.01,0.17742079496383667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,1,balanced,0.28386666377385456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,1,power_law_1.01,0.3627583980560303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,1,balanced,0.14909866452217102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,1,balanced,0.7269333203633627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,1,power_law_1.01,0.66561918258667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,1,power_law_1.01,0.2562560081481934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,1,balanced,0.3929706811904907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,1,power_law_1.01,0.1909824013710022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,1,balanced,0.2857973376909892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,1,power_law_1.01,0.3591104030609131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,1,balanced,0.7300639947255453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,1,balanced,0.1621280014514923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,1,power_law_1.01,0.6517312049865722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,1,power_law_1.01,0.2692608118057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,1,balanced,0.39811734358469647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,1,power_law_1.01,0.1960960030555725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,1,balanced,0.28970666726430255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,1,power_law_1.01,0.37036800384521484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,1,balanced,0.7329440116882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,1,balanced,0.14784533778826395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,1,power_law_1.01,0.6927680015563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,1,power_law_1.01,0.27981441020965575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,1,balanced,0.4031840165456136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,1,balanced,0.2928160031636556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,1,power_law_1.01,0.19633280038833617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,1,power_law_1.01,0.383078408241272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,1,balanced,0.7396319707234701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,1,balanced,0.15244266390800476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,1,power_law_1.01,0.6996479988098144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,1,power_law_1.01,0.2864192008972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,1,balanced,0.40879468123118085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,1,power_law_1.01,0.20504319667816162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,1,balanced,0.29948800802230835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,1,balanced,0.7442986965179443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,1,power_law_1.01,0.3869055986404419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,1,balanced,0.16473600268363953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,1,power_law_1.01,0.7037759780883789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,1,power_law_1.01,0.296287989616394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,1,balanced,0.4194186528523763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,1,power_law_1.01,0.22099840641021729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,1,balanced,0.3107946713765462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,1,balanced,0.7569279670715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,1,power_law_1.01,0.4073919773101807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,1,balanced,0.1586079994837443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,1,power_law_1.01,0.7310976028442383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,1,power_law_1.01,0.308953595161438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,1,power_law_1.01,0.208351993560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,1,balanced,0.4257599910100301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,1,balanced,0.31623999277750653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,1,balanced,0.7644000053405762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,1,power_law_1.01,0.42654080390930177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,1,power_law_1.01,0.7546495914459228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,1,balanced,0.16039466857910156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,1,power_law_1.01,0.3235840082168579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,1,power_law_1.01,0.2059328079223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,1,balanced,0.443557341893514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,1,balanced,0.3337013324101766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,1,balanced,0.7781546910603842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,1,power_law_1.01,0.43708162307739257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,1,power_law_1.01,0.7646016120910645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,1,balanced,0.17597333590189615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,1,power_law_1.01,0.34495360851287843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,1,power_law_1.01,0.22704639434814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,1,balanced,0.4729439814885457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,1,balanced,0.3705013195673625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,1,balanced,0.8069120248158773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,1,power_law_1.01,0.44464640617370604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,1,power_law_1.01,0.802086353302002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,1,balanced,0.1896373430887858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,1,power_law_1.01,0.398144006729126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,1,power_law_1.01,0.24268159866333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,1,balanced,0.5193920135498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,1,balanced,0.41182398796081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,1,balanced,0.8563253084818522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,1,power_law_1.01,0.49111042022705076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,1,power_law_1.01,0.8636863708496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,1,balanced,0.1975946625073751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.01,0.2614720106124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.01,0.4331647872924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,1,balanced,0.5678720076878866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,1,balanced,0.4687146743138631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,1,balanced,0.8968426386515299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,1,power_law_1.01,0.518720006942749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,1,power_law_1.01,0.9254528045654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,1,balanced,0.4159839948018392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.01,0.6072383880615234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.01,0.5161407947540283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,1,balanced,0.5298719803492228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,1,balanced,0.6305973529815674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,1,power_law_1.01,0.5992256164550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,1,power_law_1.01,1.0603967666625977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,1,balanced,0.9595413208007812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.01,0.686297607421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.01,0.6042751789093017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,1,power_law_1.01,0.6897088050842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,1,balanced,0.47766931851704914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,1,balanced,0.7513653437296549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,1,balanced,0.8200053373972574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,1,power_law_1.01,1.061734390258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,1,balanced,1.1766453584035237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.01,0.7896959781646729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.01,0.4274432182312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,1,power_law_1.01,0.8664383888244629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,1,power_law_1.01,1.2862591743469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,1,balanced,0.3365066846211751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.01,0.9665408134460449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,1,balanced,0.8798933029174805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,1,balanced,0.90829865137736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.01,0.46782717704772947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,1,balanced,1.2090293566385906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,1,power_law_1.01,1.0402048110961915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.01,1.3356287956237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,1,power_law_1.01,1.4701312065124512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,1,balanced,0.3447626829147339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.01,0.601804780960083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,1,balanced,1.2585546970367432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,1,balanced,1.315392017364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,1,balanced,1.6953546206156414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,1,power_law_1.01,1.3980352401733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.01,1.672287940979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,1,power_law_1.01,1.8557184219360352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.01,0.734611177444458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,1,power_law_1.01,1.7670911788940429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.01,2.359756851196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,1,power_law_1.01,2.2476287841796876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,1,balanced,0.6168746550877889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,1,balanced,1.5567253430684407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,1,balanced,1.5595253308614094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.01,1.0018495559692382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,1,balanced,1.8463199933369954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,1,power_law_1.01,2.447795104980469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.01,3.003936004638672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,1,power_law_1.01,3.033830451965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.01,1.267097568511963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,1,balanced,0.6065226793289185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,1,power_law_1.01,3.120505523681641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.01,3.697465515136719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,1,balanced,2.600378672281901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,1,balanced,2.258159955342611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,1,balanced,2.2333332697550454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,1,power_law_1.01,4.057299041748047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.01,1.5336959838867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,1,power_law_1.01,3.8153217315673826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,1,power_law_1.01,5.727571105957031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,1,balanced,0.8698186874389648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,1,power_law_1.01,4.843353652954102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,1,balanced,3.4055894215901694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,1,power_law_1.01,2.3391935348510744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,1,power_law_1.01,5.838636779785157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,1,balanced,2.942000071207682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,1,power_law_1.01,11.222342681884765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,1,balanced,2.9573065439860025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,1,power_law_1.01,7.175609588623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,1,power_law_1.01,4.508012771606445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,1,balanced,1.1389599641164143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,1,power_law_1.01,11.289830780029297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,1,power_law_1.01,13.562835693359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,1,balanced,4.524122556050618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,1,balanced,3.6237386067708335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,1,balanced,3.6265385945638022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,1,balanced,1.4117973645528157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,1,balanced,6.546474456787109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,1,balanced,5.679503758748372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,1,balanced,5.591087977091472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,1,balanced,2.232463995615641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,1,balanced,13.020304361979166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,1,balanced,11.11301294962565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,1,balanced,11.019930521647135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,1,balanced,4.421989440917969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,1,power_law_1.2,0.027558401226997375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,1,power_law_1.2,0.07402880191802978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,1,power_law_1.2,0.05203199982643127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,1,power_law_1.2,0.05335680246353149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,1,power_law_1.2,0.03232640027999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,1,power_law_1.2,0.08257279992103576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,1,power_law_1.2,0.05711359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,1,power_law_1.2,0.0597760021686554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,1,power_law_1.2,0.1526528000831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,1,power_law_1.2,0.04891520142555237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,1,power_law_1.2,0.22079999446868898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,1,power_law_1.2,0.08122239708900451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,1,power_law_1.2,0.08972799777984619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,1,power_law_1.2,0.2908416032791138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,1,power_law_1.2,0.06714239716529846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,1,power_law_1.2,0.11042560338973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,1,power_law_1.2,0.13480960130691527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,1,power_law_1.2,0.09286400079727172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,1,power_law_1.2,0.5872960090637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,1,power_law_1.2,0.1392575979232788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,1,power_law_1.2,0.17205120325088502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,1,power_law_1.2,0.6388800144195557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,1,power_law_1.2,0.15283199548721313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,1,power_law_1.2,0.23651840686798095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,1,power_law_1.2,0.3187263965606689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,1,power_law_1.2,0.6421567916870117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,1,power_law_1.2,0.25363199710845946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,1,power_law_1.2,0.1777151942253113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,1,power_law_1.2,0.3314496040344238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,1,power_law_1.2,0.6713920116424561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,1,power_law_1.2,0.25557119846343995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,1,power_law_1.2,0.17668479681015015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,1,power_law_1.2,0.34829440116882326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,1,power_law_1.2,0.663424015045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,1,power_law_1.2,0.26451199054718016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,1,power_law_1.2,0.1856511950492859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,1,power_law_1.2,0.2668031930923462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,1,power_law_1.2,0.7000639915466309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,1,power_law_1.2,0.3646784067153931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,1,power_law_1.2,0.19185919761657716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,1,power_law_1.2,0.2741375923156738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,1,power_law_1.2,0.2826560020446777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,1,power_law_1.2,0.20654079914093018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,1,power_law_1.2,0.7035967826843261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,1,power_law_1.2,0.3678719997406006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,1,power_law_1.2,0.2926784038543701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,1,power_law_1.2,0.19635839462280275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,1,power_law_1.2,0.38287360668182374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,1,power_law_1.2,0.7152575969696044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,1,power_law_1.2,0.2954495906829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,1,power_law_1.2,0.20527999401092528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,1,power_law_1.2,0.38419198989868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,1,power_law_1.2,0.7342336177825928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,1,power_law_1.2,0.3170752048492432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,1,power_law_1.2,0.22470400333404542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,1,power_law_1.2,0.3895807981491089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,1,power_law_1.2,0.7631167888641357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,1,power_law_1.2,0.32812159061431884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,1,power_law_1.2,0.21441919803619386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,1,power_law_1.2,0.7907008171081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,1,power_law_1.2,0.40789117813110354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,1,power_law_1.2,0.34823040962219237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,1,power_law_1.2,0.8175616264343262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,1,power_law_1.2,0.21616640090942382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,1,power_law_1.2,0.42337918281555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,1,power_law_1.2,0.4067967891693115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,1,power_law_1.2,0.8895808219909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,1,power_law_1.2,0.233734393119812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,1,power_law_1.2,0.4406720161437988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.2,0.44323201179504396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,1,power_law_1.2,0.9580991744995118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,1,power_law_1.2,0.24145278930664063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,1,power_law_1.2,0.4553535938262939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.2,0.5315392017364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,1,power_law_1.2,1.0787327766418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.2,0.27591040134429934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,1,power_law_1.2,0.505728006362915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.2,0.6250944137573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,1,power_law_1.2,0.5259136199951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.2,0.795795202255249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,1,power_law_1.2,1.0930944442749024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.2,0.6199039936065673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,1,power_law_1.2,0.6195263862609863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.2,0.9775232315063477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,1,power_law_1.2,1.3245823860168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.2,0.7315711975097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,1,power_law_1.2,0.7057472229003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.2,1.3422464370727538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.2,0.4587520122528076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,1,power_law_1.2,1.5224575996398926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,1,power_law_1.2,0.8927103996276855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.2,0.49411840438842775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.2,1.6834688186645508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,1,power_law_1.2,1.888185691833496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,1,power_law_1.2,1.0608896255493163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.2,0.6427711963653564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.2,2.361459159851074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,1,power_law_1.2,2.3051904678344726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.2,0.7539775848388672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,1,power_law_1.2,1.4217535972595214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.2,3.0137983322143556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,1,balanced,0.15267200271288553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,1,balanced,0.09731733798980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,1,balanced,0.15220266580581665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,1,power_law_1.2,3.0601919174194334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.2,1.0416640281677245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.2,3.702329635620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,1,power_law_1.2,1.7861248016357423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,1,balanced,0.23113600413004556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,1,balanced,0.06276800235112508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,1,balanced,0.15128533045450845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,1,balanced,0.25654399394989014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,1,power_law_1.2,4.111718368530274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.2,1.3043071746826171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,1,balanced,0.3662026723225911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,1,power_law_1.2,5.747449493408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,1,power_law_1.2,2.4654720306396483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,1,balanced,0.2547786633173625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,1,balanced,0.4640586773554484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,1,balanced,0.06818133095900218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,1,power_law_1.2,4.885753631591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,1,balanced,0.582479993502299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.2,1.5570688247680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,1,power_law_1.2,3.1556928634643553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,1,balanced,0.4715520143508911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,1,balanced,0.888480027516683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,1,power_law_1.2,11.294080352783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,1,balanced,0.0981066624323527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,1,power_law_1.2,7.253874969482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,1,balanced,1.0439733664194744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,1,power_law_1.2,3.8367359161376955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,1,balanced,0.8932212988535563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,1,power_law_1.2,2.3696640014648436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,1,balanced,1.7198452949523926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,1,balanced,2.139082590738932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,1,balanced,1.104032039642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,1,balanced,0.152837336063385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,1,balanced,1.2538080215454102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,1,balanced,2.142965316772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,1,balanced,1.1093706289927165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,1,power_law_1.2,5.826547241210937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,1,balanced,0.26506133874257404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,1,power_law_1.2,4.538124847412109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,1,balanced,1.2072106997172039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,1,balanced,2.163637320200602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,1,power_law_1.2,13.683929443359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,1,balanced,1.1140159765879314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,1,balanced,0.3270080089569092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,1,balanced,2.1734186808268228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,1,balanced,1.198367993036906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,1,balanced,0.3362079858779907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,1,balanced,1.1232853730519612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,1,balanced,2.177663962046305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,1,balanced,1.2108960151672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,1,balanced,0.3606719970703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,1,balanced,1.124186674753825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,1,power_law_1.2,11.222444915771485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,1,balanced,2.1812052726745605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,1,balanced,1.2000373204549153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,1,balanced,0.37354131539662677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,1,balanced,1.1342666943868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,1,balanced,2.1938346227010093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,1,balanced,1.1767360369364421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,1,balanced,0.3895039955774943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,1,balanced,1.1395413080851238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,1,balanced,1.2321759859720867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,1,balanced,0.38406399885813397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,1,balanced,2.193994681040446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,1,balanced,1.1429119904836018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,1,balanced,1.2289706865946453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,1,balanced,0.4029386838277181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,1,balanced,1.1555840174357097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,1,balanced,2.224682648976644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,1,balanced,0.4241120020548503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,1,balanced,1.304848035176595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,1,balanced,0.4045706590016683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,1,balanced,2.2375094095865884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,1,balanced,1.172709306081136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,1,balanced,1.2211573123931885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,1,balanced,0.42656532923380536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,1,balanced,2.238661289215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,1,balanced,1.2645973364512126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,1,balanced,1.185157299041748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,1,balanced,0.4506346782048543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,1,balanced,2.296170711517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,1,balanced,1.289562702178955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,1,balanced,1.2153759797414143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,1,balanced,0.4800693194071452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,1,balanced,2.3247413635253906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,1,balanced,1.2966612974802654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,1,balanced,1.26201597849528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,1,balanced,0.6326719919840494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,1,balanced,2.3859359423319497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,1,balanced,1.2941280206044514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,1,balanced,1.2922399838765461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,1,balanced,0.586847980817159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,1,balanced,0.7799253463745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,1,balanced,1.308677355448405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,1,balanced,1.5907786687215169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,1,balanced,3.1162986755371094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,1,balanced,0.6847679615020752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,1,balanced,1.373754660288493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,1,balanced,1.4785866737365723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,1,balanced,2.5860160191853843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,1,balanced,1.0379892985026042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,1,balanced,2.4480640093485513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,1,balanced,1.8681920369466145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,1,balanced,4.0353654225667315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,1,balanced,0.8499306837717692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,1,balanced,2.445184071858724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,1,balanced,1.865114688873291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,1,balanced,3.0280799865722656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,1,balanced,2.9133599599202475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,1,balanced,1.5208800633748372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,1,balanced,3.508965492248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,1,balanced,5.103071848551433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,1,balanced,1.5994186401367188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,1,balanced,3.041285196940104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,1,balanced,4.797279993693034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,1,balanced,4.688005447387695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,1,balanced,2.3805813789367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,1,balanced,5.2773386637369795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,1,balanced,8.637936274210611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,1,balanced,5.479178746541341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,1,balanced,3.0336907704671225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,1,balanced,5.612000147501628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,1,balanced,9.0981814066569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,1,balanced,7.7934614817301435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,1,balanced,3.108565330505371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,1,balanced,6.131568272908528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,1,balanced,9.668906529744467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,1,balanced,8.59553082784017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,1,balanced,5.568682352701823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,1,balanced,10.753349304199219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,1,power_law_1.01,0.14761600494384766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,1,balanced,17.034372965494793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,1,balanced,13.961072285970053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,1,power_law_1.01,0.23224959373474122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,1,power_law_1.01,0.3359296083450317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,1,power_law_1.01,0.6501376152038574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,1,power_law_1.01,0.9398207664489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,1,power_law_1.01,1.6123584747314452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,1,power_law_1.01,1.6955904006958007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,1,power_law_1.01,1.8546367645263673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,1,power_law_1.01,1.8618112564086915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,1,power_law_1.01,1.8821184158325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,1,power_law_1.01,0.09507200121879578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,1,power_law_1.01,1.9180543899536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,1,power_law_1.01,2.0265024185180662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,1,power_law_1.01,2.0991167068481444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,1,power_law_1.01,0.13822720050811768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,1,power_law_1.01,0.14731520414352417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,1,power_law_1.01,2.1000320434570314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,1,power_law_1.01,0.19562239646911622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,1,power_law_1.01,0.05100799798965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,1,power_law_1.01,2.2179967880249025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,1,power_law_1.01,0.1950976014137268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,1,power_law_1.01,0.32124159336090086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,1,power_law_1.01,2.2476032257080076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,1,power_law_1.01,0.4685376167297363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,1,power_law_1.01,0.28319358825683594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,1,power_law_1.01,0.07184640169143677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,1,power_law_1.01,0.44585599899291994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,1,power_law_1.01,0.8118016242980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,1,power_law_1.01,0.08572800159454345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,1,power_law_1.01,2.2724863052368165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,1,power_law_1.01,0.5837887763977051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,1,power_law_1.01,0.15676800012588502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,1,power_law_1.01,0.8887359619140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,1,power_law_1.01,2.4758655548095705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,1,power_law_1.01,0.22941439151763915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,1,power_law_1.01,0.9921536445617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,1,power_law_1.01,0.930720043182373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,1,power_law_1.01,2.7263296127319334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,1,power_law_1.01,0.3738176107406616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,1,power_law_1.01,0.9956735610961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,1,power_law_1.01,0.9195520401000976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,1,power_law_1.01,0.46161279678344724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,1,power_law_1.01,3.0902271270751953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,1,power_law_1.01,1.0576895713806151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,1,power_law_1.01,0.45009279251098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,1,power_law_1.01,0.9874815940856934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,1,power_law_1.01,3.2192768096923827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,1,power_law_1.01,1.0743488311767577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,1,power_law_1.01,0.47980799674987795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,1,power_law_1.01,1.0033599853515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,1,power_law_1.01,1.0950528144836427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,1,power_law_1.01,0.5051775932312011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,1,power_law_1.01,4.018067169189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,1,power_law_1.01,1.1371392250061034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,1,power_law_1.01,0.5238656044006348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,1,power_law_1.01,1.0405311584472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,1,power_law_1.01,4.031846237182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,1,power_law_1.01,0.5458303928375244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,1,power_law_1.01,1.1643136024475098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,1,power_law_1.01,1.0504063606262206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,1,power_law_1.01,0.5684991836547851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,1,power_law_1.01,4.898624038696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,1,power_law_1.01,0.537062406539917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,1,power_law_1.01,1.0959232330322266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,1,power_law_1.01,1.1750911712646483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,1,power_law_1.01,0.5547455787658692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,1,power_law_1.01,6.8042045593261715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,1,power_law_1.01,1.1090047836303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,1,power_law_1.01,1.182534408569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,1,power_law_1.01,0.5838016033172607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,1,power_law_1.01,1.154195213317871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,1,power_law_1.01,1.211948776245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,1,power_law_1.01,0.6649280071258545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,1,power_law_1.01,8.641932678222656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,1,power_law_1.01,1.1872447967529296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,1,power_law_1.01,0.7759424209594726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,1,power_law_1.01,1.2256704330444337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.01,0.7819136142730713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,1,power_law_1.01,1.2913472175598144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,1,power_law_1.01,10.310848236083984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,1,balanced,9.871146519978842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,1,power_law_1.01,1.2587200164794923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.01,0.9653440475463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,1,balanced,19.645397186279297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,1,power_law_1.01,1.396345615386963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,1,balanced,31.474703470865887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,1,balanced,28.180320739746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,1,power_law_1.01,12.026457977294921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,1,power_law_1.01,1.407027244567871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.01,0.874233627319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,1,power_law_1.01,1.5805888175964355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.01,1.4633472442626954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,1,power_law_1.01,17.080953979492186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.01,1.1285632133483887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,1,power_law_1.01,1.6091583251953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.01,1.6755647659301758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.01,1.2469120025634766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,1,power_law_1.01,1.9433151245117188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.01,1.9236160278320313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.01,1.4791872024536132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,1,power_law_1.01,2.290572738647461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,1,power_law_1.01,31.601150512695312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.01,2.34335994720459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.01,1.7407487869262694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,1,power_law_1.01,2.853875160217285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.01,2.7647552490234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.01,2.356243133544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,1,power_law_1.01,4.05296630859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.01,3.587763214111328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.01,2.9410688400268556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,1,power_law_1.01,5.186150360107422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,1,power_law_1.01,3.5406463623046873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.01,4.446105575561523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,1,power_law_1.01,6.140351867675781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.01,6.113484954833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,1,power_law_1.01,5.339091110229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,1,power_law_1.01,7.261593627929687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.01,7.763308715820313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,1,power_law_1.01,10.112262725830078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,1,power_law_1.01,10.7010498046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,1,power_law_1.01,9.410054779052734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,1,power_law_1.01,19.92003173828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,1,power_law_1.01,14.461984252929687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,1,power_law_1.01,28.121600341796874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,1,power_law_1.2,0.14970240592956544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,1,power_law_1.2,0.16149120330810546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,1,power_law_1.2,0.3359296083450317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,1,power_law_1.2,0.566534423828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,1,power_law_1.2,0.7515647888183594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,1,power_law_1.2,1.5699135780334472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,1,power_law_1.2,1.6733503341674805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,1,power_law_1.2,1.7954240798950196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,1,power_law_1.2,1.9096511840820312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,1,power_law_1.2,1.8947456359863282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,1,power_law_1.2,1.9901887893676757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,1,power_law_1.2,0.05088000297546387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,1,power_law_1.2,0.09806720018386841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,1,power_law_1.2,2.0907583236694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,1,power_law_1.2,0.14931199550628663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,1,power_law_1.2,2.1211776733398438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,1,power_law_1.2,0.05192959904670715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,1,power_law_1.2,0.10835839509963989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,1,power_law_1.2,0.17103999853134155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,1,power_law_1.2,2.1950143814086913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,1,balanced,0.14520000418027243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,1,power_law_1.2,0.08440319895744323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,1,power_law_1.2,0.17544959783554076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,1,power_law_1.2,0.2678719997406006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,1,power_law_1.2,2.303340721130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,1,power_law_1.2,0.2846208095550537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,1,power_law_1.2,0.14513920545578002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,1,power_law_1.2,0.3964927911758423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,1,balanced,0.2439840038617452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,1,power_law_1.2,2.325433540344238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,1,power_law_1.2,0.42687358856201174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,1,power_law_1.2,0.2144320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,1,power_law_1.2,0.5182464122772217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,1,balanced,0.4384640057881673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,1,power_law_1.2,0.8169471740722656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,1,power_law_1.2,0.40216960906982424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,1,power_law_1.2,2.382387161254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,1,balanced,0.8391679922739664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,1,power_law_1.2,0.9658368110656739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,1,power_law_1.2,0.9058560371398926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,1,power_law_1.2,0.44444799423217773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,1,power_law_1.2,2.598784065246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,1,power_law_1.2,1.0276415824890137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,1,balanced,1.6338987350463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,1,power_law_1.2,0.4670400142669678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,1,power_law_1.2,0.953593635559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,1,power_law_1.2,1.130502414703369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,1,power_law_1.2,0.49010558128356935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,1,balanced,3.1762612660725913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,1,power_law_1.2,2.8022464752197265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,1,power_law_1.2,0.9631808280944825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,1,power_law_1.2,0.5002880096435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,1,power_law_1.2,1.1492735862731933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,1,balanced,4.742207845052083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,1,power_law_1.2,1.0456512451171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,1,power_law_1.2,3.2352321624755858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,1,power_law_1.2,0.5229887962341309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,1,power_law_1.2,1.151699161529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,1,power_law_1.2,1.059059238433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,1,balanced,4.741269429524739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,1,power_law_1.2,0.5366911888122559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,1,power_law_1.2,3.383910369873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,1,power_law_1.2,1.1715200424194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,1,power_law_1.2,1.0523136138916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,1,power_law_1.2,0.5771327972412109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,1,balanced,4.745759963989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,1,power_law_1.2,1.2149184226989747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,1,power_law_1.2,4.139577484130859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,1,power_law_1.2,1.0821824073791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,1,power_law_1.2,0.5653376102447509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,1,balanced,4.752570788065593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,1,balanced,0.0969599982102712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,1,power_law_1.2,1.212947177886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,1,power_law_1.2,1.1348544120788575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,1,power_law_1.2,0.597708797454834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,1,power_law_1.2,4.148409652709961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,1,balanced,4.764911969502767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,1,power_law_1.2,1.2549887657165528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,1,power_law_1.2,1.1594623565673827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,1,balanced,0.11364266276359558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,1,power_law_1.2,0.6250879764556885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,1,balanced,0.14708800117174783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,1,power_law_1.2,1.2382399559020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,1,power_law_1.2,0.6729407787322998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,1,power_law_1.2,1.203212833404541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,1,power_law_1.2,5.001516723632813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,1,balanced,4.7831465403238935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,1,balanced,0.24425599972407022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,1,power_law_1.2,0.7964096069335938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,1,balanced,0.14593066771825156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,1,power_law_1.2,1.2502079963684083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,1,power_law_1.2,1.228223991394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,1,balanced,0.44724265734354657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,1,balanced,4.795088132222493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.2,0.7992576122283935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,1,power_law_1.2,1.3049023628234864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,1,balanced,0.2032853364944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,1,power_law_1.2,1.325817584991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,1,power_law_1.2,6.989644622802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,1,balanced,0.8480587005615234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.2,1.0043007850646972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,1,power_law_1.2,1.4185152053833008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,1,balanced,4.821231842041016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,1,balanced,0.32894400755564374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,1,power_law_1.2,1.4469823837280273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,1,power_law_1.2,8.740863800048828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,1,power_law_1.01,0.14545279741287231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,1,balanced,1.6274666786193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.2,0.9182527542114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,1,balanced,0.5753920078277588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,1,power_law_1.2,1.6481216430664063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.2,1.5446271896362305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,1,balanced,4.85315736134847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,1,balanced,2.4176905949910483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,1,balanced,1.0596960385640461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.2,1.163987159729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,1,power_law_1.2,1.6499967575073242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,1,power_law_1.2,10.492147064208984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,1,power_law_1.01,0.2407615900039673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.2,1.7414976119995118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,1,balanced,1.5537385940551758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,1,balanced,2.440336068471273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.2,1.2447487831115722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,1,power_law_1.2,1.9619199752807617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,1,balanced,4.8700745900472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,1,balanced,1.5536905924479167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,1,power_law_1.01,0.43474559783935546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,1,balanced,2.4449706077575684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.2,1.9587839126586915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.2,1.557036781311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,1,balanced,1.5594293276468914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,1,power_law_1.2,2.3205312728881835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,1,balanced,2.45194673538208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,1,power_law_1.2,12.266649627685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,1,power_law_1.01,0.5596223831176758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,1,balanced,4.924085299173991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,1,balanced,1.5662506421407063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.2,1.8195711135864259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.2,2.361952018737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,1,balanced,2.43668794631958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,1,power_law_1.2,2.917555236816406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,1,power_law_1.01,1.1551103591918945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,1,balanced,1.5729974110921223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,1,balanced,2.448410669962565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.2,2.4250240325927734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.2,2.797702407836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,1,balanced,5.008858680725098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,1,power_law_1.2,4.11902084350586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,1,balanced,1.577162742614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,1,power_law_1.01,1.7342912673950195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,1,balanced,2.458245277404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,1,power_law_1.2,17.09285125732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.2,2.9981760025024413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,1,balanced,1.5843413670857747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.2,3.5331520080566405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,1,balanced,2.473072052001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,1,power_law_1.2,5.283065414428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,1,power_law_1.01,2.0905344009399416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,1,balanced,1.6110453605651855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,1,power_law_1.2,3.5833152770996093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,1,balanced,5.0710452397664385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,1,balanced,2.4868319829305015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,1,balanced,1.6170825958251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,1,power_law_1.01,2.884556770324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.2,4.402508926391602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,1,power_law_1.2,6.186003112792969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,1,balanced,2.507925351460775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,1,balanced,1.6294827461242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,1,power_law_1.2,5.401721572875976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,1,power_law_1.01,3.0395328521728517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.2,6.011763381958008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,1,power_law_1.2,31.966574096679686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,1,balanced,1.6554346084594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,1,power_law_1.2,7.368563079833985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,1,balanced,2.530442714691162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,1,balanced,5.251909255981445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,1,power_law_1.01,3.0427200317382814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,1,balanced,1.7007039388020833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,1,balanced,2.586730639139811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.2,7.646578979492188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,1,power_law_1.2,10.111161804199218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,1,power_law_1.01,3.279993438720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,1,power_law_1.2,10.784409332275391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,1,balanced,5.3646189371744795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,1,balanced,1.736405372619629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,1,power_law_1.01,3.4281280517578123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,1,balanced,2.644517262776693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,1,power_law_1.2,9.232383728027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,1,power_law_1.01,3.533427047729492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,1,balanced,1.8220586776733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,1,balanced,2.765077273050944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,1,power_law_1.2,20.078802490234374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,1,balanced,5.682970682779948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,1,power_law_1.01,3.6632511138916017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,1,power_law_1.2,14.165338134765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,1,balanced,1.9209920565287273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,1,balanced,2.8567466735839844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,1,power_law_1.01,3.842937469482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,1,power_law_1.01,3.8358081817626952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,1,power_law_1.2,27.6611083984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,1,balanced,2.142159938812256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,1,balanced,3.0733973185221353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,1,balanced,5.733466466267903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,1,power_law_1.01,4.010176086425782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,1,power_law_1.01,4.316793441772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,1,balanced,3.28385066986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,1,balanced,2.3330507278442383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,1,power_law_1.01,4.49359359741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,1,power_law_1.01,4.955481719970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,1,balanced,6.197914759318034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,1,power_law_1.01,5.331283187866211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,1,balanced,3.716069221496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,1,power_law_1.01,6.31907844543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,1,balanced,2.719082514444987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,1,power_law_1.01,5.8272960662841795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,1,balanced,6.673269271850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,1,power_law_1.01,6.78826904296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,1,power_law_1.01,8.173286437988281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,1,balanced,4.211306571960449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,1,balanced,3.496954600016276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,1,power_law_1.01,9.831871795654298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,1,power_law_1.01,12.856224060058594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,1,balanced,7.3197282155354815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,1,power_law_1.01,14.779603576660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,1,balanced,4.886288007100423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,1,balanced,4.299471855163574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,1,power_law_1.01,18.802476501464845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,1,power_law_1.01,35.15172424316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,1,balanced,10.752890268961588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,1,balanced,6.706960042317708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,1,balanced,5.781167984008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,1,balanced,7.666234970092773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,1,balanced,12.79516855875651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,1,balanced,7.398096084594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,1,balanced,11.412218729654947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,1,balanced,16.9202880859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,1,balanced,10.691642761230469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,1,power_law_1.01,0.11080960035324097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,1,power_law_1.01,0.09356160163879394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,1,power_law_1.01,0.14381439685821534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,1,power_law_1.01,0.20139520168304442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,1,balanced,22.212793986002605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,1,power_law_1.01,0.14382079839706421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,1,power_law_1.01,0.24324479103088378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,1,power_law_1.01,0.24096639156341554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,1,power_law_1.01,0.4292031764984131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,1,power_law_1.01,0.2991872072219849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,1,power_law_1.01,0.6093696117401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,1,power_law_1.01,0.6162303924560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,1,power_law_1.01,0.7384511947631835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,1,power_law_1.01,0.9292927742004394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,1,balanced,32.14546203613281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,1,balanced,20.13704554239909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,1,power_law_1.01,0.9923263549804687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,1,power_law_1.01,1.078809642791748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,1,power_law_1.01,1.0189503669738769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,1,power_law_1.01,1.5076160430908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,1,power_law_1.01,1.0681983947753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,1,power_law_1.01,1.5384639739990233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,1,power_law_1.01,1.0971263885498046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,1,power_law_1.2,0.14485759735107423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,1,power_law_1.01,1.6471040725708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,1,power_law_1.2,0.0964352011680603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,1,power_law_1.01,1.1860544204711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,1,power_law_1.01,1.6964160919189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,1,power_law_1.2,0.2407360076904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,1,power_law_1.01,1.2148032188415527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,1,power_law_1.01,1.7640960693359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,1,power_law_1.2,0.14547200202941896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,1,power_law_1.01,1.2544639587402344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,1,power_law_1.2,0.32883200645446775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,1,power_law_1.01,1.8761024475097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,1,power_law_1.2,0.19281280040740967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,1,power_law_1.01,1.2742591857910157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,1,power_law_1.2,0.5770175933837891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,1,power_law_1.01,1.9250431060791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,1,power_law_1.2,0.31284480094909667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,1,power_law_1.2,1.0329855918884276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,1,power_law_1.01,1.294655990600586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,1,power_law_1.2,0.5410175800323487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,1,power_law_1.2,1.520576000213623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,1,power_law_1.01,1.9409984588623046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,1,power_law_1.2,0.7689663887023925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,1,power_law_1.01,1.362009620666504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,1,power_law_1.2,1.8920576095581054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,1,power_law_1.01,1.9918720245361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,1,power_law_1.2,2.647558403015137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,1,power_law_1.2,0.9875071525573731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,1,power_law_1.01,1.4803263664245605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,1,power_law_1.01,2.1038400650024416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,1,power_law_1.2,2.863520050048828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,1,power_law_1.2,1.453376007080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.01,1.5659071922302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,1,power_law_1.01,2.283929634094238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,1,power_law_1.2,3.0611648559570312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.01,1.7078079223632812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,1,power_law_1.2,1.530790424346924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,1,power_law_1.01,2.3213247299194335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,1,power_law_1.2,3.2881534576416014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.01,1.893574333190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,1,power_law_1.01,2.587929534912109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,1,power_law_1.2,1.542188835144043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,1,power_law_1.2,3.387859344482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,1,power_law_1.01,2.821785545349121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.01,2.2585792541503906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,1,power_law_1.2,3.4485824584960936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,1,power_law_1.2,1.6628543853759765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,1,power_law_1.2,3.5433982849121093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,1,power_law_1.01,3.3171199798583983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,1,power_law_1.2,1.7154367446899415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.01,2.396735954284668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,1,power_law_1.2,0.11391359567642212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,1,power_law_1.2,3.6923454284667967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,1,power_law_1.2,1.7740287780761719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,1,power_law_1.01,3.4287998199462892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.01,2.9937408447265623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,1,power_law_1.2,3.9513343811035155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,1,power_law_1.2,0.1452415943145752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,1,power_law_1.2,1.8589696884155273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,1,power_law_1.01,4.300678253173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,1,power_law_1.01,3.7649471282958986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,1,power_law_1.2,0.171942400932312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,1,power_law_1.2,1.9521087646484374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,1,power_law_1.2,4.04200325012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,1,power_law_1.01,4.759891128540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,1,power_law_1.2,0.2554559946060181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,1,power_law_1.2,1.9947328567504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,1,power_law_1.2,4.325894546508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,1,power_law_1.01,4.9883262634277346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,1,power_law_1.2,0.4023871898651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,1,power_law_1.01,6.012575912475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,1,power_law_1.2,2.152243232727051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,1,power_law_1.2,0.5466239929199219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,1,power_law_1.2,4.702329635620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,1,power_law_1.01,6.016447830200195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,1,power_law_1.2,2.304128074645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,1,power_law_1.2,0.6671872138977051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,1,power_law_1.01,8.003539276123046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,1,power_law_1.2,5.3155265808105465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,1,power_law_1.01,7.282182312011718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,1,power_law_1.2,2.4523008346557615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,1,power_law_1.2,0.9245823860168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,1,power_law_1.01,9.263046264648438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,1,power_law_1.2,5.712550354003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,1,power_law_1.01,10.892889404296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,1,power_law_1.2,0.9974656105041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,1,power_law_1.2,2.726131248474121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,1,power_law_1.01,12.380774688720702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,1,power_law_1.2,6.753932952880859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,1,power_law_1.2,1.024953556060791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,1,power_law_1.2,2.9923519134521483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,1,power_law_1.2,6.125113677978516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,1,power_law_1.2,1.0651968002319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,1,power_law_1.01,20.43193664550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,1,power_law_1.2,3.558726501464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,1,power_law_1.01,23.178303527832032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,1,power_law_1.2,1.1294655799865723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,1,power_law_1.2,7.213676452636719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,1,power_law_1.2,3.6279808044433595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,1,power_law_1.2,1.1810751914978028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,1,power_law_1.2,8.508684539794922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,1,power_law_1.2,1.2178560256958009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,1,power_law_1.2,4.490879821777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,1,power_law_1.2,1.2898367881774901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,1,power_law_1.2,10.345433807373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,1,power_law_1.2,4.922540664672852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,1,power_law_1.2,1.3300928115844726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,1,power_law_1.2,1.3379648208618165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,1,power_law_1.2,6.15544319152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,1,power_law_1.2,1.5098112106323243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,1,power_law_1.2,13.469413757324219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,1,power_law_1.2,8.432530975341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.2,1.5346752166748048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.2,1.7765823364257813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,1,power_law_1.2,9.710137939453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,1,power_law_1.2,16.17359313964844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.2,2.022604751586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.2,2.377408027648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,1,power_law_1.2,12.539366149902344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,1,power_law_1.2,19.538412475585936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.2,2.541708755493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.2,3.1615936279296877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,1,power_law_1.2,23.698789978027342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,1,power_law_1.2,3.9684799194335936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,1,power_law_1.2,35.66807556152344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,1,power_law_1.2,5.189664077758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,1,power_law_1.2,6.345267105102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,1,power_law_1.2,7.555411529541016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,1,power_law_1.2,11.104902648925782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,1,power_law_1.2,20.92332763671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,2,balanced,0.053077335158983864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,2,balanced,0.03842666745185852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,2,balanced,0.05821866790453593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,2,balanced,0.07221333185831706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,2,balanced,0.11693867047627766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,2,balanced,0.05436266462008158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,2,balanced,0.0767626663049062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,2,balanced,0.20360000928243002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,2,balanced,0.07522133489449818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,2,balanced,0.11338133613268535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,2,balanced,0.37248531977335614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,2,balanced,0.12223999698956807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,2,balanced,0.17931199073791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,2,balanced,0.379802664120992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,2,balanced,0.20609599351882935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,2,balanced,0.20945600668589273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,2,balanced,0.2938026587168376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,2,balanced,0.38121600945790607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,2,balanced,0.212336003780365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,2,balanced,0.2635200023651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,2,balanced,0.3800746599833171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,2,balanced,0.20941332976023355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,2,balanced,0.2655093272527059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,2,balanced,0.3829813400904338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,2,balanced,0.2118613322575887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,2,balanced,0.2704266707102458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,2,balanced,0.3797279993693034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,2,balanced,0.21391467253367105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,2,balanced,0.38521599769592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,2,balanced,0.25764266649882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,2,balanced,0.21463465690612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,2,balanced,0.38646399974823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,2,balanced,0.26475733518600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,2,balanced,0.21792533000310263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,2,balanced,0.25252266724904376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,2,balanced,0.21889066696166992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,2,balanced,0.39228800932566327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,2,balanced,0.2515466610590617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,2,balanced,0.22549867630004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,2,balanced,0.27265600363413495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,2,balanced,0.3936213254928589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,2,balanced,0.23227200905481973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,2,balanced,0.2527093291282654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,2,balanced,0.23586134115854898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,2,balanced,0.399402658144633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,2,balanced,0.27755733331044513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,2,balanced,0.24459733565648398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,2,balanced,0.4033973217010498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,2,balanced,0.27365867296854657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,2,balanced,0.41141335169474286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,2,balanced,0.2619946599006653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,2,balanced,0.26757333676020306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,2,balanced,0.4243466854095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,2,balanced,0.2790880004564921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,2,balanced,0.2837173342704773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,2,balanced,0.4532959858576457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,2,balanced,0.29240532716115314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,2,balanced,0.3166399995485942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,2,balanced,0.4742453495661418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,2,balanced,0.30374934275945026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,2,balanced,0.3480693499247233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,2,balanced,0.508458654085795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,2,balanced,0.3389066855112712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,2,balanced,0.4458506504694621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,2,balanced,0.6116746664047241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,2,balanced,0.5437920093536377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,2,balanced,0.5069973468780518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,2,balanced,0.6373120148976644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,2,balanced,0.5903786818186442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,2,balanced,0.7195359865824381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,2,balanced,0.8716320196787516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,2,balanced,0.8074986934661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,2,balanced,0.8588000138600668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,2,balanced,0.9696373144785563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,2,balanced,1.008021354675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,2,balanced,1.209930658340454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,2,balanced,1.3574825922648113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,2,balanced,1.4203039805094402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,2,balanced,0.0286613330245018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,2,power_law_1.01,0.07299839854240417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,2,balanced,0.03643200049797694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,2,power_law_1.01,0.06102399826049805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,2,balanced,0.05075199902057648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,2,power_law_1.01,0.09891200065612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,2,balanced,0.07253333429495494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,2,power_law_1.01,0.15409280061721803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,2,balanced,0.11965333422025044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,2,power_law_1.01,0.17549439668655395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,2,balanced,0.12353066603342693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,2,power_law_1.01,0.3209343910217285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,2,balanced,0.12770133217175803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,2,balanced,0.1406719982624054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,2,power_law_1.01,0.34377601146698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,2,balanced,0.14847466349601746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,2,power_law_1.01,0.3427328109741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,2,balanced,1.5998454093933105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,2,balanced,0.1566986640294393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,2,power_law_1.01,0.3461951971054077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,2,balanced,0.1432960033416748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,2,balanced,1.7289439837137859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,2,power_law_1.01,0.36009600162506106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,2,balanced,0.1509226659933726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,2,balanced,1.8546773592631023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,2,power_law_1.01,0.35635199546813967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,2,balanced,0.15868266423543295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,2,power_law_1.01,0.3596863985061646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,2,balanced,0.15289066235224405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,2,power_law_1.01,0.3758080005645752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,2,power_law_1.01,0.3891455888748169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,2,balanced,0.16316266854604086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,2,power_law_1.01,0.401964807510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,2,balanced,0.17156267166137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,2,power_law_1.01,0.417248010635376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,2,power_law_1.01,0.05271040201187134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,2,balanced,0.1867520014444987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,2,power_law_1.01,0.4195903778076172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,2,power_law_1.01,0.07841920256614685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,2,power_law_1.01,0.04764800071716309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,2,balanced,0.2135039965311686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,2,power_law_1.01,0.466534423828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,2,balanced,1.9574079513549805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,2,power_law_1.01,0.07422720193862915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,2,balanced,2.1185545921325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,2,power_law_1.01,0.06106240153312683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,2,balanced,2.3006505966186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,2,balanced,0.24227199951807657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,2,power_law_1.01,0.49733757972717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,2,power_law_1.01,0.10721919536590577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,2,power_law_1.01,0.08956800103187561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,2,power_law_1.01,0.027167999744415285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,2,balanced,0.3702666759490967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,2,power_law_1.01,0.5549248218536377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,2,power_law_1.01,0.13570560216903688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,2,power_law_1.01,0.12037119865417481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,2,power_law_1.01,0.028307199478149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,2,balanced,0.3487306833267212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,2,power_law_1.01,0.5727168083190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,2,power_law_1.01,0.17120640277862548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,2,power_law_1.01,0.17777279615402222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,2,balanced,3.0509866078694663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,2,balanced,3.2959893544514975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,2,balanced,3.6302293141682944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,2,power_law_1.01,0.23777918815612792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,2,power_law_1.01,0.04467200040817261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,2,power_law_1.01,0.6993535995483399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,2,power_law_1.01,0.18890880346298217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,2,balanced,0.33689598242441815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,2,power_law_1.01,0.7836671829223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,2,power_law_1.01,0.24641919136047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,2,power_law_1.01,0.06026880145072937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,2,power_law_1.01,0.1897536039352417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,2,balanced,0.3389120101928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,2,power_law_1.01,1.0572863578796388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,2,power_law_1.01,0.24309120178222657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,2,power_law_1.01,0.06747519969940186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,2,power_law_1.01,0.19379199743270875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,2,power_law_1.01,0.10165760517120362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,2,power_law_1.01,0.24342401027679444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,2,power_law_1.01,1.264198398590088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,2,power_law_1.01,0.10698879957199096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,2,power_law_1.01,0.2541311979293823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,2,power_law_1.01,0.20061440467834474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,2,power_law_1.01,1.6294336318969727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,2,power_law_1.01,0.11303679943084717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,2,balanced,0.6128960053126017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,2,power_law_1.01,0.2560447931289673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,2,power_law_1.01,0.20090880393981933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,2,power_law_1.01,0.11983360052108764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,2,power_law_1.01,0.26282880306243894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,2,power_law_1.01,2.1108928680419923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,2,power_law_1.01,0.2090751886367798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,2,power_law_1.01,0.1260543942451477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,2,power_law_1.01,0.2548223972320557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,2,power_law_1.01,0.2160192012786865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,2,power_law_1.01,2.586476707458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,2,power_law_1.01,0.12117120027542114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,2,power_law_1.01,0.2726016044616699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,2,power_law_1.01,0.22220160961151122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,2,power_law_1.01,0.12783360481262207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,2,power_law_1.01,0.27174398899078367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,2,power_law_1.01,3.891417694091797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,2,power_law_1.01,0.2347520112991333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,2,power_law_1.01,0.1360703945159912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,2,power_law_1.01,0.2798912048339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,2,balanced,0.5985653400421143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,2,power_law_1.01,0.25049600601196287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,2,power_law_1.01,7.754905700683594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,2,power_law_1.01,0.29461119174957273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,2,power_law_1.01,0.12774399518966675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,2,power_law_1.01,0.2505408048629761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,2,power_law_1.01,0.30198400020599364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,2,power_law_1.01,0.13422080278396606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.01,0.3408063888549805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,2,power_law_1.01,0.2843967914581299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,2,power_law_1.01,0.14673919677734376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,2,power_law_1.01,0.301363205909729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.01,0.4135807991027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,2,power_law_1.01,0.1677248001098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,2,power_law_1.01,0.3612159967422485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.01,0.47722878456115725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,2,power_law_1.01,0.20102400779724122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,2,power_law_1.01,0.40251522064208983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.01,0.24476161003112792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.01,0.6108607769012451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,2,power_law_1.01,0.49932160377502444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.01,0.3069119930267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,2,balanced,0.8648586273193359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.01,0.710041618347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.01,0.2606015920639038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,2,power_law_1.01,0.6044352054595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.01,0.9540927886962891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.01,0.2954495906829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,2,power_law_1.01,0.7877952098846436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,2,balanced,5.946741104125977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.01,1.1965951919555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.01,0.32395520210266116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,2,power_law_1.01,1.0799551963806153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.01,1.7342592239379884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.01,0.4331007957458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,2,balanced,6.5424957275390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,2,power_law_1.01,1.492300796508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.01,0.5386879920959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,2,power_law_1.01,1.8046592712402343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.01,2.1884288787841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,2,balanced,7.1725279490153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,2,power_law_1.01,2.278745651245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.01,0.7220928192138671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,2,power_law_1.01,2.648019218444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,2,power_law_1.01,3.257215881347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.01,0.9263551712036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,2,power_law_1.01,6.929011535644531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,2,power_law_1.01,4.049331283569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,2,power_law_1.01,1.1008511543273927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,2,balanced,1.1333920160929363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,2,power_law_1.01,1.759596824645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,2,power_law_1.01,8.345247650146485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,2,power_law_1.01,3.451385498046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,2,balanced,1.4046026865641277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,2,balanced,2.2244799931844077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,2,balanced,4.417632102966309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,2,power_law_1.2,0.0729088008403778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,2,power_law_1.2,0.052095997333526614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,2,power_law_1.2,0.061926400661468504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,2,power_law_1.2,0.04441600143909454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,2,power_law_1.2,0.09390079975128174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,2,power_law_1.2,0.060838401317596436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,2,power_law_1.2,0.13004159927368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,2,power_law_1.2,0.08241279721260071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,2,power_law_1.2,0.17091840505599976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,2,power_law_1.2,0.11267199516296386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,2,power_law_1.2,0.3217472076416016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,2,power_law_1.2,0.18074239492416383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,2,power_law_1.2,0.34485759735107424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,2,power_law_1.2,0.19029120206832886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,2,power_law_1.2,0.3534656047821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,2,power_law_1.2,0.19491839408874512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,2,power_law_1.2,0.3595072031021118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,2,power_law_1.2,0.2023360013961792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,2,power_law_1.2,0.20025599002838135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,2,power_law_1.2,0.360319995880127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,2,power_law_1.2,0.20673279762268065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,1,balanced,0.0540533314148585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,2,power_law_1.2,0.025567999482154845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,2,power_law_1.2,0.3724224090576172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,2,power_law_1.2,0.07553279995918274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,2,power_law_1.2,0.21419520378112794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,1,balanced,0.0746559997399648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,2,power_law_1.2,0.3719871997833252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,2,power_law_1.2,0.03139840066432953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,1,balanced,0.11948266625404358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,2,power_law_1.2,0.38322560787200927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,2,power_law_1.2,0.04322560131549835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,2,power_law_1.2,0.2157439947128296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,1,balanced,0.20578134059906006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,2,power_law_1.2,0.07329279780387879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,1,balanced,0.038015998899936676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,1,balanced,0.3726400136947632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,2,power_law_1.2,0.3980799913406372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,2,power_law_1.2,0.05475839972496033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,2,power_law_1.2,0.22602880001068115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,2,power_law_1.2,0.09294080138206481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,1,balanced,0.3737013339996338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,2,power_law_1.2,0.0656000018119812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,2,power_law_1.2,0.4190976142883301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,1,balanced,0.056608001391092934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,2,power_law_1.2,0.23991041183471679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,1,balanced,0.050240000089009605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,2,power_law_1.2,0.12151679992675782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,1,balanced,0.3778666655222575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,2,power_law_1.2,0.1016767978668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,2,power_law_1.2,0.15222400426864624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,2,power_law_1.2,0.2479935884475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,2,power_law_1.2,0.42272000312805175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,1,balanced,0.0800799975792567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,1,balanced,0.3792159954706828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,1,balanced,0.05946133534113566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,2,power_law_1.2,0.10804480314254761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,2,power_law_1.2,0.24035201072692872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,2,power_law_1.2,0.2545344114303589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,1,balanced,0.38041067123413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,1,balanced,0.1292693316936493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,2,power_law_1.2,0.43875842094421386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,2,power_law_1.2,0.2542464017868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,2,power_law_1.2,0.11487360000610351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,1,balanced,0.3840800126393636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,2,power_law_1.2,0.2895616054534912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,1,balanced,0.07853333155314128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,2,power_law_1.2,0.503110408782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,1,balanced,0.21682665745417276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,2,power_law_1.2,0.2562432050704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,1,balanced,0.38656000296274823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,2,power_law_1.2,0.12327680587768555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,2,power_law_1.2,0.3146559953689575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,2,power_law_1.2,0.2578752040863037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,1,balanced,0.11737599968910217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,2,power_law_1.2,0.5306816101074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,1,balanced,0.39055999120076496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,1,balanced,0.21735467513402304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,2,power_law_1.2,0.1302847981452942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,2,power_law_1.2,0.3836927890777588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,2,power_law_1.2,0.2606719970703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,1,balanced,0.18088533480962118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,1,balanced,0.2198773423830668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,1,balanced,0.3943733374277751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,2,power_law_1.2,0.5982207775115966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,2,power_law_1.2,0.12325119972229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,1,balanced,0.40061867237091064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,1,balanced,0.1879040002822876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,1,balanced,0.21961599588394165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,2,power_law_1.2,0.2587264060974121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,2,power_law_1.2,0.4164224147796631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,1,balanced,0.40906135241190594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,1,balanced,0.19179733594258627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,2,power_law_1.2,0.609062385559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,2,power_law_1.2,0.12624000310897826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,1,balanced,0.22004799048105875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,2,power_law_1.2,0.2692543983459473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,1,balanced,0.22933866580327353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,2,power_law_1.2,0.520633602142334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,1,balanced,0.4161546627680461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,1,balanced,0.1907306710879008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,2,power_law_1.2,0.736518383026123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,2,power_law_1.2,0.13690240383148194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,2,power_law_1.2,0.27178239822387695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,1,balanced,0.2262453238169352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,2,power_law_1.2,0.27831039428710935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,1,balanced,0.23408534129460654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,1,balanced,0.19193067153294882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,2,power_law_1.2,0.6511231899261475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,1,balanced,0.4288906653722127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,2,power_law_1.2,0.1297152042388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,2,power_law_1.2,0.8506112098693848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,2,power_law_1.2,0.28041601181030273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,1,balanced,0.23611734310785928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,1,balanced,0.2002613345781962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,2,power_law_1.2,0.13797119855880738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,1,balanced,0.2409600019454956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,2,power_law_1.2,0.8509311676025391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,2,power_law_1.2,0.28506879806518554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,1,balanced,0.19550933440526327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,2,power_law_1.2,1.1251392364501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,1,balanced,0.45389866828918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,1,balanced,0.2512320081392924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,2,power_law_1.2,0.14966399669647218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,2,power_law_1.2,1.1596351623535157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,2,power_law_1.2,0.3023551940917969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,1,balanced,0.2032053271929423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,2,power_law_1.2,1.4115263938903808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,2,power_law_1.2,0.168230402469635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,2,power_law_1.2,1.4415488243103027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,1,balanced,0.25731199979782104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,1,balanced,0.48875200748443604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,2,power_law_1.2,0.3281536102294922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,1,balanced,0.20569066206614176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,2,power_law_1.2,1.8851455688476562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,2,power_law_1.2,0.20777599811553954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.2,0.3632447957992554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,1,balanced,0.2121653358141581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,1,balanced,0.2709760069847107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,2,power_law_1.2,1.9528640747070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.2,0.24304640293121338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,2,power_law_1.2,2.3046592712402343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,1,balanced,0.5276266733805338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.2,0.4162879943847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,1,balanced,0.22446932395299277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,2,power_law_1.2,2.546451187133789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,1,balanced,0.29588266213734943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.2,0.3118272066116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,2,power_law_1.2,2.760723114013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,1,balanced,0.5842880010604858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.2,0.5165823936462403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,1,balanced,0.23335999250411987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.2,0.25452799797058107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,1,balanced,0.3307146628697713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,2,power_law_1.2,3.868102264404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.2,0.3074048042297363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,1,balanced,0.23932266235351562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.2,0.603443193435669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,2,power_law_1.2,4.379046249389648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,1,balanced,0.7310559749603271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,1,balanced,0.39849599202473956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,2,power_law_1.2,7.655219268798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.2,0.33860480785369873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.2,0.7395520210266113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,1,balanced,0.2743413249651591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,2,power_law_1.2,9.15051498413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,1,balanced,0.4602666695912679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.2,0.4469888210296631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,1,balanced,0.8022879759470621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.2,1.0022784233093263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,1,balanced,0.3070613344510396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,1,balanced,0.6206826766331991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.2,1.356383991241455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.2,0.5616576194763183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,1,balanced,0.3633013168970744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.2,1.7730432510375977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.2,0.7456704139709472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.2,2.198476791381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,1,balanced,0.4163839817047119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,1,balanced,0.021514666577180225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,1,balanced,0.7232693036397299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,1,balanced,1.1138559977213542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.2,1.054361629486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,2,power_law_1.2,2.7733055114746095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,1,balanced,0.02531733363866806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,2,power_law_1.2,1.1959808349609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,2,power_law_1.2,4.500057601928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,1,balanced,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,1,balanced,0.5906560023625692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,1,balanced,0.047797332207361855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,2,power_law_1.2,1.7972543716430665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,1,balanced,0.07467199862003326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,1,balanced,1.040725310643514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,2,power_law_1.2,9.078707122802735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,1,balanced,0.07656000057856242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,2,power_law_1.2,3.6885887145996095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,1,balanced,1.3147146701812744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,1,balanced,0.6912906964619955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,1,balanced,0.08078399797280629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,1,balanced,0.08390399813652039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,1,balanced,1.262229363123576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,1,balanced,0.08729599912961324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,1,balanced,0.09237333138783772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,1,balanced,1.0133972962697346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,1,balanced,0.09884799520174663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,1,balanced,0.09272000193595886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,1,balanced,1.8892052968343098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,1,balanced,0.09515733520189922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,1,balanced,0.1030399998029073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,1,balanced,1.2655146916707356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,1,balanced,0.09697600205739339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,1,balanced,1.8277759552001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,1,balanced,0.10286399722099304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,1,balanced,0.11125333110491435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,1,balanced,0.12032000223795573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,1,balanced,0.1267253359158834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,1,balanced,0.2347466746966044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,1,balanced,2.476640065511068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,1,balanced,1.8451199531555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,1,balanced,0.2781013250350952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,1,balanced,2.389861265818278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,1,balanced,0.21727999051411948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,1,balanced,0.22460265954335532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,1,balanced,2.426186720530192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,1,balanced,0.38337600231170654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,1,balanced,3.0403947830200195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,1,balanced,2.966298739115397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,1,balanced,0.386896014213562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,1,balanced,3.0352001190185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,1,balanced,0.560975988705953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,1,balanced,4.648810704549153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,1,balanced,0.7342186768849691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,1,balanced,4.633877436319987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,1,balanced,4.718154589335124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,1,balanced,0.9088746706644694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,1,balanced,1.4431734085083008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,1,balanced,9.105365117390951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,1,power_law_1.01,0.050457602739334105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,1,power_law_1.01,0.03404799997806549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,1,balanced,9.09070905049642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,1,power_law_1.01,0.059443199634552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,1,power_law_1.01,0.09622399806976319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,1,power_law_1.01,0.044268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,1,power_law_1.01,0.14355839490890504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,1,power_law_1.01,0.06618880033493042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,1,power_law_1.01,0.19052799940109252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,1,power_law_1.01,0.09188479781150818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,1,power_law_1.01,0.11797120571136474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,1,power_law_1.01,0.309004807472229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,1,power_law_1.01,0.18407679796218873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,1,power_law_1.01,0.3275840044021606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,1,power_law_1.01,0.1906559944152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,1,power_law_1.01,0.3450495958328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,1,power_law_1.01,0.1998528003692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,1,power_law_1.01,0.3467711925506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,1,power_law_1.01,0.35847039222717286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,1,power_law_1.01,0.3665791988372803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,1,power_law_1.01,0.04759039878845215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,1,power_law_1.01,0.20392320156097413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,1,balanced,9.142405192057291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,1,power_law_1.01,0.3720383882522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,1,power_law_1.01,0.05214080214500427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,1,power_law_1.01,0.2095168113708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,1,power_law_1.01,0.37725439071655276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,1,power_law_1.01,0.02126079946756363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,1,power_law_1.01,0.3923840045928955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,1,power_law_1.01,0.06924160122871399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,1,power_law_1.01,0.21971840858459474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,1,power_law_1.01,0.40729598999023436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,1,power_law_1.01,0.09423360228538513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,1,power_law_1.01,0.0240447998046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,1,power_law_1.01,0.4130112171173096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,1,power_law_1.01,0.11421439647674561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,1,power_law_1.01,0.22355198860168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,1,power_law_1.01,0.03304960131645203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,1,power_law_1.01,0.4250368118286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,1,power_law_1.01,0.15985920429229736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,1,power_law_1.01,0.46655998229980467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,1,power_law_1.01,0.1678272008895874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,1,power_law_1.01,0.2277888059616089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,1,power_law_1.01,0.04713599979877472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,1,power_law_1.01,0.5194431781768799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,1,power_law_1.01,0.1729024052619934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,1,power_law_1.01,0.06529279947280883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,1,power_law_1.01,0.23735039234161376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,1,power_law_1.01,0.17867519855499267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,1,power_law_1.01,0.5923840045928955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,1,power_law_1.01,0.09278720021247863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,1,power_law_1.01,0.1814911961555481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,1,power_law_1.01,0.6277184009552002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,1,power_law_1.01,0.25002241134643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,1,power_law_1.01,0.18659199476242067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,1,power_law_1.01,0.09513599872589111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,1,power_law_1.01,0.7696447849273682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,1,power_law_1.01,0.10210560560226441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,1,power_law_1.01,0.19414399862289428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,1,power_law_1.01,0.26073598861694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,1,power_law_1.01,0.2707007884979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,1,power_law_1.01,0.10677119493484497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,1,power_law_1.01,0.19831039905548095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,1,power_law_1.01,0.91397123336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,1,power_law_1.01,0.3103231906890869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,1,power_law_1.01,0.11449600458145141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,1,power_law_1.01,0.21099519729614258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,1,power_law_1.01,0.34914560317993165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,1,power_law_1.01,1.1947648048400878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,1,power_law_1.01,0.12059520483016968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,1,power_law_1.01,0.2265536069869995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,1,power_law_1.01,0.4237823963165283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,1,power_law_1.01,0.11541759967803955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,1,power_law_1.01,0.23556480407714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,1,power_law_1.01,1.505945587158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,1,power_law_1.01,0.48952322006225585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,1,power_law_1.01,0.11669119596481323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,1,power_law_1.01,0.24453120231628417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,1,power_law_1.01,0.6385727882385254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,1,power_law_1.01,2.032723236083984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,1,power_law_1.01,0.12510720491409302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,1,power_law_1.01,0.28713600635528563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,1,power_law_1.01,0.7971776008605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,1,power_law_1.01,0.12145919799804687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.01,0.3091200113296509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,1,power_law_1.01,1.092505645751953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,1,power_law_1.01,2.5818944931030274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,1,power_law_1.01,0.1263360023498535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.01,0.3797568082809448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,1,power_law_1.01,1.3598655700683593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,1,power_law_1.01,0.13587839603424073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.01,0.45015678405761717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,1,balanced,2.845696131388346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,1,power_law_1.01,3.151840019226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,1,power_law_1.01,0.1439039945602417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,1,power_law_1.01,1.9089664459228515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.01,0.6011007785797119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.01,0.15864319801330568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.01,0.7425151824951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,1,power_law_1.01,2.4738239288330077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,1,power_law_1.01,4.955558395385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.01,0.3385855913162231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.01,1.029094409942627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,1,power_law_1.01,3.0366847991943358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.01,0.38407039642333984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.01,1.3073344230651855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.01,0.251526403427124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,1,power_law_1.01,4.66363525390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,1,power_law_1.01,9.376223754882812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.01,1.868729591369629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.01,0.29563519954681394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,1,power_law_1.01,9.073811340332032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.01,0.3689471960067749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.01,2.4414655685424806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.01,0.4604928016662598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.01,3.0077695846557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.01,0.627948808670044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.01,0.799616003036499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,1,power_law_1.01,4.62848014831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.01,0.9669568061828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,1,power_law_1.01,9.186707305908204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,1,power_law_1.01,1.52805757522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,1,power_law_1.01,2.887820816040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,1,power_law_1.2,0.05230720043182373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,1,power_law_1.2,0.057868802547454835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,1,power_law_1.2,0.09026560187339783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,1,power_law_1.2,0.13181439638137818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,1,power_law_1.2,0.16655999422073364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,1,power_law_1.2,0.31125760078430176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,1,power_law_1.2,0.03324800133705139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,1,power_law_1.2,0.3286272048950195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,1,power_law_1.2,0.3444927930831909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,1,power_law_1.2,0.043584001064300534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,1,power_law_1.2,0.35676159858703616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,1,power_law_1.2,0.061561602354049685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,1,power_law_1.2,0.3659136056900024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,1,power_law_1.2,0.0869055986404419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,1,power_law_1.2,0.3661887884140015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,1,power_law_1.2,0.10577280521392822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,1,power_law_1.2,0.37194879055023194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,1,power_law_1.2,0.18622080087661744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,1,power_law_1.2,0.3772736072540283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,1,power_law_1.2,0.19393279552459716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,1,power_law_1.2,0.39881598949432373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,1,power_law_1.2,0.20600318908691406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,1,power_law_1.2,0.40787200927734374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,1,power_law_1.2,0.20794239044189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,1,power_law_1.2,0.42304000854492185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,1,power_law_1.2,0.21359360218048096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,1,power_law_1.2,0.4307392120361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,1,power_law_1.2,0.21712000370025636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,1,power_law_1.2,0.4826687812805176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,1,power_law_1.2,0.2249216079711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,1,power_law_1.2,0.5270080089569091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,1,power_law_1.2,0.22826240062713624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,1,power_law_1.2,0.6089087963104248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,1,power_law_1.2,0.24049921035766603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,1,power_law_1.2,0.04789760112762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,1,power_law_1.2,0.25331840515136717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,1,power_law_1.2,0.6430272102355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,1,power_law_1.2,0.0537280023097992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,1,power_law_1.2,0.2617664098739624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,1,power_law_1.2,0.7850240230560303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,1,power_law_1.2,0.06599680185317994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,1,power_law_1.2,0.2762687921524048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,1,power_law_1.2,0.08419839739799499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,1,power_law_1.2,0.9287551879882813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,1,power_law_1.2,0.31383039951324465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,1,power_law_1.2,0.10368640422821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,1,power_law_1.2,1.2049983978271483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,1,power_law_1.2,0.16378240585327147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,1,power_law_1.2,0.35651841163635256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,1,power_law_1.2,0.16921600103378295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,1,power_law_1.2,1.5055487632751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,1,power_law_1.2,0.43374080657958985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,1,power_law_1.2,0.1749119997024536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,1,power_law_1.2,0.4951039791107178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,1,power_law_1.2,2.063020706176758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,1,power_law_1.2,0.18318079710006713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,1,power_law_1.2,0.6469696044921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,1,power_law_1.2,0.18517119884490968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,1,power_law_1.2,2.6291072845458983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,1,power_law_1.2,0.19090559482574462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,1,power_law_1.2,0.8060671806335449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,1,power_law_1.2,0.19673600196838378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,1,power_law_1.2,3.183078384399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,1,power_law_1.2,1.0963135719299317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,1,power_law_1.2,0.20416638851165772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,1,power_law_1.2,1.3537023544311524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,1,power_law_1.2,0.214630389213562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,1,power_law_1.2,4.964812850952148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,1,power_law_1.2,0.23145599365234376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,1,power_law_1.2,1.9041088104248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,1,power_law_1.2,0.2401344060897827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,1,power_law_1.2,2.4751104354858398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,1,power_law_1.2,0.24978559017181395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,1,power_law_1.2,9.413721466064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,1,power_law_1.2,0.29212160110473634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,1,power_law_1.2,3.0400064468383787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.2,0.31975679397583007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.2,0.39089920520782473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,1,power_law_1.2,4.657855987548828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.2,0.4608767986297607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.2,0.6082880020141601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.2,0.7504767894744873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,1,power_law_1.2,9.046669006347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,2,balanced,0.10248532891273499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.2,1.032479953765869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.2,1.3137344360351562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,2,balanced,0.14909332990646362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,1,power_law_1.2,0.02114560008049011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.2,1.8698368072509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,2,balanced,0.25381867090861004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,2,balanced,0.46641600131988525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.2,2.455129623413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,1,power_law_1.2,0.02399359941482544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,2,balanced,0.8897919654846191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,1,power_law_1.2,0.03312000036239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.2,3.0251903533935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,2,balanced,1.1053120295206706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,1,power_law_1.2,0.04519039988517761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,2,balanced,1.1066559950510662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,1,power_law_1.2,0.06167680025100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,1,power_law_1.2,4.639257431030273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,2,balanced,0.07308266560236613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,2,balanced,1.1097493171691895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,1,power_law_1.2,0.09228799939155578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,1,power_law_1.2,0.10057599544525146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,2,balanced,1.1146346728007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,2,balanced,0.09935466448465984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,2,balanced,0.10519466797510783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,1,power_law_1.2,9.325567626953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,1,power_law_1.2,0.10580480098724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,2,balanced,1.1133973598480225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,2,balanced,0.15106133619944254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,1,power_law_1.2,0.10992000102996827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,2,balanced,0.2639946738878886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,2,balanced,0.15009066462516785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,2,balanced,1.1196906566619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,2,balanced,0.47867735226949054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,1,power_law_1.2,0.11257599592208863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,2,balanced,0.22831465800603232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,2,balanced,1.1265013217926025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,2,balanced,0.5800640185674032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,1,power_law_1.2,0.12299519777297974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,2,balanced,0.5825173457463583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,1,power_law_1.2,0.11542400121688842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,2,balanced,1.1238986651102703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,2,balanced,0.5840799808502197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,2,balanced,0.37514134248097736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,2,balanced,0.5896053314208984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,1,power_law_1.2,0.11978240013122558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,2,balanced,1.1318613688151042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,2,balanced,0.6255893309911092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,2,balanced,0.5886613527933756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,1,power_law_1.2,0.12718080282211303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,2,balanced,0.5947893460591634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,2,balanced,0.7106080055236816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,2,balanced,1.148586670557658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,2,balanced,0.6026026805241903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,1,power_law_1.2,0.1188480019569397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,2,balanced,0.05468800167242686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,2,balanced,0.6957866350809733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,2,balanced,0.6002293427785238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,2,balanced,1.1537493069966633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,2,balanced,0.6770666440327963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,1,power_law_1.2,0.12672640085220338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,2,balanced,0.6079146862030029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,2,balanced,0.0682666649421056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,2,balanced,0.6554986635843912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,2,balanced,1.1655680338541667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,1,power_law_1.2,0.13962240219116212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,2,balanced,0.6192640066146851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,2,balanced,0.09828799962997437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,2,balanced,0.6585813363393148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,1,power_law_1.2,0.14920320510864257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,2,balanced,1.3574560483296711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,2,balanced,0.6234720150629679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.2,0.158188796043396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,2,balanced,0.15291200081507364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,2,balanced,0.6693706512451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,2,balanced,0.6362719933191935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,2,balanced,1.2340213457743328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.2,0.339737606048584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,2,balanced,0.2606773376464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,2,balanced,0.7125226656595866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,2,balanced,0.7262559731801351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.2,0.3933183908462524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,2,balanced,1.489125410715739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,2,balanced,0.3260586659113566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,2,balanced,0.6717387040456136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,2,balanced,0.6838666598002116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.2,0.26696960926055907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,2,balanced,0.33818666140238446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.2,0.3065664052963257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,2,balanced,0.704474687576294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,2,balanced,0.8432160218556722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,2,balanced,0.3534506559371948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,2,balanced,1.3319786389668782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.2,0.3705343961715698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,2,balanced,0.36600534121195477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,2,balanced,0.7299893697102865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.2,0.468287992477417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,2,balanced,0.7935733000437418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,2,balanced,0.3803786834081014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.2,0.6355584144592286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,2,balanced,0.6715839703877767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,2,balanced,0.37250133355458576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,2,balanced,1.711125373840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.2,0.8032511711120606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,2,balanced,1.009434700012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,2,balanced,0.7149759928385416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,2,balanced,0.39103468259175617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,2,balanced,0.4102133512496948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.2,0.9754176139831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,2,balanced,0.6988053321838379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,2,balanced,0.38525867462158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,2,balanced,1.5697174072265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,2,balanced,0.40355201562245685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,2,balanced,1.0213812987009685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,1,power_law_1.2,1.5500672340393067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,2,balanced,0.6944479942321777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,2,balanced,0.42551998297373456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,2,balanced,0.4745440085728963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,1,power_law_1.2,2.8969280242919924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,2,balanced,0.707589308420817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,2,balanced,0.5359679857889811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,2,balanced,2.670735994974772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,2,balanced,1.5724533398946126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,2,balanced,0.47729067007700604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,2,balanced,0.7253706455230713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,2,balanced,0.8317920366923014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,2,balanced,1.3263786633809407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,2,balanced,0.8219359715779623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,2,balanced,1.6466186841328938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,2,balanced,2.480090618133545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,2,balanced,1.283941348393758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,2,balanced,0.8492000102996826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,2,balanced,0.8365546862284342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,2,balanced,1.8814613024393718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,2,balanced,2.7993812561035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,2,balanced,4.204501469930013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,2,balanced,1.5090187390645344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,2,balanced,2.4951574007670083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,2,balanced,1.5844799677530925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,2,balanced,3.0118773778279624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,2,balanced,4.674794514973958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,2,balanced,2.855104128519694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,2,balanced,2.365861256917318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,2,power_law_1.01,0.15538560152053832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,2,power_law_1.01,0.2337023973464966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,2,power_law_1.01,0.21137280464172364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,2,balanced,3.4207468032836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,2,power_law_1.01,0.359878396987915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,2,balanced,4.94865067799886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,2,power_law_1.01,0.48215041160583494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,2,balanced,4.119461377461751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,2,power_law_1.01,0.819820785522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,2,power_law_1.01,0.8628031730651855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,2,power_law_1.01,0.9451904296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,2,balanced,3.0256532033284507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,2,power_law_1.01,0.9510080337524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,2,power_law_1.01,1.0074687957763673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,2,power_law_1.01,0.09544960260391236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,2,power_law_1.01,0.9801024436950684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,2,power_law_1.01,1.0455360412597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,2,power_law_1.01,1.0573504447937012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,2,power_law_1.01,0.12942080497741698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,2,power_law_1.01,1.060307216644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,2,power_law_1.01,1.1455103874206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,2,power_law_1.01,0.112716805934906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,2,power_law_1.01,1.1429951667785645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,2,power_law_1.01,1.2316991806030273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,2,power_law_1.01,0.21319680213928222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,2,power_law_1.01,1.381663990020752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,2,power_law_1.01,0.2780735969543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,2,power_law_1.01,1.3369471549987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,2,power_law_1.01,0.45345277786254884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,2,power_law_1.01,1.4982463836669921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,2,balanced,4.431450525919597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,2,power_law_1.01,0.48010878562927245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,2,power_law_1.01,1.502086353302002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,2,power_law_1.01,0.47788801193237307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,2,power_law_1.01,1.8083776473999023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,2,power_law_1.01,0.5027520179748535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,2,balanced,3.091498692830404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,2,power_law_1.01,2.1443199157714843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,2,power_law_1.01,0.14079359769821168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,2,power_law_1.01,0.510316801071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,2,power_law_1.01,2.5177536010742188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,2,power_law_1.01,0.525491189956665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,2,power_law_1.01,0.15980160236358643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,2,balanced,5.451690673828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,2,power_law_1.01,0.5598464012145996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,2,power_law_1.01,3.5149505615234373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,2,balanced,8.464314778645834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,2,power_law_1.01,0.16867200136184693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,2,power_law_1.01,4.723590469360351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,2,power_law_1.01,0.2762752056121826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,2,power_law_1.01,0.5463232040405274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,2,power_law_1.01,0.35265278816223145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,2,power_law_1.01,0.5792960166931153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,2,power_law_1.01,5.564563369750976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,2,power_law_1.01,0.5535999774932862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,2,power_law_1.01,0.6067520141601562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,2,power_law_1.01,0.5787903785705566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,2,power_law_1.01,6.39738883972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,2,power_law_1.01,0.6136703968048096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,2,power_law_1.01,0.5813504219055176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,2,power_law_1.01,0.6560063838958741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,2,power_law_1.01,0.6113024234771729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,2,power_law_1.01,9.315500640869141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,2,power_law_1.01,0.7225984096527099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,2,power_law_1.01,0.6277056217193604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,2,power_law_1.01,0.7452991962432861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,2,power_law_1.01,0.6463679790496826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,2,power_law_1.01,0.8607616424560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,2,power_law_1.01,0.6434688091278076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,2,power_law_1.01,17.866259765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,2,power_law_1.01,0.6309887886047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,2,power_law_1.01,0.8910335540771485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,2,power_law_1.01,0.6716864109039307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,2,power_law_1.01,1.0760831832885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,2,power_law_1.01,0.6741695880889893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,2,power_law_1.01,1.262559986114502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,2,power_law_1.01,0.6659840106964111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,2,power_law_1.01,1.6072000503540038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,2,power_law_1.01,0.7072127819061279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,2,power_law_1.01,2.196505546569824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,2,power_law_1.01,0.7519999980926514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.01,0.8227456092834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,2,power_law_1.01,2.9389375686645507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,2,balanced,7.139797210693359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.01,0.9299967765808106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,2,power_law_1.01,3.5299392700195313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.01,1.0722240447998046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,2,balanced,5.540677388509114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.01,1.2970175743103027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,2,power_law_1.01,4.227967834472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.01,1.5562111854553222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,2,power_law_1.01,5.957420730590821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.01,2.0838783264160154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.01,2.45733757019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,2,power_law_1.01,11.680332946777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.01,3.4202945709228514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.01,4.540864181518555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,2,power_law_1.01,5.237395095825195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,2,power_law_1.01,8.562828826904298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,2,power_law_1.01,14.732736206054687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,2,balanced,10.284512201944986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,2,balanced,15.69146728515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,2,power_law_1.01,0.05371519923210144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,2,power_law_1.01,0.06734079718589783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,2,power_law_1.01,0.07390720248222352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,2,power_law_1.01,0.10961920022964478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,2,power_law_1.01,0.15896960496902465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,2,power_law_1.01,0.25128319263458254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,2,power_law_1.01,0.2658047914505005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,2,power_law_1.01,0.28211839199066163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,2,power_law_1.01,0.2874752044677734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,2,power_law_1.01,0.30309760570526123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,2,power_law_1.01,0.3078464031219482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,2,balanced,14.337632497151693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,2,power_law_1.01,0.3216896057128906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,2,power_law_1.01,0.3395391941070557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,2,balanced,9.85748291015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,2,power_law_1.01,0.330841588973999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,2,power_law_1.01,0.3393984079360962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,2,power_law_1.01,0.36119039058685304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,2,power_law_1.01,0.4041855812072754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,2,power_law_1.01,0.47832322120666504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.01,0.4150847911834717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.01,0.5098944187164307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.01,0.6234240055084228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.01,0.6738175868988037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.01,0.7495359897613525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.01,1.0065279960632325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.01,1.1700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.01,1.6416639328002929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.01,2.12554874420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,2,power_law_1.01,2.585369682312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,2,power_law_1.01,3.999923324584961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,2,power_law_1.01,7.554617309570313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,2,power_law_1.2,0.16005760431289673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,2,power_law_1.2,0.14056960344314576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,2,power_law_1.2,0.1661247968673706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,2,power_law_1.2,0.32814080715179444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,2,power_law_1.2,0.436902379989624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,2,power_law_1.2,0.8635199546813965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,2,power_law_1.2,0.9192383766174317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,2,power_law_1.2,0.9445440292358398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,2,power_law_1.2,1.0057215690612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,2,power_law_1.2,1.0002623558044434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,2,power_law_1.2,1.0490816116333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,2,power_law_1.2,1.077286434173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,2,power_law_1.2,1.1117312431335449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,2,power_law_1.2,1.079417610168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,2,power_law_1.2,1.1737088203430175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,2,power_law_1.2,1.1827456474304199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,2,power_law_1.2,1.2734527587890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,2,power_law_1.2,1.4252351760864257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,2,power_law_1.2,1.3824959754943849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,2,power_law_1.2,1.6349632263183593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,2,power_law_1.2,1.6482368469238282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,2,power_law_1.2,1.9077823638916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,2,power_law_1.2,0.09589759707450866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,2,power_law_1.2,2.294291114807129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,2,power_law_1.2,0.08370559811592101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,2,power_law_1.2,2.67127685546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,2,power_law_1.2,0.12569600343704224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,2,power_law_1.2,3.818745422363281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,2,power_law_1.2,0.17381759881973266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,2,power_law_1.2,0.2564608097076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,2,power_law_1.2,4.995251083374024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,2,power_law_1.2,0.4437888145446777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,2,power_law_1.2,6.358003234863281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,2,power_law_1.2,0.45993599891662595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,2,power_law_1.2,0.5100927829742432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,2,power_law_1.2,7.13768310546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,2,power_law_1.2,0.5050816059112548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,2,power_law_1.2,0.5296703815460205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,2,power_law_1.2,10.000806427001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,2,power_law_1.2,0.5467135906219482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,2,power_law_1.2,0.5825600147247314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,2,power_law_1.2,0.5829631805419921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,2,power_law_1.2,0.6120512008666992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,2,power_law_1.2,19.28297576904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,2,power_law_1.2,0.6260992050170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,2,power_law_1.2,0.6480447769165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,2,power_law_1.2,0.676204776763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,2,power_law_1.2,0.7659520149230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,2,power_law_1.2,0.7776000022888183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,2,power_law_1.2,0.8847231864929199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,1,balanced,0.0997920036315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,2,power_law_1.2,0.9183551788330078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,2,power_law_1.2,0.05128960013389587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,2,power_law_1.2,0.1410048007965088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,2,power_law_1.2,1.1254719734191894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,1,balanced,0.15315733353296915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,2,power_law_1.2,0.048467200994491574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,1,balanced,0.256549338499705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,2,power_law_1.2,1.3318528175354003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,2,power_law_1.2,0.12751359939575196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,1,balanced,0.4695626497268677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,2,power_law_1.2,0.0729088008403778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,2,power_law_1.2,1.6988096237182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,2,power_law_1.2,0.16726399660110475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,1,balanced,0.8853546778361002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,2,power_law_1.2,0.11241600513458253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,2,power_law_1.2,2.381318473815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,2,power_law_1.2,0.24216959476470948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,1,balanced,1.102842648824056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,2,power_law_1.2,0.3400768041610718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,2,power_law_1.2,0.1388479948043823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,2,power_law_1.2,3.087833595275879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,1,balanced,1.1035199960072835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,2,power_law_1.2,0.519052791595459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,2,power_law_1.2,0.2515968084335327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,1,balanced,0.07094933092594147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,1,balanced,0.11335466305414836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,2,power_law_1.2,0.2774143934249878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,2,power_law_1.2,0.5962560176849365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,1,balanced,1.1043039957682292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,1,balanced,0.09825600186983745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,2,power_law_1.2,0.2732032060623169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,1,balanced,1.1125866572062175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,2,power_law_1.2,0.6040832042694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,2,power_law_1.2,3.6425472259521485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,1,balanced,0.15652267138163248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,1,balanced,0.1507306694984436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,2,power_law_1.2,0.29633278846740724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,2,power_law_1.2,0.6388415813446044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,1,balanced,1.114261309305827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,1,balanced,0.2286613384882609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,1,balanced,0.2616693377494812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,2,power_law_1.2,0.30443520545959474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,2,power_law_1.2,0.6450111865997314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,1,balanced,1.1268906593322754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,1,balanced,0.3701333204905192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,1,balanced,0.47620264689127606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,2,power_law_1.2,4.504249572753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,2,power_law_1.2,0.3097471952438354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,1,balanced,0.5821546713511149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,1,balanced,0.6248106559117635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,2,power_law_1.2,0.6518208026885987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,1,balanced,1.1309653123219807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,2,power_law_1.2,0.32690560817718506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,1,balanced,0.5890506505966187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,2,power_law_1.2,6.470259094238282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,1,balanced,0.5927733182907104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,2,power_law_1.2,0.34431359767913816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,2,power_law_1.2,0.670137596130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,1,balanced,1.1347520351409912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,1,balanced,0.7112533251444498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,1,balanced,0.5960533221562704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,2,power_law_1.2,0.3416960000991821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,2,power_law_1.2,0.6880447864532471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,1,balanced,0.5988586743672689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,1,balanced,0.6850293477376302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,1,balanced,1.1477279663085938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,2,power_law_1.2,0.3548095941543579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,2,power_law_1.2,13.912141418457031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,2,power_law_1.2,0.6871103763580322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,1,balanced,0.6043359835942587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,1,balanced,0.6842933495839437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,2,power_law_1.2,0.38730878829956056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,2,power_law_1.2,0.6894336223602295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,1,balanced,1.1622933546702068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,1,balanced,0.6113653182983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,1,balanced,0.7007893721262614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,2,power_law_1.2,0.43555197715759275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,2,power_law_1.2,0.6984960079193115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,1,balanced,0.6124906539916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,1,balanced,1.166543960571289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,1,balanced,0.683568000793457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,2,power_law_1.2,0.5035200119018555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,2,power_law_1.2,0.7429952144622802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,1,balanced,0.6238773266474406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,1,balanced,0.6759893099466959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.2,0.4415679931640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,2,power_law_1.2,0.7894527912139893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,1,balanced,1.1900266806284587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,1,balanced,0.6364320119222006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,1,balanced,0.6749707063039144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.2,0.5121024131774903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.2,0.8433088302612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,1,balanced,1.3584319750467937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,1,balanced,0.6488106648127238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,1,balanced,0.6970346768697103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.2,0.6322879791259766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,1,balanced,0.04238399863243103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,1,balanced,0.7156213124593099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.2,0.9815296173095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,1,balanced,0.6678826808929443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,1,balanced,1.270901362101237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.2,0.7077760219573974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,1,balanced,0.04411733150482178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,1,balanced,0.6998453140258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.2,1.1066752433776856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,1,balanced,0.06150400141874949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.2,0.8107328414916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,1,balanced,0.7639146645863851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,1,balanced,1.4943307240804036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,1,balanced,0.708463986714681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.2,1.3505599975585938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,1,balanced,0.08855467041333516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.2,1.009779167175293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,1,balanced,0.7401119867960612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,1,balanced,0.7478293577829996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,1,balanced,1.4120532671610515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.2,1.6011072158813477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,1,balanced,0.14843733112017313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,1,balanced,0.7358773549397787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.2,2.0401792526245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.2,1.2729855537414552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,1,balanced,0.9212053616841634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,1,balanced,0.18331199884414673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.2,2.5441408157348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,1,balanced,2.081018606821696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,1,balanced,0.7611786524454752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.2,1.7379199981689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,1,balanced,0.19130667050679526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.2,3.4583038330078124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,1,balanced,0.9047359625498453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,1,balanced,0.19006399313608804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.2,2.1711551666259767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,1,balanced,0.8187200228373209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,1,balanced,0.19617066780726114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.2,4.65494384765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,1,balanced,0.20122667153676352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,1,balanced,1.3375946680704753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,2,power_law_1.2,2.6071039199829102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,1,balanced,1.7241172790527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,1,balanced,0.21284266312917074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,2,power_law_1.2,6.015481567382812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,1,balanced,0.8816800117492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,1,balanced,0.22419732809066772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,2,power_law_1.2,4.059641647338867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,1,balanced,0.23779733975728354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,2,power_law_1.2,8.02520980834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,1,balanced,0.22243199745814005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,1,balanced,1.2725120385487874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,1,balanced,1.488437334696452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,1,balanced,0.23439466953277588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,2,power_law_1.2,7.437586975097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,1,balanced,0.24758932987848917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,2,power_law_1.2,17.816677856445313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,1,balanced,2.6437973976135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,1,balanced,0.27187200387318927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,1,balanced,0.35924800237019855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,1,balanced,1.5249120394388835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,1,balanced,1.9119253158569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,1,balanced,0.3479626576105754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,1,balanced,0.4627946615219116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,1,balanced,2.753706614176432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,1,balanced,2.1812373797098794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,1,balanced,0.4298400084177653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,1,balanced,2.156895955403646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,1,balanced,0.5925333499908447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,1,balanced,2.935082753499349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,1,balanced,0.4924000104268392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,1,balanced,4.899584134419759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,1,balanced,3.293386777242025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,1,balanced,0.8302026589711508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,1,balanced,3.6280959447224936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,1,power_law_1.01,0.09703680276870727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,1,power_law_1.01,0.1363584041595459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,1,power_law_1.01,0.19022079706192016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,1,power_law_1.01,0.34476799964904786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,1,balanced,0.8955360253651937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,1,balanced,5.113802591959636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,1,power_law_1.01,0.4751999855041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,1,power_law_1.01,0.7935167789459229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,1,balanced,4.062282562255859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,1,power_law_1.01,0.8923775672912597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,1,power_law_1.01,0.9308544158935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,1,power_law_1.01,0.948851203918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,1,power_law_1.01,0.9925824165344238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,1,power_law_1.01,0.9922816276550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,1,power_law_1.01,1.048896026611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,1,power_law_1.01,1.0454527854919433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,1,balanced,5.021381378173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,1,power_law_1.01,1.1048959732055663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,1,balanced,1.2981812953948975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,1,power_law_1.01,1.1254783630371095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,1,power_law_1.01,1.1443391799926759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,1,power_law_1.01,1.1937984466552733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,1,power_law_1.01,1.3219200134277345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,1,balanced,5.596250534057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,1,power_law_1.01,1.3083392143249513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,1,balanced,4.584320068359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,1,power_law_1.01,1.4865983963012694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,1,power_law_1.01,1.6867584228515624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,1,power_law_1.01,2.032966423034668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,1,power_law_1.01,2.1069440841674805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,1,power_law_1.01,2.587808036804199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,1,balanced,1.691765308380127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,1,power_law_1.01,3.7064064025878904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,1,balanced,5.742229461669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,1,power_law_1.01,4.7126209259033205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,1,power_law_1.01,5.5350593566894535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,1,power_law_1.01,6.583071899414063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,1,power_law_1.01,10.367411041259766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,1,power_law_1.01,18.337452697753907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,1,balanced,1.7960586547851562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,1,balanced,10.237775802612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,1,balanced,7.459679921468099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,1,power_law_1.01,0.06849920153617858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,1,power_law_1.01,0.08734719753265381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,1,balanced,9.10594113667806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,1,power_law_1.01,0.11875840425491332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,1,power_law_1.01,0.1950976014137268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,1,power_law_1.01,0.26179840564727785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,1,power_law_1.01,0.4521791934967041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,1,power_law_1.01,0.47374720573425294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,1,power_law_1.01,0.5028863906860351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,1,power_law_1.01,0.5123392105102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,1,power_law_1.01,0.525273609161377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,1,power_law_1.01,0.5531519889831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,1,power_law_1.01,0.5577407836914062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,1,power_law_1.01,0.5595263957977294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,1,power_law_1.01,0.5826816082000732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,1,power_law_1.01,0.6085696220397949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,1,power_law_1.01,0.6292223930358887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,1,power_law_1.01,0.6683648109436036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,1,balanced,3.0741653442382812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,1,power_law_1.01,0.7499584197998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,1,power_law_1.01,0.775545597076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,1,power_law_1.01,0.9235584259033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,1,power_law_1.01,1.0450240135192872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,1,power_law_1.01,1.3208895683288575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,1,power_law_1.01,1.4500288009643554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,1,power_law_1.01,1.8920768737792968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,1,power_law_1.01,2.434726333618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,1,power_law_1.01,3.2689151763916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,1,power_law_1.01,4.30909423828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,1,power_law_1.01,5.162726211547851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,1,power_law_1.01,7.422700500488281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,1,power_law_1.01,14.433311462402344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,1,balanced,17.947210947672527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,1,balanced,14.342783610026041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,1,power_law_1.01,0.03199360072612763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,1,balanced,17.98204803466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,1,power_law_1.01,0.10707839727401733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,1,power_law_1.01,0.0450111985206604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,1,power_law_1.01,0.1323199987411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,1,power_law_1.01,0.05315840244293213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,1,power_law_1.01,0.18024319410324097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,1,power_law_1.01,0.2636735916137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,1,power_law_1.01,0.08966400027275086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,1,power_law_1.01,0.35688319206237795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,1,power_law_1.01,0.13479039669036866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,1,power_law_1.01,0.5502463817596436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,1,power_law_1.01,0.5779839992523194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,1,power_law_1.01,0.20855679512023925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,1,power_law_1.01,0.60000638961792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,1,power_law_1.01,0.6292287826538085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,1,power_law_1.01,0.2446592092514038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,1,power_law_1.01,0.6218751907348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,1,power_law_1.01,0.24659841060638427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,1,power_law_1.01,0.6397056102752685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,1,power_law_1.01,0.26204800605773926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,1,power_law_1.01,0.6389567852020264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,1,power_law_1.01,0.25941760540008546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,1,power_law_1.01,0.6520192146301269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,1,balanced,5.654517491658528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,1,power_law_1.01,0.27239038944244387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,1,power_law_1.01,0.6553152084350586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,1,power_law_1.01,0.2783423900604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,1,power_law_1.01,0.6785664081573486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,1,power_law_1.01,0.295084810256958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,1,power_law_1.01,0.6767936229705811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,1,power_law_1.01,0.7288959980010986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,1,power_law_1.01,0.29003520011901857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,1,power_law_1.01,0.8025471687316894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,1,power_law_1.01,0.3030783891677856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.01,0.855622386932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.01,1.000723171234131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,1,power_law_1.01,0.30310399532318116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.01,1.1194111824035644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,1,power_law_1.01,0.32255361080169676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.01,1.4071359634399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,1,power_law_1.01,0.40250239372253416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.01,1.683807945251465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.01,0.4266496181488037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.01,2.2006399154663088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.01,0.5377024173736572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.01,2.747238349914551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.01,0.5112895965576172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.01,3.803263854980469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.01,0.651091194152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.01,4.864601516723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.01,0.6746816158294677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,1,power_law_1.01,5.922828674316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.01,0.8205760002136231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.01,0.9714752197265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,1,power_law_1.01,9.050835418701173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.01,1.3271424293518066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.01,1.6681024551391601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,1,power_law_1.01,17.516455078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,1,power_law_1.01,2.0031103134155273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,1,power_law_1.01,3.0332799911499024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,1,power_law_1.01,5.769267272949219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,1,power_law_1.2,0.09721599817276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,1,power_law_1.2,0.10557440519332886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,1,power_law_1.2,0.17860480546951293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,1,power_law_1.2,0.29329919815063477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,1,power_law_1.2,0.4220223903656006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,1,power_law_1.2,0.8242688179016113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,1,power_law_1.2,0.9064384460449219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,1,power_law_1.2,0.9735872268676757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,1,power_law_1.2,0.9606016159057618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,1,power_law_1.2,0.9731328010559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,1,power_law_1.2,1.0326208114624023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,1,power_law_1.2,1.059340763092041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,1,power_law_1.2,1.0935551643371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,1,power_law_1.2,1.1108480453491212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,1,power_law_1.2,1.1728704452514649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,1,power_law_1.2,1.204089641571045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,1,power_law_1.2,1.246291160583496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,1,power_law_1.2,1.37194881439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,1,power_law_1.2,1.3653823852539062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,1,power_law_1.2,1.5390463829040528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,1,power_law_1.2,1.7295808792114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,1,power_law_1.2,2.090809631347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,1,power_law_1.2,2.1622976303100585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,1,power_law_1.2,2.6337408065795898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,1,power_law_1.2,3.8116992950439452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,1,power_law_1.2,4.8149055480957035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,1,power_law_1.2,5.6489921569824215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,1,power_law_1.2,6.706918334960937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,1,power_law_1.2,10.203250885009766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,1,power_law_1.2,18.568479919433592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,1,power_law_1.2,0.07106559872627258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,1,power_law_1.2,0.10581760406494141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,1,power_law_1.2,0.0800320029258728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,1,power_law_1.2,0.11755520105361938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,1,power_law_1.2,0.11285120248794556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,1,power_law_1.2,0.16759040355682372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,1,power_law_1.2,0.17535359859466554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,1,power_law_1.2,0.2383807897567749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,1,power_law_1.2,0.2415616035461426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,1,power_law_1.2,0.4419072151184082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,1,power_law_1.2,0.3149888038635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,1,power_law_1.2,0.03214719891548157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,1,power_law_1.2,0.4865407943725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,1,power_law_1.2,0.5340928077697754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,2,balanced,0.09243733684221904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,1,power_law_1.2,0.5028607845306396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,1,power_law_1.2,0.038867199420928956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,1,power_law_1.2,0.5886528015136718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,1,power_law_1.2,0.532147216796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,2,balanced,0.14257599910100302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,1,power_law_1.2,0.597977590560913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,1,power_law_1.2,0.05422080159187317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,1,power_law_1.2,0.5374591827392579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,2,balanced,0.24254933993021646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,1,power_law_1.2,0.6425471782684327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,1,power_law_1.2,0.08221439719200134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,2,balanced,0.4407680034637451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,1,power_law_1.2,0.5664639949798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,1,power_law_1.2,0.67259521484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,1,power_law_1.2,0.11320960521697998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,2,balanced,0.8328479925791422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,1,power_law_1.2,0.5746047973632813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,1,power_law_1.2,0.6802175998687744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,1,power_law_1.2,0.21889920234680177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,2,balanced,1.6165493329366047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,1,power_law_1.2,0.5944640159606933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,1,power_law_1.2,0.6718656063079834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,1,power_law_1.2,0.2344896078109741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,2,balanced,2.404266675313314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,1,power_law_1.2,0.6135231971740722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,1,power_law_1.2,0.69169921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,1,power_law_1.2,0.24567680358886718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,2,balanced,2.400181293487549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,1,power_law_1.2,0.6184895992279053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,1,power_law_1.2,0.6841279983520507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,1,power_law_1.2,0.2508352041244507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,2,balanced,2.407690684000651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,1,power_law_1.2,0.6433856010437011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,1,power_law_1.2,0.7139647960662842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,1,power_law_1.2,0.2711424112319946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,2,balanced,2.4104320208231607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,1,power_law_1.2,0.7366911888122558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,1,power_law_1.2,0.6856639862060547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,1,power_law_1.2,0.27947518825531004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,2,balanced,2.407562732696533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,1,power_law_1.2,0.7528768062591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,1,power_law_1.2,0.29120640754699706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,1,power_law_1.2,0.7878719806671143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,2,balanced,2.417461395263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,1,power_law_1.2,0.8358719825744629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,1,power_law_1.2,0.3064127922058105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,1,power_law_1.2,0.806611156463623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,2,balanced,2.4205493927001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,1,power_law_1.2,0.2962239980697632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.2,0.9036992073059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,2,balanced,2.4566453297932944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,1,power_law_1.2,0.9506112098693847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,2,balanced,0.07097066442171733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,1,power_law_1.2,0.3065792083740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.2,1.0215999603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,2,balanced,2.472442626953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,1,power_law_1.2,1.07258882522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,1,power_law_1.2,0.31931519508361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,2,balanced,0.09522666533788045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.2,1.1470399856567384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,2,balanced,2.4811946551005044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,1,power_law_1.2,1.3353280067443847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,1,power_law_1.2,0.35415680408477784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,2,balanced,0.14407466848691305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.2,1.439065647125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,1,power_law_1.2,1.4651904106140137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,2,balanced,2.480624039967855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,1,power_law_1.2,0.41179518699645995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,2,balanced,0.24775999784469604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.2,1.7064447402954102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,2,balanced,0.45119468371073407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,1,power_law_1.2,1.908345603942871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.2,2.2424575805664064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,2,balanced,2.510277271270752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.2,0.42340478897094724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,2,balanced,0.8386080265045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,1,power_law_1.2,2.449113655090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.2,0.5499199867248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.2,2.7708160400390627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,2,balanced,2.545701344807943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,2,balanced,1.234496037165324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.2,0.5077119827270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,1,power_law_1.2,3.311692810058594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,2,balanced,1.2364479700724285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.2,3.801881790161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,2,balanced,2.6052160263061523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.2,0.6686272144317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,2,balanced,1.241967995961507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,1,power_law_1.2,4.388608169555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,2,balanced,1.2402026653289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.2,0.6872064113616944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.2,4.854419326782226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,2,balanced,1.2443040211995442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,2,balanced,2.699029286702474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,1,power_law_1.2,5.249324798583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.2,0.8291584014892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,2,balanced,1.2526453336079915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,1,power_law_1.2,5.940857696533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.2,0.993126392364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,1,power_law_1.2,7.482112121582031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,2,balanced,1.2530506451924641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,2,balanced,2.8230934143066406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.2,1.348262405395508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,2,balanced,1.2655519644419353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,2,balanced,1.2794880072275798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,1,power_law_1.2,9.087289428710937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.2,1.7021503448486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,1,power_law_1.2,14.614726257324218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,2,balanced,1.2897066275278728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,1,power_law_1.2,2.0283584594726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,2,balanced,1.3003253142038982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,2,balanced,2.9190241495768228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,1,power_law_1.2,17.42909393310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,1,power_law_1.2,3.077548789978027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,2,balanced,1.3415412902832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,1,power_law_1.2,5.809088134765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,2,balanced,1.3727092742919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,2,balanced,3.139258702596029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,2,balanced,0.07441066702206929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,2,balanced,1.4251359303792317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,2,balanced,0.09519466757774353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,2,power_law_1.01,0.14090880155563354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,2,balanced,0.12553067008654276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,2,balanced,0.1896053353945414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,2,balanced,1.4913333257039387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,2,power_law_1.01,0.23909759521484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,2,balanced,0.31282132863998413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,2,balanced,3.434000015258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,2,balanced,0.5549920002619425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,2,power_law_1.01,0.43449602127075193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,2,balanced,0.797274669011434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,2,balanced,0.8013652960459391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,2,balanced,0.804101308186849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,2,balanced,1.5987946192423503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,2,power_law_1.01,0.329203200340271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,2,balanced,0.8077653249104818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,2,balanced,0.8088533083597819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,2,power_law_1.01,0.652729606628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,2,balanced,0.8164373238881429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,2,power_law_1.01,1.002143955230713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,2,balanced,0.8204692999521891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,2,power_law_1.01,1.1340415954589844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,2,balanced,0.8277173042297363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,2,balanced,1.7293334007263184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,2,balanced,0.8369173208872477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,2,power_law_1.01,1.5163519859313965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,2,balanced,3.8107306162516275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,2,balanced,0.8433866500854492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,2,power_law_1.01,1.559059238433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,2,balanced,0.8610826333363851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,2,power_law_1.01,1.6971904754638671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,2,balanced,0.8842559655507406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,2,power_law_1.01,1.7265727996826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,2,balanced,1.9458400408426921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,2,power_law_1.01,1.8007808685302735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,2,balanced,0.9059092998504639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,2,power_law_1.01,1.7951295852661133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,2,balanced,0.95469864209493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,2,power_law_1.01,1.8740415573120117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,2,balanced,1.007322629292806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,2,power_law_1.01,1.9869375228881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,2,power_law_1.01,2.034572792053223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,2,balanced,1.1278186639149983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,2,power_law_1.01,2.161446380615234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,2,balanced,2.278997262318929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,2,balanced,5.392730712890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,2,power_law_1.01,2.2504512786865236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,2,power_law_1.01,2.4269567489624024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,2,balanced,1.2453546524047852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,2,power_law_1.01,2.654745674133301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,2,power_law_1.01,3.0742143630981444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,2,power_law_1.01,3.616204833984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,2,balanced,1.456725279490153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,2,balanced,2.711541175842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,2,power_law_1.01,2.9855487823486326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,2,power_law_1.01,3.6125694274902345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,2,balanced,1.8710239728291829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,2,power_law_1.01,4.317119979858399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,2,balanced,6.245157241821289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,2,power_law_1.01,5.1115264892578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,2,power_law_1.01,6.9486335754394535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,2,power_law_1.01,8.1512451171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,2,balanced,3.6642185846964517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,2,power_law_1.01,10.485056304931641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,2,balanced,2.2941013971964517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,2,power_law_1.01,19.23020782470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,2,power_law_1.01,0.09523839950561523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,2,balanced,3.104485193888346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,2,power_law_1.01,0.144268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,2,balanced,4.195791880289714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,2,power_law_1.01,0.24061439037322999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,2,balanced,8.6243896484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,2,power_law_1.01,0.20342400074005126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,2,power_law_1.01,0.33648641109466554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,2,power_law_1.01,0.4890304088592529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,2,power_law_1.01,0.5850048065185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,2,power_law_1.01,0.7857791900634765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,2,power_law_1.01,0.809875202178955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,2,power_law_1.01,0.8803584098815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,2,power_law_1.01,0.9030336380004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,2,power_law_1.01,0.930720043182373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,2,balanced,3.8723678588867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,2,power_law_1.01,0.9543744087219238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,2,power_law_1.01,0.9812416076660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,2,power_law_1.01,1.0098496437072755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,2,power_law_1.01,1.0448512077331542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,2,power_law_1.01,1.076095962524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,2,power_law_1.01,1.1796992301940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,2,power_law_1.01,1.198796844482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,2,power_law_1.01,1.3928895950317384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,2,power_law_1.01,1.4966400146484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,2,balanced,6.297418594360352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,2,power_law_1.01,1.772198486328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,2,power_law_1.01,1.783046340942383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,2,power_law_1.01,2.1886079788208006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,2,power_law_1.01,2.596588706970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,2,power_law_1.01,3.2989952087402346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,2,power_law_1.01,4.542208099365235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,2,power_law_1.01,5.234579086303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,2,power_law_1.01,6.99438705444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,2,balanced,5.700554529825847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,2,power_law_1.01,12.947193908691407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,2,power_law_1.01,0.09232640266418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,2,balanced,16.410655975341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,2,power_law_1.01,0.125491201877594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,2,power_law_1.01,0.18035199642181396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,2,power_law_1.01,0.16120320558547974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,2,power_law_1.01,0.23948159217834472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,2,power_law_1.01,0.35163519382476804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,2,power_law_1.01,0.4078207969665527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,2,power_law_1.01,0.5166143894195556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,2,power_law_1.01,0.5510975837707519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,2,power_law_1.2,0.1406272053718567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,2,power_law_1.01,0.5764480113983155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,2,power_law_1.2,0.23975679874420167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,2,power_law_1.01,0.5874112129211426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,2,power_law_1.2,0.09301120042800903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,2,power_law_1.2,0.25246078968048097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,2,power_law_1.01,0.6268928050994873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,2,power_law_1.2,0.27633280754089357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,2,power_law_1.01,0.6217088222503662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,2,power_law_1.2,0.14261759519577027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,2,power_law_1.01,0.6495232105255127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,2,power_law_1.2,0.154310405254364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,2,power_law_1.2,0.5721536159515381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,2,power_law_1.01,0.6871935844421386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,2,power_law_1.2,0.1893952012062073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,2,power_law_1.2,0.8039872169494628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,2,power_law_1.01,0.7082687854766846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,2,power_law_1.2,0.2975680112838745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,2,power_law_1.01,0.7152512073516846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,2,power_law_1.2,0.43363199234008787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,2,power_law_1.2,0.996889591217041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,2,power_law_1.01,0.7848320007324219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,2,power_law_1.2,0.5432703971862793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,2,power_law_1.01,0.8088000297546387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,2,power_law_1.2,1.39749116897583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,2,power_law_1.2,0.8024959564208984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,2,power_law_1.01,0.9229439735412598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,2,power_law_1.2,1.5452159881591796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,2,balanced,11.994725545247396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,2,power_law_1.2,0.7894527912139893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,2,power_law_1.01,1.010476779937744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,2,power_law_1.2,1.5082624435424805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,2,power_law_1.2,0.8047039985656739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,2,power_law_1.01,1.2522944450378417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,2,power_law_1.2,1.6828800201416017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,2,power_law_1.2,0.8901632308959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,2,power_law_1.01,1.3156864166259765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,2,power_law_1.2,1.7176511764526368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,2,power_law_1.2,0.9126463890075683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,2,power_law_1.01,1.6330495834350587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,2,power_law_1.2,1.778374481201172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,2,power_law_1.2,0.9227519989013672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,2,power_law_1.01,2.0786111831665037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,2,power_law_1.2,0.9862272262573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,2,power_law_1.2,1.8736448287963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,2,power_law_1.01,2.8010175704956053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,2,power_law_1.2,1.024665641784668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,2,power_law_1.2,1.8544191360473632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,2,power_law_1.01,3.4750720977783205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,2,power_law_1.2,1.0301055908203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,2,balanced,10.949701944986979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,2,power_law_1.2,1.9803712844848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,2,power_law_1.01,4.163103866577148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,2,power_law_1.2,1.1020480155944825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,2,power_law_1.2,1.1883328437805176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,2,power_law_1.2,2.0916927337646483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,2,power_law_1.01,6.2229759216308596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,2,power_law_1.2,1.2756671905517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,2,power_law_1.2,2.328486442565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,2,power_law_1.2,1.4416511535644532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,2,power_law_1.2,2.5186880111694334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,2,power_law_1.01,11.946803283691406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,2,power_law_1.2,1.5406463623046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,2,power_law_1.2,2.952934455871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,2,power_law_1.2,1.8380607604980468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,2,power_law_1.2,3.306515121459961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,2,power_law_1.2,1.8985984802246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,2,power_law_1.2,3.832767868041992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,2,power_law_1.2,2.3062976837158202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,2,power_law_1.2,3.1927104949951173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,2,power_law_1.2,2.740300750732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,2,power_law_1.2,3.8340480804443358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,2,power_law_1.2,4.532774353027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,2,power_law_1.2,3.509715270996094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,2,power_law_1.2,5.710835266113281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,2,power_law_1.2,4.730527877807617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,2,power_law_1.2,7.434559631347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,2,power_law_1.2,5.530956649780274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,2,power_law_1.2,8.835398101806641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,2,power_law_1.2,7.544569396972657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,2,power_law_1.2,11.970649719238281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,2,power_law_1.2,13.600845336914062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,2,power_law_1.2,0.0906495988368988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,2,power_law_1.2,20.379808044433595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,2,power_law_1.2,0.12477439641952515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,2,power_law_1.2,0.12622079849243165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,2,power_law_1.2,0.15111039876937865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,2,power_law_1.2,0.22685439586639405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,2,power_law_1.2,0.3062592029571533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,2,power_law_1.2,0.3578111886978149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,2,power_law_1.2,0.5147647857666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,2,power_law_1.2,0.5163584232330323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,2,power_law_1.2,0.5583168029785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,2,power_law_1.2,0.5712704181671142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,2,power_law_1.2,0.5957119941711426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,2,power_law_1.2,0.6206975936889648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,2,power_law_1.2,0.6471936225891113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,2,power_law_1.2,0.6822144031524658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,2,power_law_1.2,0.6990464210510254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,2,power_law_1.2,0.7149055957794189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,2,power_law_1.2,0.7863232135772705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,2,power_law_1.2,0.8398719787597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,2,power_law_1.2,0.9608768463134766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,2,power_law_1.2,1.04017915725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,2,power_law_1.2,1.3071167945861817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,2,power_law_1.2,1.3951295852661132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,2,power_law_1.2,1.7628351211547852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,2,power_law_1.2,2.23819522857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,2,power_law_1.2,2.874496078491211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,2,power_law_1.2,3.723923110961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,2,power_law_1.2,4.148012924194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,2,power_law_1.2,6.495378875732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,2,power_law_1.2,12.848019409179688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,1,balanced,0.09481066465377808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,1,balanced,0.0727946658929189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,1,balanced,0.14648000399271646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,1,balanced,0.24147733052571616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,1,balanced,0.44277334213256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,1,balanced,0.09751466910044353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,1,balanced,0.8445866902669271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,1,balanced,0.14987732966740927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,1,balanced,1.621114730834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,1,balanced,0.25192532936731976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,1,balanced,2.4064906438191733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,1,balanced,0.4493653376897176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,1,balanced,2.412597338358561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,1,balanced,0.8456640243530273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,1,balanced,2.413424015045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,1,balanced,1.2402079900105794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,1,balanced,2.417893409729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,1,balanced,1.2426400184631348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,1,balanced,2.4232053756713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,1,balanced,1.2469759782155354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,1,balanced,2.439141273498535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,1,balanced,1.2495360374450684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,1,balanced,2.447322686513265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,1,balanced,1.2653760115305583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,1,balanced,2.457359949747721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,1,balanced,1.2706773281097412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,1,balanced,2.4758079846700034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,1,balanced,1.2767679691314697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,1,balanced,2.483738740285238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,1,balanced,1.290783961613973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,1,balanced,2.5081547101338706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,1,balanced,1.301642656326294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,1,balanced,2.5535839398701987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,1,balanced,1.3139039675394695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,1,balanced,1.3449920018513997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,1,balanced,2.61024538675944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,1,balanced,1.392581303914388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,1,balanced,2.729152043660482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,1,balanced,1.4465440114339192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,1,balanced,2.8043092091878257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,1,balanced,1.5510986646016438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,1,balanced,1.6132799784342449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,1,balanced,3.003413200378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,1,balanced,1.824885368347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,1,balanced,3.1632585525512695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,1,balanced,1.9747360547383626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,1,balanced,3.564901351928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,1,balanced,0.09894933303197224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,1,power_law_1.01,0.0939903974533081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,1,balanced,2.355818589528402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,1,balanced,0.11755200227101643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,1,balanced,0.15105066696802774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,1,balanced,0.21979733308156332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,1,power_law_1.01,0.14392319917678834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,1,balanced,0.3491520086924235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,1,balanced,0.5973546504974365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,1,power_law_1.01,0.23941760063171386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,1,balanced,4.026186625162761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,1,balanced,0.8533066908518473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,1,power_law_1.01,0.31896960735321045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,1,balanced,0.8613920211791992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,1,balanced,0.8666186332702637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,1,balanced,2.8607571919759116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,1,power_law_1.01,0.6279232025146484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,1,balanced,0.8712586561838785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,1,power_law_1.01,0.9370176315307617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,1,balanced,0.878986676534017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,1,power_law_1.01,0.07340800166130065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,1,balanced,0.8898506959279379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,1,power_law_1.01,1.1115391731262207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,1,balanced,0.9127946694691976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,1,power_law_1.01,0.09744639992713929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,1,power_law_1.01,1.4615743637084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,1,balanced,0.9003573258717855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,1,power_law_1.01,0.1464319944381714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,1,balanced,0.9116213321685791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,1,power_law_1.01,1.558572769165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,1,balanced,4.63106123606364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,1,balanced,0.938368002573649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,1,power_law_1.01,0.18567039966583251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,1,power_law_1.01,1.5908672332763671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,1,balanced,0.9605173269907633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,1,power_law_1.01,0.34615681171417234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,1,power_law_1.01,1.6523775100708007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,1,balanced,3.568981488545736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,1,balanced,1.011402686436971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,1,power_law_1.01,0.492460823059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,1,power_law_1.01,1.7589632034301759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,1,balanced,1.0635039806365967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,1,power_law_1.01,1.8349824905395509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,1,power_law_1.01,0.5858560085296631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,1,balanced,1.1767573356628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,1,power_law_1.01,0.7715968132019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,1,power_law_1.01,1.863577651977539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,1,power_law_1.01,0.8308863639831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,1,power_law_1.01,1.9796096801757812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,1,balanced,1.23853333791097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,1,power_law_1.01,0.8551168441772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,1,power_law_1.01,1.977440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,1,balanced,6.159247716267903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,1,power_law_1.01,0.8973631858825684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,1,power_law_1.01,0.09063040018081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,1,power_law_1.01,2.1089344024658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,1,power_law_1.01,0.9352831840515137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,1,balanced,1.4204479853312175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,1,power_law_1.01,0.1133247971534729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,1,power_law_1.01,2.2084800720214846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,1,power_law_1.01,0.9452863693237304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,1,balanced,4.867541313171387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,1,power_law_1.01,0.14638079404830934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,1,power_law_1.01,1.00067195892334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,1,power_law_1.2,0.09484800100326538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,1,power_law_1.01,2.3884607315063477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,1,power_law_1.01,0.17167359590530396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,1,power_law_1.01,1.0339584350585938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,1,balanced,1.6079519589742024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,1,power_law_1.01,0.27077760696411135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,1,power_law_1.01,2.6841407775878907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,1,power_law_1.2,0.14439680576324462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,1,power_law_1.01,1.0713983535766602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,1,power_law_1.01,0.3864703893661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,1,power_law_1.01,2.6213823318481446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,1,power_law_1.2,0.20032639503479005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,1,power_law_1.01,1.1111807823181152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,1,power_law_1.01,0.4430399894714355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,1,power_law_1.2,0.32232959270477296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,1,power_law_1.01,2.9870847702026366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,1,power_law_1.01,1.1996671676635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,1,power_law_1.01,0.5523776054382324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,1,power_law_1.2,0.5364031791687012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,1,power_law_1.01,3.2653377532958983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,1,power_law_1.01,0.5773759841918945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,1,power_law_1.01,1.337343978881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,1,balanced,1.9805173873901367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,1,power_law_1.2,0.779033613204956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,1,power_law_1.01,0.6079616069793701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,1,power_law_1.01,3.841126251220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,1,power_law_1.01,1.5492480278015137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,1,power_law_1.2,0.9357119560241699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,1,power_law_1.01,0.644761610031128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,1,power_law_1.01,1.5964863777160645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,1,power_law_1.01,4.352364730834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,1,power_law_1.2,1.3716608047485352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,1,power_law_1.01,0.6670207977294922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,1,balanced,7.074394861857097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,1,power_law_1.01,1.9351680755615235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,1,power_law_1.2,1.4744000434875488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,1,power_law_1.01,5.428672027587891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,1,power_law_1.01,0.6906176090240479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,1,power_law_1.01,2.020102310180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,1,power_law_1.2,1.5908672332763671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,1,power_law_1.01,0.7151040077209473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,1,balanced,2.5983947118123374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,1,balanced,6.212362925211589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,1,power_law_1.2,1.6602176666259765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,1,power_law_1.01,2.5259328842163087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,1,power_law_1.01,7.463603210449219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,1,power_law_1.01,0.7518720149993896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,1,power_law_1.2,1.7134719848632813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,1,power_law_1.01,3.1107200622558593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,1,power_law_1.01,0.7700928211212158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,1,power_law_1.01,8.671526336669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,1,power_law_1.2,1.7814144134521483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,1,power_law_1.01,0.8080960273742676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,1,power_law_1.01,4.053900909423828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,1,power_law_1.01,0.8872960090637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,1,power_law_1.2,1.8878847122192384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,1,power_law_1.01,5.115801620483398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,1,power_law_1.01,11.183647918701173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.01,0.9657728195190429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,1,power_law_1.2,1.9851392745971679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,1,power_law_1.01,6.210950469970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.01,1.1239808082580567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,1,power_law_1.2,2.0331584930419924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,1,balanced,3.346554756164551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.01,1.1915264129638672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,1,power_law_1.2,2.0983039855957033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,1,power_law_1.01,9.188243103027343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,1,power_law_1.01,21.468704223632812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.01,1.4539135932922362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,1,power_law_1.2,2.2679231643676756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.01,1.6533952713012696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,1,power_law_1.2,2.503545570373535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.01,2.141881561279297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,1,power_law_1.01,17.441786193847655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,1,power_law_1.2,2.848678398132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,1,power_law_1.01,2.794175910949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,1,power_law_1.2,2.823967933654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,1,power_law_1.01,3.716185760498047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,1,power_law_1.01,4.559942245483398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,1,power_law_1.2,3.271129608154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,1,balanced,10.39031982421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,1,power_law_1.2,3.4737728118896483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,1,balanced,4.554586728413899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,1,power_law_1.01,5.470662307739258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,1,power_law_1.2,4.152691268920899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,1,balanced,8.724186579386393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,1,power_law_1.01,8.3169921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,1,power_law_1.2,4.671756744384766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,1,power_law_1.2,5.835686492919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,1,power_law_1.01,15.630213928222656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,1,power_law_1.2,7.908589172363281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,1,power_law_1.2,9.047014617919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,1,power_law_1.2,0.07044479846954346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,1,power_law_1.2,11.632927703857423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,1,balanced,5.730202356974284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,1,power_law_1.2,0.09482240080833435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,1,power_law_1.2,21.988479614257812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,1,power_law_1.2,0.11640959978103638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,1,power_law_1.2,0.1865664005279541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,1,power_law_1.2,0.29870080947875977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,1,power_law_1.2,0.4361599922180176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,1,power_law_1.2,0.5183872222900391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,1,power_law_1.2,0.733190393447876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,1,power_law_1.2,0.7863167762756348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,1,power_law_1.2,0.8276224136352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,1,power_law_1.2,0.8910847663879394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,1,power_law_1.2,0.9146368026733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,1,power_law_1.2,0.9540736198425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,1,power_law_1.2,0.976921558380127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,1,balanced,8.387253443400065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,1,power_law_1.2,1.0316096305847169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,1,power_law_1.2,1.0612735748291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,1,power_law_1.2,1.1225088119506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,1,power_law_1.2,1.2248191833496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,1,balanced,20.13707224527995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,1,power_law_1.2,1.3882687568664551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,1,power_law_1.2,1.5786815643310548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,1,balanced,16.617407480875652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,1,power_law_1.2,1.6743616104125976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,1,power_law_1.2,2.0442623138427733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,1,power_law_1.2,2.129318428039551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,1,power_law_1.2,2.653561592102051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,1,power_law_1.2,3.2468288421630858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,1,power_law_1.2,4.189785766601562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,1,power_law_1.2,5.29356803894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,1,power_law_1.2,6.360697555541992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,1,power_law_1.2,9.464345550537109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,1,power_law_1.2,17.404115295410158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,1,balanced,15.837781270345053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,4,balanced,0.0355679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,1,power_law_1.2,0.0927232027053833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,4,balanced,0.05407999952634176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,4,balanced,0.07543466488520305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,1,power_law_1.2,0.1142016053199768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,4,balanced,0.12240533034006755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,4,balanced,0.2053333322207133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,1,power_law_1.2,0.1294592022895813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,4,balanced,0.20900267362594604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,1,power_law_1.2,0.17667200565338134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,4,balanced,0.2104746699333191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,1,power_law_1.2,0.2557248115539551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,4,balanced,0.20803199211756387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,4,balanced,0.20772800842920938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,1,power_law_1.2,0.34231679439544677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,4,balanced,0.2079306642214457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,1,power_law_1.2,0.4022655963897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,4,balanced,0.20944533745447794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,4,balanced,0.03305066625277201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,1,power_law_1.2,0.5337855815887451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,4,balanced,0.2109546661376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,4,balanced,0.21448000272115073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,1,power_law_1.2,0.5761280059814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,4,balanced,0.033530667424201965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,4,balanced,0.2132479945818583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,4,balanced,0.051962668697039284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,1,power_law_1.2,0.5862656116485596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,4,balanced,0.07865599791208903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,4,balanced,0.22082666556040445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,1,power_law_1.2,0.6165952205657959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,4,balanced,0.12076266606648763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,4,balanced,0.22308266162872314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,4,balanced,0.12378133336702983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,1,power_law_1.2,0.6471424102783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,4,balanced,0.12532800436019897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,4,balanced,0.2264159917831421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,1,power_law_1.2,0.6787456035614013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,4,balanced,0.123471995194753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,1,power_law_1.2,0.6948095798492432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,4,balanced,0.2325013279914856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,4,balanced,0.12442666292190552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,1,power_law_1.2,0.7336512088775635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,4,balanced,0.12587733070055643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,4,balanced,0.24102399746576944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,1,power_law_1.2,0.7679999828338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,4,balanced,0.12570666273434958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,1,power_law_1.2,0.8351424217224122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,4,balanced,0.03839466720819473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,4,balanced,0.12913599610328674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,4,balanced,0.28169600168863934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,4,balanced,0.1293280025323232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,1,power_law_1.2,0.9164863586425781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,4,balanced,0.039621333281199135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,4,balanced,0.13156267007191977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.2,1.013907241821289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,4,balanced,0.045882667104403176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,4,balanced,0.1402133305867513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,4,balanced,0.28014934062957764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.2,1.1706111907958985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,4,balanced,0.069882666071256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,4,balanced,0.14004266262054443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,4,balanced,0.09748799602190654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.2,1.2466112136840821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,4,balanced,0.09877866506576538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,4,balanced,0.14510933558146158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.2,1.5095232009887696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,4,balanced,0.09961066643397014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,4,balanced,0.3366080125172933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,4,balanced,0.09937066833178203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,4,balanced,0.15606932838757834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.2,1.7274879455566405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,4,balanced,0.10097066561381023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.2,2.2208255767822265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,4,balanced,0.10222933689753215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,4,balanced,0.16938134034474692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,4,balanced,0.10156800349553426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,1,power_law_1.2,2.849772834777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,4,balanced,0.35919467608133954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,4,balanced,0.10437333583831787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,1,power_law_1.2,3.809811019897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,4,balanced,0.19368533293406168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,4,balanced,0.10475732882817586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,1,power_law_1.2,4.612883377075195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,4,balanced,0.10718400279680888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,4,balanced,0.11576533317565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,4,balanced,0.2098133365313212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,1,power_law_1.2,5.600665664672851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,4,balanced,0.11599999666213989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,4,balanced,0.47145601113637287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,4,balanced,0.12143466869990031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,1,power_law_1.2,8.464832305908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,4,balanced,0.26344533761342365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,4,balanced,0.12915733456611633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,4,balanced,0.14353600144386292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,4,balanced,0.025231999655564625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,1,power_law_1.2,16.095858764648437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,4,balanced,0.15930666526158652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,4,balanced,0.02716800073782603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,4,balanced,0.3054080009460449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,4,balanced,0.03540800015131632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,4,balanced,0.5305866797765096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,4,balanced,0.04970666766166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,4,balanced,0.17874133586883545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,4,balanced,0.07057066758473714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,4,balanced,0.07539199789365132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,4,balanced,0.0772213339805603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,4,balanced,0.07912533481915791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,4,balanced,0.23573867479960123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,4,balanced,0.42236268520355225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,4,balanced,0.09895466764767964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,4,balanced,0.09284266829490662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,4,balanced,0.09479999542236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,4,balanced,0.1442506710688273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,4,balanced,0.13059733311335245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,4,balanced,0.13557866215705872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,4,balanced,0.2696959972381592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,4,balanced,0.1627893348534902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,4,balanced,0.7198987007141113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,4,balanced,0.1662986675898234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,4,power_law_1.01,0.07649919986724854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,4,balanced,0.49820268154144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,4,balanced,0.18341867129007974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,4,power_law_1.01,0.05626239776611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,4,balanced,0.18832000096638998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,4,balanced,0.36128000418345135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,4,power_law_1.01,0.07198079824447631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,4,balanced,0.19812800486882529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,4,power_law_1.01,0.10375679731369018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,4,power_law_1.01,0.1165503978729248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,4,balanced,0.21285865704218546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,4,power_law_1.01,0.18858879804611206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,4,power_law_1.01,0.18994560241699218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,4,balanced,0.228602667649587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,4,power_law_1.01,0.1937343955039978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,4,power_law_1.01,0.1994752049446106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,4,balanced,0.4543306827545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,4,balanced,0.6980906327565511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,4,power_law_1.01,0.1944640040397644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,4,balanced,0.23988799254099527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,4,power_law_1.01,0.19965440034866333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,4,balanced,0.9100053310394287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,4,power_law_1.01,0.19569920301437377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,4,power_law_1.01,0.2059839963912964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,4,power_law_1.01,0.21416959762573243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,4,power_law_1.01,0.23019518852233886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,4,balanced,0.262442668279012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,4,power_law_1.01,0.23549439907073974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,4,power_law_1.01,0.23883519172668458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,4,power_law_1.01,0.26328320503234864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,4,power_law_1.01,0.051737600564956666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,4,power_law_1.01,0.2841536045074463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,4,balanced,0.6395466725031534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,4,power_law_1.01,0.3376895904541016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,4,power_law_1.01,0.047225600481033324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,4,balanced,0.4322773218154907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,4,power_law_1.01,0.3308288097381592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,4,power_law_1.01,0.0493120014667511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,4,balanced,0.9011840025583903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,4,power_law_1.01,0.06783360242843628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,4,power_law_1.01,0.40752639770507815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,4,power_law_1.01,0.07719039916992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,4,power_law_1.01,0.11056640148162841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,4,power_law_1.01,0.45182080268859864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,4,power_law_1.01,0.11423360109329224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,4,power_law_1.01,0.11752959489822387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,4,power_law_1.01,0.6456704139709473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,4,power_law_1.01,0.11253119707107544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,4,balanced,1.1011786460876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,4,balanced,0.4664906660715739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,4,power_law_1.01,0.1174847960472107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,4,power_law_1.01,0.7277952194213867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,4,power_law_1.01,0.12201600074768067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,4,power_law_1.01,0.04647679924964905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,4,power_law_1.01,0.1256127953529358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,4,power_law_1.01,0.9729984283447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,4,power_law_1.01,0.12949119806289672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,4,balanced,0.8188373247782389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,4,power_law_1.01,0.040268799662590025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,4,power_law_1.01,0.13485440015792846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,4,power_law_1.01,0.04917120039463043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,4,power_law_1.01,0.14083199501037597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,4,power_law_1.01,1.3053248405456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,4,power_law_1.01,0.06554880142211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,4,power_law_1.01,0.14676480293273925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,4,power_law_1.01,0.07089279890060425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,4,power_law_1.01,0.15269119739532472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,4,power_law_1.01,1.5004159927368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,4,power_law_1.01,0.08956800103187561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,4,balanced,1.1033066908518474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,4,balanced,0.6725866794586182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,4,power_law_1.01,0.16855039596557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,4,power_law_1.01,0.09231359958648681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,4,power_law_1.01,0.18558080196380616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,4,power_law_1.01,2.408185577392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,4,power_law_1.01,0.09091839790344239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,4,power_law_1.01,0.22498559951782227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,4,power_law_1.01,0.09275519847869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,4,power_law_1.01,0.2542399883270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,4,power_law_1.01,0.0961471974849701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,4,power_law_1.01,0.3282815933227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,4,power_law_1.01,4.529740905761718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,4,power_law_1.01,0.09797120094299316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,4,power_law_1.01,0.3877824068069458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,4,power_law_1.01,0.10095360279083251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,4,power_law_1.01,0.10185600519180298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,4,power_law_1.01,0.5166528224945068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,4,power_law_1.01,0.10913920402526855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,4,balanced,1.0025546550750732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,4,power_law_1.01,0.6590400218963623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,4,power_law_1.01,0.11535359621047973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,4,power_law_1.01,0.11900800466537476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,4,power_law_1.01,0.8964287757873535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,4,balanced,0.8820959726969401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,4,power_law_1.01,0.12747520208358765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,4,power_law_1.01,1.178547191619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,4,balanced,1.7001226743062336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,4,power_law_1.01,0.1450368046760559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.01,0.15741440057754516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,4,power_law_1.01,1.3473216056823731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.01,0.1910912036895752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.01,0.22121601104736327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,4,power_law_1.01,2.1948352813720704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.01,0.2780479907989502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.01,0.3334399938583374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,4,power_law_1.01,4.374534225463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.01,0.4809279918670654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.01,0.6086463928222656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,4,balanced,1.708405335744222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.01,0.8311615943908691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.01,1.0879167556762694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,4,power_law_1.01,1.300102424621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,4,balanced,1.0961973667144775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,4,power_law_1.01,2.149196815490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,4,power_law_1.01,4.315833663940429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,4,balanced,1.5550452868143718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,4,balanced,1.7358026504516602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,4,balanced,3.306133270263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,4,balanced,3.3366400400797525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,4,power_law_1.01,0.026495999097824095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,4,power_law_1.01,0.02781440019607544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,4,balanced,3.0254185994466147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,4,power_law_1.01,0.03248000144958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,4,power_law_1.01,0.04539520144462585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,4,power_law_1.01,0.048511999845504764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,4,power_law_1.01,0.0632960021495819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,4,power_law_1.01,0.06949759721755981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,4,power_law_1.01,0.07032960057258605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,4,power_law_1.01,0.077920001745224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,4,power_law_1.01,0.07438719868659974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,4,power_law_1.01,0.07370880246162415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,4,power_law_1.01,0.1028607964515686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,4,power_law_1.01,0.10197759866714477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,4,power_law_1.01,0.08954240083694458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,4,power_law_1.01,0.14272639751434327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,4,power_law_1.01,0.14004479646682738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,4,balanced,3.4389546712239585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,4,power_law_1.01,0.16327680349349977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,4,power_law_1.01,0.13455359935760497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.01,0.12861440181732178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.01,0.1469375967979431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.01,0.17036160230636596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.01,0.19562239646911622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.01,0.2352384090423584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.01,0.30788478851318357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.01,0.4026303768157959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.01,0.5441855907440185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.01,0.7014912128448486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,4,power_law_1.01,0.8560383796691895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,4,power_law_1.01,1.3637951850891112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,4,power_law_1.01,2.8096639633178713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,4,power_law_1.2,0.0755840003490448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,4,power_law_1.2,0.053548800945281985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,4,power_law_1.2,0.06646400094032287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,4,power_law_1.2,0.0913536012172699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,4,power_law_1.2,0.11886080503463745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,4,power_law_1.2,0.05115519762039185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,4,power_law_1.2,0.17088639736175537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,4,power_law_1.2,0.1883903980255127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,4,power_law_1.2,0.03893760144710541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,4,power_law_1.2,0.19500800371170043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,4,power_law_1.2,0.1966655969619751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,4,power_law_1.2,0.05173119902610779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,4,power_law_1.2,0.19129600524902343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,4,power_law_1.2,0.060249602794647215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,4,power_law_1.2,0.20821120738983154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,4,power_law_1.2,0.07187839746475219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,4,power_law_1.2,0.20343680381774903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,4,power_law_1.2,0.10855679512023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,4,power_law_1.2,0.19452799558639527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,4,power_law_1.2,0.11430399417877198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,4,power_law_1.2,0.22171521186828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,4,power_law_1.2,0.1173248052597046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,4,power_law_1.2,0.23328640460968017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,4,power_law_1.2,0.11817599534988403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,4,power_law_1.2,0.04614399969577789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,4,power_law_1.2,0.2452415943145752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,4,power_law_1.2,0.12021759748458863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,4,power_law_1.2,0.24696319103240966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,4,power_law_1.2,0.12515840530395508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,4,power_law_1.2,0.041382399201393125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,4,power_law_1.2,0.2766207933425903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,4,power_law_1.2,0.12693120241165162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,4,power_law_1.2,0.29748480319976806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,4,power_law_1.2,0.04702079892158508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,4,power_law_1.2,0.12942719459533691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,4,power_law_1.2,0.35976319313049315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,4,power_law_1.2,0.055795198678970336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,4,power_law_1.2,0.1398911952972412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,4,power_law_1.2,0.06951040029525757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,4,power_law_1.2,0.35187199115753176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,4,power_law_1.2,0.14920320510864257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,4,power_law_1.2,0.08888319730758668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,4,power_law_1.2,0.09068160057067871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,4,power_law_1.2,0.15355520248413085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,4,power_law_1.2,0.41396479606628417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,4,power_law_1.2,0.09560319781303406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,4,power_law_1.2,0.1513152003288269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,4,power_law_1.2,0.09774720072746276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,4,power_law_1.2,0.5533887863159179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,4,power_law_1.2,0.18279680013656616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,4,power_law_1.2,0.09818879961967468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,4,power_law_1.2,0.652243185043335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,4,power_law_1.2,0.09836800098419189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,4,power_law_1.2,0.20064001083374022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,4,power_law_1.2,0.9243647575378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,4,power_law_1.2,0.10157439708709717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,4,power_law_1.2,0.2412031888961792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,4,power_law_1.2,0.10685440301895141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,4,power_law_1.2,1.2044032096862793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,4,power_law_1.2,0.25284481048583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,4,power_law_1.2,0.10953600406646728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,4,power_law_1.2,0.3314496040344238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,4,power_law_1.2,1.617804718017578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,4,power_law_1.2,0.11683839559555054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,4,power_law_1.2,0.427507209777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,4,power_law_1.2,0.12364799976348877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,4,power_law_1.2,2.0320959091186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,4,power_law_1.2,0.5582655906677246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,4,power_law_1.2,0.13103359937667847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,4,power_law_1.2,0.1500480055809021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,4,power_law_1.2,0.6757952213287354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,4,power_law_1.2,3.132736015319824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.2,0.16316159963607788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,4,power_law_1.2,0.02595199942588806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,4,power_law_1.2,1.0992959976196288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.2,0.19208320379257202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,4,power_law_1.2,0.023897600173950196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.2,0.2172032117843628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,4,power_law_1.2,1.4340800285339355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,4,power_law_1.2,5.393644714355469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,4,power_law_1.2,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.2,0.3077440023422241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,4,power_law_1.2,1.7976512908935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,4,power_law_1.2,0.03919360041618347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.2,0.372761607170105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,4,power_law_1.2,0.04774399995803833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.2,0.5814976215362548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,4,power_law_1.2,0.06720640063285828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,4,power_law_1.2,2.532492828369141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.2,0.6613823890686035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,4,power_law_1.2,0.06844159960746765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,4,power_law_1.2,0.06926079988479614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.2,0.8308480262756348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,4,power_law_1.2,0.07314559817314148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,4,power_law_1.2,5.418086242675781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.2,1.4720319747924804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,4,power_law_1.2,0.0714303970336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,4,power_law_1.2,0.0744383990764618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,4,power_law_1.2,1.412825584411621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,2,balanced,0.035946667194366455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,4,power_law_1.2,0.0979200005531311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,4,power_law_1.2,0.09278720021247863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,2,balanced,0.05378133555253347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,4,power_law_1.2,2.8832895278930666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,4,power_law_1.2,0.08819839954376221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,2,balanced,0.07600000003973643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,4,power_law_1.2,0.14540799856185913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,2,balanced,0.12086400389671326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,4,power_law_1.2,0.14542720317840577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,4,power_law_1.2,5.177862548828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,2,balanced,0.2048906683921814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,2,balanced,0.20602667331695557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,4,power_law_1.2,0.15700479745864868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,2,balanced,0.20765332380930582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,4,power_law_1.2,0.13358720541000366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,2,balanced,0.20865066846211752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,2,balanced,0.033887999753157295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.2,0.14312319755554198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,2,balanced,0.20811200141906738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.2,0.14445439577102662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,2,balanced,0.035530666510264076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,2,balanced,0.20949333906173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.2,0.16324479579925538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,2,balanced,0.2124533255894979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,2,balanced,0.05455466608206431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.2,0.2005631923675537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,2,balanced,0.21437333027521768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,2,balanced,0.08281066517035167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,2,balanced,0.2178879976272583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.2,0.23625600337982178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,2,balanced,0.12962133685747781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,2,balanced,0.21790399154027304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,2,balanced,0.12691733241081238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,2,balanced,0.2252053419748942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.2,0.2942528009414673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,2,balanced,0.12972799936930338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,2,balanced,0.2300106684366862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.2,0.4171584129333496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,2,balanced,0.12981333335240683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,2,balanced,0.2352799971898397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,2,balanced,0.13329600294431052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,2,balanced,0.13512532909711203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.2,0.5977280139923096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,2,balanced,0.24752533435821533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,2,balanced,0.13501333196957907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,2,balanced,0.13782933354377747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.2,0.7609087944030761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,2,balanced,0.2629280090332031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,2,balanced,0.13979199528694153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,4,power_law_1.2,0.8952832221984863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,2,balanced,0.1432213286558787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,2,balanced,0.30450665950775146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,2,balanced,0.15148799618085226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,4,power_law_1.2,1.4745920181274415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,2,balanced,0.1546293298403422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,2,balanced,0.3250826597213745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,2,balanced,0.16293866435686746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,4,power_law_1.2,2.754355239868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,2,balanced,0.18290666739145914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,2,balanced,0.19814932346343994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,2,balanced,0.39964266618092853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,2,balanced,0.05230399966239929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,2,balanced,0.23679467042287192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,2,balanced,0.06222933530807495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,2,balanced,0.44308801492055255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,2,balanced,0.267685333887736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,2,balanced,0.07912533481915791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,2,balanced,0.11992533008257548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,2,balanced,0.18918933471043906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,2,balanced,0.17108267545700073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,2,balanced,0.3459786574045817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,2,balanced,0.17203734318415323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,2,balanced,0.16597867012023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,2,balanced,0.6049600044886271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,2,balanced,0.16471466422080994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,2,balanced,0.17008533080418906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,2,balanced,0.4129226605097453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,2,balanced,0.16032000382741293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,2,balanced,0.16451199849446616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,2,balanced,0.1678346594174703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,2,balanced,0.16220266620318094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,2,balanced,0.18178133169809976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,2,balanced,0.5832533439000448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,2,balanced,0.7018933296203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,2,balanced,0.17634665966033936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,2,balanced,0.17892799774805704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,2,balanced,0.18959999084472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,2,balanced,0.20525866746902466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,2,balanced,0.02288000037272771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,2,balanced,0.2268213431040446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,2,balanced,0.7293333212534586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,2,balanced,0.02515200028816859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,2,balanced,0.03169066707293192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,2,balanced,0.251802663008372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,2,balanced,0.04972266654173533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,2,balanced,0.07318399846553802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,2,balanced,0.9760053157806396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,2,balanced,0.0734986662864685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,2,balanced,0.07704533139864604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,2,balanced,0.3877760171890259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,2,balanced,0.08080533146858215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,2,balanced,0.08372799555460612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,2,balanced,0.08748799562454224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,2,balanced,0.08617066343625386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,2,balanced,0.439248005549113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,2,balanced,0.09081066648165385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,2,balanced,1.040725310643514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,2,balanced,0.09645866354306538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,2,balanced,0.0906880001227061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,2,balanced,0.09683199723561604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,2,balanced,0.10327999790509541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,2,balanced,0.6119999885559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,2,balanced,1.2694453398386638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,2,balanced,0.11497599879900615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,2,balanced,0.1306666632493337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,2,balanced,0.1507200002670288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,2,balanced,1.3521493275960286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,2,balanced,0.77292799949646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,2,balanced,0.22457599639892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,2,balanced,0.21732799212137857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,2,balanced,0.21524266401926676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,2,balanced,1.5482293764750164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,2,balanced,0.2238933245340983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,2,balanced,1.1117599805196126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,2,balanced,0.38626666863759357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,2,balanced,1.6762453715006511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,2,balanced,0.39405866463979083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,2,balanced,1.4341707229614258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,2,balanced,0.5670773188273112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,2,balanced,2.4222346941630044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,2,balanced,1.7629547119140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,2,balanced,2.542202631632487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,2,balanced,0.7430453300476074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,2,power_law_1.01,0.05268480181694031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,2,power_law_1.01,0.051744002103805545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,2,power_law_1.01,0.03509120047092438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,2,power_law_1.01,0.05873280167579651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,2,power_law_1.01,0.0985472023487091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,2,power_law_1.01,0.03687039911746979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,2,power_law_1.01,0.11470719575881957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,2,power_law_1.01,0.04604800045490265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,2,power_law_1.01,0.17017600536346436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,2,power_law_1.01,0.18862719535827638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,2,power_law_1.01,0.06616960167884826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,2,power_law_1.01,0.19058560132980346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,2,power_law_1.01,0.07704319953918456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,2,power_law_1.01,0.18859519958496093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,2,power_law_1.01,0.11331199407577515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,2,power_law_1.01,0.19082880020141602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,2,power_law_1.01,0.11741440296173096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,2,power_law_1.01,0.1984063982963562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,2,power_law_1.01,0.11927679777145386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,2,power_law_1.01,0.204915189743042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,2,power_law_1.01,0.05987200140953064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,2,power_law_1.01,0.1245568037033081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,2,power_law_1.01,0.20977919101715087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,2,power_law_1.01,0.12424960136413574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,2,power_law_1.01,0.22013440132141113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,2,power_law_1.01,0.05626879930496216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,2,balanced,0.9195146560668945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,2,power_law_1.01,0.1280895948410034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,2,power_law_1.01,0.22652161121368408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,2,power_law_1.01,0.0639360010623932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,2,power_law_1.01,0.13378560543060303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,2,balanced,2.7494560877482095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,2,power_law_1.01,0.23736960887908937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,2,power_law_1.01,0.23943040370941163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,2,power_law_1.01,0.09162880182266235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,2,power_law_1.01,0.13192960023880004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,2,power_law_1.01,0.25489919185638427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,2,power_law_1.01,0.14199680089950562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,2,power_law_1.01,0.10802559852600098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,2,power_law_1.01,0.28533759117126467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,2,power_law_1.01,0.3368704080581665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,2,power_law_1.01,0.1471935987472534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,2,power_law_1.01,0.1541632056236267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,2,power_law_1.01,0.3573760032653809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,2,power_law_1.01,0.15660799741744996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,2,power_law_1.01,0.1520640015602112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,2,power_law_1.01,0.16768640279769897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,2,power_law_1.01,0.15724159479141236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,2,power_law_1.01,0.4389887809753418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,2,power_law_1.01,0.16200319528579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,2,power_law_1.01,0.1985983967781067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,2,power_law_1.01,0.5129280090332031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,2,power_law_1.01,0.1620352029800415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,2,power_law_1.01,0.2118272066116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,2,power_law_1.01,0.6757823944091796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,2,power_law_1.01,0.02129279971122742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,2,power_law_1.01,0.16629120111465454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,2,power_law_1.01,0.25982720851898194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,2,power_law_1.01,0.8224384307861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,2,power_law_1.01,0.021913599967956544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,2,power_law_1.01,0.1667456030845642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,2,power_law_1.01,0.28703999519348145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,2,balanced,4.737397193908691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,2,power_law_1.01,0.028966400027275085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,2,power_law_1.01,1.178598403930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,2,power_law_1.01,0.17326719760894777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,2,power_law_1.01,0.39844479560852053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,2,power_law_1.01,0.04048640131950378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,2,power_law_1.01,0.17171839475631714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,2,power_law_1.01,1.48024320602417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,2,power_law_1.01,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,2,power_law_1.01,0.47360639572143554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,2,power_law_1.01,0.17749760150909424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,2,power_law_1.01,0.06109439730644226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,2,power_law_1.01,0.6410304069519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,2,power_law_1.01,1.8425024032592774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,2,balanced,1.4493227005004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,2,power_law_1.01,0.1811776041984558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,2,power_law_1.01,0.06481919884681701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,2,power_law_1.01,0.8337344169616699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,2,power_law_1.01,0.19586559534072875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,2,power_law_1.01,0.06789759993553161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,2,power_law_1.01,2.6799295425415037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,2,power_law_1.01,0.21484160423278809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,2,power_law_1.01,1.197209644317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,2,power_law_1.01,0.07117440104484558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.01,0.2340928077697754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,2,balanced,4.962965329488118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,2,power_law_1.01,0.07393919825553893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,2,power_law_1.01,1.4881855964660644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.01,0.28840959072113037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,2,power_law_1.01,5.358643341064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,2,power_law_1.01,0.07383040189743043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.01,0.32881920337677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,2,power_law_1.01,1.9336639404296876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,2,power_law_1.01,0.07623040080070495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.01,0.4342336177825928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,2,power_law_1.01,0.08001919984817504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.01,0.5216703891754151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,2,power_law_1.01,2.8082176208496095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,2,power_law_1.01,0.07791360020637512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.01,0.7286079883575439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,2,power_law_1.01,0.08128640055656433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.01,0.8982015609741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,2,power_law_1.01,5.232083129882812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,2,power_law_1.01,0.08490880131721497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.01,1.2829952239990234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,2,power_law_1.01,0.1028480052947998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.01,1.5697983741760253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,2,power_law_1.01,0.11792639493942261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,2,power_law_1.01,2.083180809020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.01,0.14346879720687866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.01,0.17957760095596315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,2,power_law_1.01,3.0458879470825195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.01,0.15786240100860596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.01,0.1814463973045349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.01,0.21770238876342773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,2,power_law_1.01,6.1119743347167965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.01,0.2809983968734741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.01,0.34045441150665284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.01,0.4762239933013916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.01,0.6218048095703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,2,power_law_1.01,0.7258048057556152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,2,power_law_1.01,1.2122495651245118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,2,balanced,5.357413609822591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,2,power_law_1.01,2.184147262573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,2,balanced,2.851594607035319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,2,power_law_1.2,0.05075200200080872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,2,power_law_1.2,0.0431551992893219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,2,power_law_1.2,0.06647040247917176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,2,power_law_1.2,0.08885759711265565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,2,power_law_1.2,0.10678399801254272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,2,power_law_1.2,0.17249280214309692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,2,power_law_1.2,0.1891136050224304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,2,power_law_1.2,0.18691200017929077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,2,power_law_1.2,0.189792001247406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,2,power_law_1.2,0.2008512020111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,2,power_law_1.2,0.0357120007276535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,2,power_law_1.2,0.19415040016174318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,2,power_law_1.2,0.061919999122619626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,2,power_law_1.2,0.20145280361175538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,2,power_law_1.2,0.03653759956359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,2,power_law_1.2,0.21086080074310304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,2,power_law_1.2,0.04755840003490448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,2,power_law_1.2,0.057599997520446776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,2,power_law_1.2,0.22209279537200927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,2,power_law_1.2,0.06096000075340271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,2,power_law_1.2,0.0735040009021759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,2,power_law_1.2,0.06954240202903747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,2,power_law_1.2,0.23889920711517335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,2,power_law_1.2,0.10767359733581543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,2,power_law_1.2,0.24440319538116456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,2,power_law_1.2,0.08823680281639099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,2,power_law_1.2,0.11596159934997559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,2,power_law_1.2,0.2432703971862793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,2,power_law_1.2,0.0992576003074646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,2,power_law_1.2,0.12043520212173461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,2,power_law_1.2,0.2749311923980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,2,power_law_1.2,0.1512511968612671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,2,power_law_1.2,0.12535680532455445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,2,power_law_1.2,0.29160959720611573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,2,power_law_1.2,0.15563520193099975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,2,power_law_1.2,0.1263360023498535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,2,power_law_1.2,0.3532799959182739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,1,balanced,0.036101333796978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,2,power_law_1.2,0.12968319654464722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,2,power_law_1.2,0.15858559608459472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,2,power_law_1.2,0.3892607927322388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,2,power_law_1.2,0.021324799954891206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,1,balanced,0.0547680010398229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,2,power_law_1.2,0.1336575984954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,2,power_law_1.2,0.16657919883728028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,1,balanced,0.0763626645008723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,2,power_law_1.2,0.4699840068817139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,2,power_law_1.2,0.13891839981079102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,2,power_law_1.2,0.16414079666137696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,1,balanced,0.12304533521334331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,2,power_law_1.2,0.023148800432682037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,2,power_law_1.2,0.554150390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,2,power_law_1.2,0.1441472053527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,2,power_law_1.2,0.17081600427627563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,1,balanced,0.2107093334197998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,2,power_law_1.2,0.15314559936523436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,2,power_law_1.2,0.7127039909362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,2,power_law_1.2,0.17147519588470458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,2,power_law_1.2,0.027955201268196107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,1,balanced,0.2139093279838562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,2,power_law_1.2,0.16332800388336183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,2,power_law_1.2,0.16992000341415406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,2,power_law_1.2,0.8819456100463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,1,balanced,0.2156053384145101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,2,power_law_1.2,0.03777920007705689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,2,power_law_1.2,0.17515519857406617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,2,power_law_1.2,0.1752768039703369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,2,power_law_1.2,1.290214443206787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,1,balanced,0.21975467602411905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,2,power_law_1.2,0.1866047978401184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,2,power_law_1.2,0.20576000213623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,2,power_law_1.2,0.04221439957618713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,2,power_law_1.2,1.627359962463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,2,power_law_1.2,0.1774399995803833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,1,balanced,0.21733333667119345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,2,power_law_1.2,0.21468799114227294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,2,power_law_1.2,0.06311039924621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,2,power_law_1.2,0.20138239860534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,1,balanced,0.22789865732192993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,2,power_law_1.2,1.9055679321289063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,1,balanced,0.0643093337615331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,2,power_law_1.2,0.25742719173431394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,1,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,2,power_law_1.2,0.06704000234603882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,2,power_law_1.2,0.21994240283966066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,1,balanced,0.22374399503072104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,2,power_law_1.2,0.2933120012283325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,2,power_law_1.2,0.06694399714469909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.2,0.24455680847167968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,2,power_law_1.2,3.068998336791992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,1,balanced,0.04179200033346812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,1,balanced,0.07858133316040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,1,balanced,0.22750399510065714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.2,0.2953279972076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,2,power_law_1.2,0.07243520021438599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,2,power_law_1.2,0.39699199199676516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,1,balanced,0.05811200042565664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,1,balanced,0.22944533824920654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,1,balanced,0.10214933753013611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.2,0.34303998947143555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,2,power_law_1.2,5.572326278686523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,2,power_law_1.2,0.07508479952812194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,1,balanced,0.0885599950949351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,1,balanced,0.1483786702156067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,1,balanced,0.2339679996172587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,2,power_law_1.2,0.507750415802002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.2,0.4406720161437988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,2,power_law_1.2,0.07434239983558655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,1,balanced,0.13377599914868674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,1,balanced,0.021221332252025604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,1,balanced,0.23161600033442178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,1,balanced,0.24215465784072876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,2,power_law_1.2,0.6664512157440186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.2,0.547276782989502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,2,power_law_1.2,0.0787392020225525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,1,balanced,0.1406719982624054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,1,balanced,0.21707733472188315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,1,balanced,0.021344001094500225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,1,balanced,0.2493333419164022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,2,power_law_1.2,0.8559552192687988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.2,0.6944320201873779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,1,balanced,0.1439733306566874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,1,balanced,0.21993066867192587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,2,power_law_1.2,0.08033279776573181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,1,balanced,0.023402666052182514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,1,balanced,0.2577386697133382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,1,balanced,0.2116480072339376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,1,balanced,0.14180800318717957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.2,0.8662783622741699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,2,power_law_1.2,1.288576030731201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,2,power_law_1.2,0.07895680069923401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,1,balanced,0.03201599915822347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,1,balanced,0.14460800091425577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,1,balanced,0.21280000607172647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.2,1.2629440307617188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,2,power_law_1.2,0.08217599987983704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,1,balanced,0.051701332132021584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,1,balanced,0.15410133202870688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,2,power_law_1.2,1.7254079818725585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,1,balanced,0.2801706592241923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,1,balanced,0.215503990650177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.2,1.6083839416503907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,2,power_law_1.2,0.08934400081634522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,1,balanced,0.05386666456858317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,1,balanced,0.1492959956328074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,2,power_law_1.2,1.9535167694091797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,2,power_law_1.2,0.10506880283355713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,1,balanced,0.21861332654953003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,2,power_law_1.2,2.181439971923828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,1,balanced,0.3031733234723409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,1,balanced,0.054842665791511536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,1,balanced,0.1564906636873881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,2,power_law_1.2,3.0586496353149415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,1,balanced,0.2175253431002299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,2,power_law_1.2,0.12538880109786987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,2,power_law_1.2,3.279206466674805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,1,balanced,0.1609226663907369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,1,balanced,0.37626667817433673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,1,balanced,0.05570133527119955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,1,balanced,0.22041600942611694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,2,power_law_1.2,5.9264575958251955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,1,balanced,0.05760000149408976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,1,balanced,0.1655786633491516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,1,balanced,0.2220319906870524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.2,0.14648319482803346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,1,balanced,0.05961599946022034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,1,balanced,0.17523199319839478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,1,balanced,0.23497599363327026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,2,power_law_1.2,6.385030364990234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,1,balanced,0.06054399907588959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,1,balanced,0.41222933928171795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.2,0.18643840551376342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,1,balanced,0.18110400438308716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,1,balanced,0.24336532751719156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,1,balanced,0.06605866551399231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.2,0.16288000345230103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,1,balanced,0.1926400065422058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,1,balanced,0.24703466892242432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,1,balanced,0.0681386689345042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,1,balanced,0.2237493395805359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,1,balanced,0.5434826612472534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.2,0.18743040561676025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,1,balanced,0.27875200907389325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,1,balanced,0.06540266672770183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.2,0.21746559143066407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,1,balanced,0.2513759930928548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,1,balanced,0.07022933165232341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,1,balanced,0.30606400966644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.2,0.29607040882110597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,1,balanced,0.07277866701285045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.2,0.3431936025619507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,1,balanced,0.3234933416048686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,1,balanced,0.6203573147455851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,1,balanced,0.08212266862392426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,1,balanced,0.3644853432973226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.2,0.48317441940307615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,1,balanced,0.10291733344395955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.2,0.6017280101776123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,1,balanced,0.3612213134765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,1,balanced,0.41633065541585285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,1,balanced,0.1067573328812917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,2,power_law_1.2,0.7774271965026855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,1,balanced,0.8526079654693604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,1,balanced,0.5072960058848063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,1,balanced,0.6449013153711954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,2,power_law_1.2,1.2332927703857421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,1,balanced,0.14729066689809164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,2,power_law_1.2,2.301375961303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,1,balanced,0.13525333007176718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,1,balanced,0.6111306746800741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,1,balanced,0.7565546830495199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,1,balanced,0.17531200249989828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,1,balanced,1.0383520126342773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,1,balanced,1.0709333419799805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,1,balanced,0.8693280220031738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,1,balanced,0.18321067094802856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,1,balanced,0.295413335164388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,1,balanced,1.3713173866271973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,1,balanced,1.1066880226135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,1,balanced,1.4885066350301106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,1,balanced,0.31462933619817096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,1,balanced,1.9794027010599773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,1,balanced,1.6068746248881023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,1,balanced,1.924575964609782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,1,balanced,0.44657599925994873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,1,balanced,2.597658634185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,1,balanced,2.081941286722819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,1,balanced,2.367087999979655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,1,balanced,0.5817013184229533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,1,balanced,3.2005227406819663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,1,balanced,2.5691307385762534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,1,balanced,0.7238293488820394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,1,balanced,3.6699307759602866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,1,power_law_1.01,0.032364800572395325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,1,power_law_1.01,0.04191359877586365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,1,power_law_1.01,0.06421120166778564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,1,power_law_1.01,0.0905344009399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,1,power_law_1.01,0.11189119815826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,1,balanced,5.03055985768636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,1,power_law_1.01,0.1785215973854065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,1,balanced,4.0327574412028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,1,power_law_1.01,0.18636159896850585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,1,power_law_1.01,0.19447040557861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,1,power_law_1.01,0.20297598838806152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,1,power_law_1.01,0.20368640422821044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,1,power_law_1.01,0.21052160263061523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,1,balanced,1.1411573092142742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,1,power_law_1.01,0.21671040058135987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,1,power_law_1.01,0.2204416036605835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,1,power_law_1.01,0.22375040054321288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,1,power_law_1.01,0.23310720920562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,1,power_law_1.01,0.24068479537963866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,1,power_law_1.01,0.2550335884094238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,1,power_law_1.01,0.28545279502868653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,1,power_law_1.01,0.31820800304412844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,1,power_law_1.01,0.3928639888763428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,1,power_law_1.01,0.4354239940643311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,1,power_law_1.01,0.5531263828277588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,1,power_law_1.01,0.6681663990020752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,1,power_law_1.01,0.8887999534606934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,1,power_law_1.01,1.1092800140380858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,1,power_law_1.01,1.5369152069091796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,1,power_law_1.01,1.9751615524291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,1,power_law_1.01,2.4129791259765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,1,power_law_1.01,3.679673767089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,1,power_law_1.01,7.10025634765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,1,power_law_1.01,0.03323520123958588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,1,power_law_1.01,0.038406398892402646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,1,power_law_1.01,0.062118399143219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,1,balanced,7.133120218912761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,1,power_law_1.01,0.05030400156974792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,1,power_law_1.01,0.06771199703216553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,1,power_law_1.01,0.06526079773902893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,1,power_law_1.01,0.09160320162773132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,1,power_law_1.01,0.08464000225067139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,1,power_law_1.01,0.11555839776992798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,1,power_law_1.01,0.11969280242919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,1,power_law_1.01,0.1336575984954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,1,power_law_1.01,0.12341760396957398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,1,power_law_1.01,0.19111039638519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,1,power_law_1.01,0.1298815965652466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,1,power_law_1.01,0.1971392035484314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,1,power_law_1.01,0.13532160520553588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,1,power_law_1.01,0.19867520332336425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,1,power_law_1.01,0.14179199934005737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,1,power_law_1.01,0.2006592035293579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,1,balanced,9.906421025594076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,1,power_law_1.01,0.20619521141052247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,1,balanced,7.8249867757161455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,1,power_law_1.01,0.1503872036933899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,1,power_law_1.01,0.01895039975643158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,1,power_law_1.01,0.2096384048461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,1,power_law_1.01,0.15909759998321532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,1,power_law_1.01,0.21459200382232665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,1,power_law_1.01,0.1679487943649292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,1,power_law_1.01,0.020204800367355346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,1,power_law_1.01,0.2190079927444458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,1,power_law_1.01,0.16201599836349487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,1,power_law_1.01,0.2228480100631714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,1,power_law_1.01,0.02426239997148514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,1,power_law_1.01,0.1745344042778015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,1,power_law_1.01,0.23046400547027587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,1,balanced,2.2322932879130044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,1,power_law_1.01,0.18246400356292725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,1,power_law_1.01,0.03299840092658997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,1,power_law_1.01,0.23824639320373536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,1,power_law_1.01,0.19881600141525269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,1,power_law_1.01,0.2598335981369019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,1,power_law_1.01,0.23827838897705078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,1,power_law_1.01,0.042368000745773314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,1,power_law_1.01,0.060057598352432254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,1,power_law_1.01,0.3001728057861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,1,power_law_1.01,0.26119680404663087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,1,power_law_1.01,0.06183680295944214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,1,power_law_1.01,0.3303040027618408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.01,0.3328383922576904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,1,power_law_1.01,0.06542720198631287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,1,power_law_1.01,0.3834239959716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.01,0.4150847911834717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,1,power_law_1.01,0.06680960059165955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,1,power_law_1.01,0.5131775856018066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.01,0.48645758628845215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,1,power_law_1.01,0.06936320066452026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,1,power_law_1.01,0.6328192234039307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.01,0.6420608043670655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,1,power_law_1.01,0.07119359970092773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.01,0.7951488018035888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,1,power_law_1.01,0.8860735893249512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,1,power_law_1.01,0.07514879703521729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,1,power_law_1.01,1.1228672027587892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.01,1.1013119697570801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,1,power_law_1.01,0.0770367980003357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,1,power_law_1.01,1.610201644897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.01,1.4026432037353516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,1,power_law_1.01,0.07490559816360473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.01,2.0081472396850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,1,power_law_1.01,2.115827178955078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,1,power_law_1.01,0.08048639893531799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,1,power_law_1.01,0.08375040292739869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.01,2.6136703491210938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,1,power_law_1.01,2.602828788757324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,1,power_law_1.01,0.09324799776077271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.01,3.2153854370117188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,1,power_law_1.01,0.11168639659881592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,1,power_law_1.01,4.027852630615234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.01,0.12718720436096193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,1,power_law_1.01,5.009369659423828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.01,0.16311039924621581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.01,0.15045759677886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,1,power_law_1.01,7.837324523925782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.01,0.1911679983139038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,1,power_law_1.01,9.882854461669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.01,0.22851200103759767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.01,0.2911936044692993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.01,0.3562112092971802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.01,0.48490238189697266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.01,0.617625617980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.01,0.7530752182006836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,1,power_law_1.01,1.1529088020324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,1,power_law_1.01,2.2387903213500975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,1,power_law_1.2,0.03235200047492981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,1,power_law_1.2,0.040012800693511964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,1,power_law_1.2,0.061267197132110596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,1,power_law_1.2,0.07999359965324401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,1,power_law_1.2,0.10728960037231446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,1,power_law_1.2,0.17848960161209107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,1,power_law_1.2,0.1930816054344177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,1,power_law_1.2,0.1998528003692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,1,power_law_1.2,0.20640640258789061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,1,power_law_1.2,0.20638720989227294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,1,power_law_1.2,0.21184000968933106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,1,power_law_1.2,0.22305920124053955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,1,power_law_1.2,0.22225921154022216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,1,power_law_1.2,0.23059840202331544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,1,power_law_1.2,0.23905279636383056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,1,power_law_1.2,0.24828801155090333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,1,power_law_1.2,0.2610879898071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,1,power_law_1.2,0.2923264026641846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,1,power_law_1.2,0.32751998901367185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,1,power_law_1.2,0.39565439224243165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,1,power_law_1.2,0.4431039810180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,1,power_law_1.2,0.5569791793823242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,1,power_law_1.2,0.6741824150085449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,1,power_law_1.2,0.8949248313903808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,1,power_law_1.2,1.1102463722229003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,1,power_law_1.2,1.5383808135986328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,1,power_law_1.2,0.06183680295944214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,1,power_law_1.2,0.03278720080852508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,1,power_law_1.2,1.976595115661621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,1,power_law_1.2,0.0191551998257637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,1,power_law_1.2,0.06704639792442321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,1,power_law_1.2,0.037894400954246524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,1,power_law_1.2,2.4242944717407227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,1,power_law_1.2,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,1,power_law_1.2,0.08366079926490784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,1,power_law_1.2,0.020742399990558623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,1,power_law_1.2,0.06103039979934692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,1,power_law_1.2,3.6789825439453123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,1,power_law_1.2,0.10285439491271972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,1,power_law_1.2,0.07677440047264099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,1,power_law_1.2,0.02407039999961853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,1,power_law_1.2,0.1211135983467102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,1,power_law_1.2,0.12229119539260865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,1,power_law_1.2,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,1,power_law_1.2,0.13011200428009034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,1,power_law_1.2,0.187225604057312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,1,power_law_1.2,7.080601501464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,4,balanced,0.07228800157705943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,1,power_law_1.2,0.1361791968345642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,1,power_law_1.2,0.04197759926319122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,1,power_law_1.2,0.19790719747543334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,1,power_law_1.2,0.1385472059249878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,1,power_law_1.2,0.05774719715118408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,4,balanced,0.10564266641934712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,1,power_law_1.2,0.201529598236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,1,power_law_1.2,0.14270080327987672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,1,power_law_1.2,0.06254079937934875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,1,power_law_1.2,0.14926079511642457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,1,power_law_1.2,0.06580479741096497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,4,balanced,0.15892266233762106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,1,power_law_1.2,0.20750720500946046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,1,power_law_1.2,0.1598207950592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,1,power_law_1.2,0.16862720251083374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,1,power_law_1.2,0.06999040246009827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,4,balanced,0.26200000445048016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,1,power_law_1.2,0.20556159019470216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,1,power_law_1.2,0.163481605052948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,1,power_law_1.2,0.07109119892120361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,4,balanced,0.4717973470687866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,1,power_law_1.2,0.17696640491485596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,1,power_law_1.2,0.2114880084991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,4,balanced,0.5758346716562907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,1,power_law_1.2,0.18558080196380616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,1,power_law_1.2,0.07059839963912964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,1,power_law_1.2,0.2024384021759033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,4,balanced,0.5772159894307455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,1,power_law_1.2,0.21270399093627929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,1,power_law_1.2,0.07525759935379028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,1,power_law_1.2,0.24095358848571777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,4,balanced,0.5824319918950399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,1,power_law_1.2,0.21843841075897216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,1,power_law_1.2,0.0777728021144867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,1,power_law_1.2,0.2615231990814209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,4,balanced,0.5825173457463583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,1,power_law_1.2,0.22012159824371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,1,power_law_1.2,0.3337023973464966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,1,power_law_1.2,0.23653120994567872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,4,balanced,0.5800000031789144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,1,power_law_1.2,0.07608960270881653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,1,power_law_1.2,0.38599679470062254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,1,power_law_1.2,0.24188799858093263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,4,balanced,0.5841706593831381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,1,power_law_1.2,0.08058879971504211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,1,power_law_1.2,0.5168320178985596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,1,power_law_1.2,0.2599679946899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,1,power_law_1.2,0.6373695850372314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,1,power_law_1.2,0.08402559757232667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,1,power_law_1.2,0.30271999835968016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,4,balanced,0.5858186483383179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,1,power_law_1.2,0.8815808296203613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.2,0.33829119205474856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,1,power_law_1.2,0.09431040287017822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,4,balanced,0.051856001218159996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,4,balanced,0.587168018023173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,1,power_law_1.2,1.1220671653747558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.2,0.41635842323303224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,1,power_law_1.2,0.11336319446563721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,1,power_law_1.2,1.5975232124328613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,4,balanced,0.5916959842046102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,4,balanced,0.07215466598669688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.2,0.48699522018432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.2,0.1290112018585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,4,balanced,0.08025600016117096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,1,power_law_1.2,2.1128704071044924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,4,balanced,0.6031893491744995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,4,balanced,0.10021866361300151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.2,0.16291199922561644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.2,0.6458559989929199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,4,balanced,0.10744532942771912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,1,power_law_1.2,2.6041664123535155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,4,balanced,0.1553813318411509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.2,0.15028480291366578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,4,balanced,0.6017013390858968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.2,0.792742395401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,4,balanced,0.05348266661167145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,4,balanced,0.1379039982954661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,1,power_law_1.2,4.029702377319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,4,balanced,0.2616213361422221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,4,balanced,0.23930666844050089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,4,balanced,0.3141653339068095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.2,0.19382400512695314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.2,1.0967616081237792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,4,balanced,0.6117226680119833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,4,balanced,0.05221333106358846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,4,balanced,0.3593759934107463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,4,balanced,0.31622399886449176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,1,power_law_1.2,7.824358367919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,4,balanced,0.06717866659164429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,4,balanced,0.41018132368723553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,4,balanced,0.3208906650543213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.2,0.23537919521331788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.2,1.38985595703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,4,balanced,0.09483733773231506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,4,balanced,0.6922240257263184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,4,balanced,0.38913599650065106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,4,balanced,0.32231465975443524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,4,balanced,0.14596800009409586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,4,balanced,0.386677344640096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,4,balanced,0.32174400488535565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.2,0.29585919380187986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.2,1.9923135757446289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,4,balanced,0.1809706687927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,4,balanced,0.3770560026168823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,4,balanced,0.6403359969456991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,4,balanced,0.3240320086479187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,4,balanced,0.19285333156585693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,4,balanced,0.38412801424662274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.2,0.35624959468841555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.2,2.5910463333129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,4,balanced,0.18522133429845175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,4,balanced,0.39029868443806964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,4,balanced,0.3277759949366252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,4,balanced,0.8313600222269694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.2,0.48349437713623045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.2,3.1920192718505858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,4,balanced,0.3295360008875529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,4,balanced,0.1918613314628601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,4,balanced,0.3738826513290405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.2,0.6279168128967285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,1,power_law_1.2,4.983219146728516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,4,balanced,0.33326399326324463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,4,balanced,0.7062453428904215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,4,balanced,0.19621866941452026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.2,0.7523327827453613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,4,balanced,0.34036266803741455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,4,balanced,0.21040532986323038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,4,balanced,0.38997864723205566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,4,balanced,0.34357333183288574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,1,power_law_1.2,9.788524627685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,1,power_law_1.2,1.1645183563232422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,4,balanced,0.22406933705012003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,4,balanced,0.9082346757253011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,4,balanced,0.3513813416163127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,4,balanced,0.3996266523996989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,4,balanced,0.23708800474802652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,1,power_law_1.2,2.2462976455688475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,4,balanced,0.40120001633961994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,4,balanced,0.24357332785924277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,4,balanced,0.38253335158030194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,4,balanced,0.2638453245162964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,4,balanced,0.3838239908218384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,4,balanced,0.3800319830576579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,4,balanced,0.8445493380228678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,4,balanced,0.2855253418286641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,4,balanced,0.3997866710027059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,4,balanced,0.3078453342119853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,4,balanced,0.466378649075826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,4,balanced,0.39446401596069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,4,balanced,0.3051519989967346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,4,balanced,0.379802664120992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,4,balanced,0.6056053241093954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,4,balanced,0.44618133703867596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,4,balanced,1.353882630666097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,4,balanced,0.7440533638000488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,4,balanced,0.3954399824142456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,4,balanced,0.7014826933542887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,4,balanced,0.503770669301351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,4,balanced,0.40219199657440186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,4,balanced,0.5885973374048868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,4,balanced,1.313968022664388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,4,balanced,0.5623039801915487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,4,balanced,0.7286720275878906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,4,balanced,0.5961173375447592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,4,balanced,0.7066453297932943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,4,balanced,0.8771039644877116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,4,balanced,0.9956693649291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,4,balanced,2.3370240529378257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,4,balanced,1.0137706597646077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,4,balanced,0.9429013729095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,4,balanced,1.1238773663838704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,4,balanced,1.3543680508931477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,4,balanced,1.491322676340739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,4,balanced,2.431813398996989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,4,balanced,1.669040044148763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,4,balanced,1.5211680730183919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,4,balanced,1.7215946515401204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,4,balanced,2.5732852617899575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,4,balanced,2.145930608113607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,4,balanced,2.2096640268961587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,4,balanced,1.915877342224121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,4,balanced,2.3706186612447104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,4,balanced,2.358853340148926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,4,balanced,4.540794690450032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,4,power_law_1.01,0.155622398853302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,4,power_law_1.01,0.18863999843597412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,4,power_law_1.01,0.13365119695663452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,4,power_law_1.01,0.22216320037841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,4,power_law_1.01,0.2900736093521118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,4,power_law_1.01,0.46316161155700686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,4,power_law_1.01,0.5032192230224609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,4,power_law_1.01,0.4998976230621338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,4,power_law_1.01,0.49485440254211427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,4,balanced,3.12224547068278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,4,power_law_1.01,0.5215104103088379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,4,power_law_1.01,0.5373375892639161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,4,power_law_1.01,0.5391488075256348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,4,power_law_1.01,0.532371187210083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,4,power_law_1.01,0.5773568153381348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,4,power_law_1.01,0.6002048015594482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,4,power_law_1.01,0.6156352043151856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,4,balanced,4.062469482421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,4,power_law_1.01,0.6517375946044922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,4,power_law_1.01,0.7167935848236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,4,power_law_1.01,0.7859776020050049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,4,power_law_1.01,0.9145279884338379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,4,power_law_1.01,0.8404224395751954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,4,balanced,3.7239147822062173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,4,power_law_1.01,1.0605888366699219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,4,power_law_1.01,1.150233554840088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,4,power_law_1.01,1.5484095573425294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,4,power_law_1.01,2.110848045349121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,4,power_law_1.01,2.6079679489135743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,4,power_law_1.01,3.3047744750976564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,4,power_law_1.01,3.91833610534668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,4,power_law_1.01,5.942425537109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,4,power_law_1.01,10.252159881591798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,4,power_law_1.01,0.09459199905395507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,4,power_law_1.01,0.12119040489196778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,4,power_law_1.01,0.084307199716568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,4,power_law_1.01,0.13175679445266725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,4,power_law_1.01,0.1734015941619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,4,power_law_1.01,0.255020809173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,4,power_law_1.01,0.2701184034347534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,4,balanced,8.248170852661133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,4,power_law_1.01,0.27675518989562986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,4,power_law_1.01,0.28092160224914553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,4,power_law_1.01,0.2876607894897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,4,power_law_1.01,0.29607040882110597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,4,power_law_1.01,0.3080384016036987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,4,power_law_1.01,0.30037119388580324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,4,power_law_1.01,0.13575680255889894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,4,power_law_1.01,0.3181504011154175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,4,power_law_1.01,0.34254720211029055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,4,power_law_1.01,0.051692801713943484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,4,power_law_1.01,0.16455680131912231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,4,power_law_1.01,0.11273599863052368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,4,power_law_1.01,0.3518527984619141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,4,balanced,5.902128219604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,4,power_law_1.01,0.19994879961013795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,4,power_law_1.01,0.35683839321136473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,4,power_law_1.01,0.07916160225868225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,4,power_law_1.01,0.21989760398864747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,4,power_law_1.01,0.4218111991882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,4,power_law_1.01,0.05498239994049072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,4,power_law_1.01,0.32473599910736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,4,power_law_1.01,0.41527681350708007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,4,power_law_1.01,0.07709439992904663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,4,power_law_1.01,0.3510976076126099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,4,power_law_1.01,0.5004799842834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,4,power_law_1.01,0.09403520226478576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,4,power_law_1.01,0.3567744016647339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,4,power_law_1.01,0.6045567989349365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,4,power_law_1.01,0.15008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,4,balanced,7.524250666300456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,4,power_law_1.01,0.36611199378967285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,4,power_law_1.01,0.7517248153686523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,4,power_law_1.01,0.1544319987297058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,4,power_law_1.01,0.3599168062210083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,4,power_law_1.01,0.15455360412597657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,4,power_law_1.01,0.7741888046264649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,4,power_law_1.01,0.3668479919433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,4,power_law_1.01,0.15279359817504884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,4,power_law_1.01,1.013657569885254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,4,power_law_1.01,0.16526080369949342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,4,power_law_1.01,0.3723455905914307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,4,power_law_1.01,1.3119999885559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,4,power_law_1.01,0.17098239660263062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,4,power_law_1.01,0.3734463930130005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,4,balanced,7.49239985148112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,4,power_law_1.01,1.7275455474853516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,4,power_law_1.01,0.18324480056762696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,4,power_law_1.01,0.3700223922729492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,4,power_law_1.01,0.19140479564666749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,4,power_law_1.01,2.0971328735351564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,4,power_law_1.01,0.3722624063491821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,4,power_law_1.01,0.19478399753570558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,4,power_law_1.01,0.4024384021759033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,4,power_law_1.01,0.20244479179382324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,4,power_law_1.01,0.3951807975769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,4,power_law_1.01,2.7952640533447264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,4,power_law_1.01,0.2185920000076294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,4,power_law_1.01,0.4511104106903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,4,power_law_1.01,0.20997118949890137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.01,0.4848447799682617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,4,power_law_1.01,3.9705150604248045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.01,0.5511360168457031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,4,power_law_1.01,0.2247488021850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.01,0.649625587463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.01,0.5140416145324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.01,0.7276224136352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.01,0.6348608016967774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,4,power_law_1.01,7.2885185241699215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.01,0.34650239944458006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.01,0.8915776252746582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.01,0.4409791946411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.01,1.202175998687744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.01,1.570246410369873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.01,0.4782464027404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.01,2.0302976608276366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.01,0.6470399856567383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.01,2.63558406829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.01,0.8654720306396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,4,power_law_1.01,3.4012287139892576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.01,1.1601920127868652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,4,power_law_1.01,5.159539031982422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.01,1.4064448356628418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,4,power_law_1.01,1.8285247802734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,4,power_law_1.01,9.396697235107421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,4,power_law_1.01,2.698137664794922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,4,power_law_1.01,5.627872085571289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,4,power_law_1.2,0.15631359815597534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,4,power_law_1.2,0.1030079960823059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,4,power_law_1.2,0.14535679817199706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,4,power_law_1.2,0.2119296073913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,4,power_law_1.2,0.24525439739227295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,4,power_law_1.2,0.4564095973968506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,4,power_law_1.2,0.4688576221466064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,4,power_law_1.2,0.4936063766479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,4,power_law_1.2,0.535481595993042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,4,power_law_1.2,0.5340415954589843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,4,power_law_1.2,0.5466752052307129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,4,power_law_1.2,0.5523903846740723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,4,power_law_1.2,0.5681024074554444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,4,power_law_1.2,0.611840009689331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,4,power_law_1.2,0.597599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,4,power_law_1.2,0.6411327838897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,4,power_law_1.2,0.6620543956756592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,4,power_law_1.2,0.09789440035820007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,4,power_law_1.2,0.7721727848052978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,4,power_law_1.2,0.09020159840583801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,4,power_law_1.2,0.7876287937164307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,4,power_law_1.2,1.00447359085083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,4,power_law_1.2,0.09294080138206481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,4,power_law_1.2,0.12972160577774047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,4,power_law_1.2,0.8935104370117187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,4,power_law_1.2,0.16588159799575805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,4,power_law_1.2,1.1877759933471679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,4,power_law_1.2,0.25449600219726565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,4,power_law_1.2,1.2983872413635253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,4,power_law_1.2,0.27949440479278564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,4,power_law_1.2,1.6504640579223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,4,power_law_1.2,0.2797696113586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,4,power_law_1.2,0.2899775981903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,4,power_law_1.2,2.3835775375366213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,4,power_law_1.2,0.2955904006958008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,4,power_law_1.2,2.9929664611816404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,4,power_law_1.2,0.30935680866241455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,4,power_law_1.2,0.3195904016494751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,4,power_law_1.2,3.8484161376953123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,4,power_law_1.2,0.3229248046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,4,power_law_1.2,0.3377984046936035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,4,power_law_1.2,4.169689559936524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,4,power_law_1.2,0.3575808048248291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,4,power_law_1.2,0.13922560214996338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,4,power_law_1.2,0.3588671922683716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,4,power_law_1.2,6.056793594360352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,4,power_law_1.2,0.05140479803085327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,4,power_law_1.2,0.378003191947937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,4,power_law_1.2,0.11260160207748413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,4,power_law_1.2,0.4443647861480713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,4,power_law_1.2,0.12783360481262207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,4,power_law_1.2,0.045772799849510194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,4,power_law_1.2,0.4517888069152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,4,power_law_1.2,0.1743872046470642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,4,power_law_1.2,0.05902079939842224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,4,power_law_1.2,11.114182281494141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,4,power_law_1.2,0.52674560546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,4,power_law_1.2,0.2215104103088379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,4,power_law_1.2,0.07765759825706482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,4,power_law_1.2,0.621619176864624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,4,power_law_1.2,0.3311295986175537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,4,power_law_1.2,0.09153280258178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,2,balanced,0.0703359991312027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,4,power_law_1.2,0.3467072010040283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,4,power_law_1.2,0.14227839708328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,2,balanced,0.09756799538930257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,4,power_law_1.2,0.35019519329071047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,4,power_law_1.2,0.1505344033241272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,4,power_law_1.2,0.786521577835083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,4,power_law_1.2,0.36319360733032224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,2,balanced,0.1500746707121531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,4,power_law_1.2,0.1593984007835388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,2,balanced,0.2548266649246216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,4,power_law_1.2,0.37345280647277834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,4,power_law_1.2,0.8592000007629395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,4,power_law_1.2,0.16526080369949342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,2,balanced,0.46855465571085614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,4,power_law_1.2,0.37969279289245605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,4,power_law_1.2,0.16647679805755616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,2,balanced,0.5790293216705322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,4,power_law_1.2,1.0894975662231445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,4,power_law_1.2,0.38570239543914797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,2,balanced,0.5781013170878092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,4,power_law_1.2,0.17863039970397948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,4,power_law_1.2,0.3877376079559326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,2,balanced,0.5837706724802653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,4,power_law_1.2,0.18245760202407837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,4,power_law_1.2,1.3515456199645997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,2,balanced,0.05197866757710775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,2,balanced,0.5885279973347982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,4,power_law_1.2,0.1931264042854309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,4,power_law_1.2,0.3851072072982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,2,balanced,0.07194666564464569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,4,power_law_1.2,2.002374458312988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,2,balanced,0.5903199911117554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,4,power_law_1.2,0.19390079975128174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,2,balanced,0.1002293328444163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,4,power_law_1.2,0.39756801128387453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,2,balanced,0.15340266625086466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,2,balanced,0.5894773403803507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,4,power_law_1.2,0.22033920288085937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,2,balanced,0.26262933015823364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,4,power_law_1.2,0.4095168113708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,2,balanced,0.5959626833597819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,4,power_law_1.2,0.23366398811340333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,4,power_law_1.2,2.271699142456055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,2,balanced,0.31569067637125653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,2,balanced,0.5987840096155802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,2,balanced,0.31811734040578205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,4,power_law_1.2,0.4416895866394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,4,power_law_1.2,0.21622400283813475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,2,balanced,0.3202986717224121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,2,balanced,0.5977760155995687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,4,power_law_1.2,0.236409592628479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,4,power_law_1.2,0.46110081672668457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,2,balanced,0.3248853286107381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,4,power_law_1.2,2.8853952407836916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,2,balanced,0.60807998975118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.2,0.5023231983184815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.2,0.48969597816467286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,2,balanced,0.3248746593793233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.2,0.6556096076965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,2,balanced,0.3296426733334859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,2,balanced,0.6124106645584106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.2,0.5705920219421386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,2,balanced,0.3355040152867635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.2,0.37000958919525145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.2,0.6783167839050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,2,balanced,0.6262720028559366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,4,power_law_1.2,4.202656173706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,2,balanced,0.335098663965861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,2,balanced,0.08121599753697713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.2,0.4322815895080566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.2,0.8085247993469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,2,balanced,0.341264009475708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,2,balanced,0.665615995724996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.2,0.5165567874908448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,2,balanced,0.10732266306877136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,2,balanced,0.348746657371521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.2,0.9853376388549805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,2,balanced,0.1370560030142466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,2,balanced,0.043791999419530235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.2,0.6574912071228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,4,power_law_1.2,8.506886291503907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,2,balanced,0.3535733222961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,2,balanced,0.6700692971547445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.2,1.440447998046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,2,balanced,0.22988800207773843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.2,0.815283203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,2,balanced,0.045408000548680626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,2,balanced,0.3673653205235799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,2,balanced,0.060959999759991966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.2,1.2444352149963378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.2,1.5323264122009277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,2,balanced,0.35068798065185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,2,balanced,0.8080533345540365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,2,balanced,0.0890773336092631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,2,balanced,0.39131732781728107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.2,1.5199935913085938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.2,2.3563264846801757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,2,balanced,0.4073226849238078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,2,balanced,0.14708800117174783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,4,power_law_1.2,2.0655040740966797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,2,balanced,0.40910931428273517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,2,balanced,0.17881600062052408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,2,balanced,0.7483092943827311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.2,3.0090688705444335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,2,balanced,0.37965333461761475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,2,balanced,0.18428800503412882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,4,power_law_1.2,3.229964828491211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,2,balanced,0.3832159837086995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,4,power_law_1.2,3.4547264099121096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,2,balanced,0.19403733809789023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,2,balanced,0.512117346127828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,2,balanced,0.9423519770304362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,2,balanced,0.20195200045903525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,2,balanced,0.38789868354797363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,4,power_law_1.2,5.574950408935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,4,power_law_1.2,6.204460906982422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,2,balanced,0.21136534214019775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,2,balanced,0.51582932472229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,2,balanced,0.37142399946848553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,2,balanced,0.20583999156951904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,2,balanced,0.3724106550216675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,2,balanced,0.21659199396769205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,4,power_law_1.2,11.7042236328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,2,balanced,0.9229119618733724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,2,balanced,0.38685333728790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,2,balanced,0.2286400000254313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,2,balanced,0.6610560019810995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,2,balanced,0.3752906719843547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,2,balanced,0.39774401982625324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,2,balanced,0.213919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,2,balanced,1.4274932543436687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,2,balanced,0.4004799922307332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,2,balanced,0.22538665930430093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,2,balanced,0.376581350962321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,2,balanced,0.24067199230194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,2,balanced,0.7225279808044434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,2,balanced,0.39635733763376874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,2,balanced,0.2605973283449809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,2,balanced,0.4065973361333211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,2,balanced,0.30100266138712567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,2,balanced,0.41414399941762287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,2,balanced,0.27507734298706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,2,balanced,1.4427040417989094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,2,balanced,1.0610986550649006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,2,balanced,0.4645013411839803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,2,balanced,0.4405706723531087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,2,balanced,0.4535893201828003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,2,balanced,0.47658665974934894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,2,balanced,0.4873173236846924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,2,balanced,1.2192053000132244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,2,balanced,2.5114506085713706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,2,balanced,0.7747253576914469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,2,balanced,0.490389347076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,2,balanced,0.8148000240325928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,2,balanced,0.8221759796142578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,2,balanced,1.9694080352783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,2,balanced,1.1648693084716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,2,balanced,2.6523893674214682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,2,balanced,0.8806400299072266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,2,balanced,1.519978682200114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,2,balanced,2.283573309580485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,2,balanced,1.3024799823760986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,2,balanced,2.8921868006388345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,2,balanced,1.8958667119344075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,2,balanced,1.6883947054545085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,2,balanced,2.6464053789774575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,2,balanced,2.6089332898457847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,2,balanced,5.025381406148274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,2,balanced,1.786128044128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,2,balanced,2.961029370625814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,2,power_law_1.01,0.10324480533599853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,2,power_law_1.01,0.11924480199813843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,2,power_law_1.01,0.12802560329437257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,2,power_law_1.01,0.19625600576400756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,2,balanced,4.178677241007487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,2,power_law_1.01,0.28584320545196534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,2,power_law_1.01,0.4378367900848389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,2,power_law_1.01,0.4799808025360107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,2,power_law_1.01,0.4933760166168213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,2,power_law_1.01,0.5044928073883057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,2,power_law_1.01,0.5230591773986817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,2,power_law_1.01,0.5160831928253173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,2,power_law_1.01,0.5385087966918946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,2,power_law_1.01,0.558739185333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,2,power_law_1.01,0.5652736186981201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,2,power_law_1.01,0.5860095977783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,2,power_law_1.01,0.6130688190460205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,2,power_law_1.01,0.6224192142486572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,2,power_law_1.01,0.6925055980682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,2,power_law_1.01,0.7247935771942139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,2,balanced,3.063098589579264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,2,power_law_1.01,0.8615039825439453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,2,power_law_1.01,0.8246975898742676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,2,balanced,4.6973012288411455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,2,power_law_1.01,1.0372544288635255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,2,power_law_1.01,1.164735984802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,2,power_law_1.01,1.4100607872009276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,2,power_law_1.01,2.061631965637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,2,power_law_1.01,2.6331071853637695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,2,power_law_1.01,3.1611455917358398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,2,power_law_1.01,3.8268352508544923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,2,power_law_1.01,5.728351974487305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,2,power_law_1.01,10.694233703613282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,2,power_law_1.01,0.06900479793548583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,2,power_law_1.01,0.10239360332489014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,2,balanced,9.310805638631185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,2,power_law_1.01,0.07842559814453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,2,power_law_1.01,0.1337407946586609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,2,power_law_1.01,0.08011519908905029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,2,power_law_1.01,0.11341439485549927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,2,power_law_1.01,0.1310528039932251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,2,power_law_1.01,0.18287359476089476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,2,power_law_1.01,0.16241919994354248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,2,power_law_1.01,0.21566081047058105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,2,power_law_1.01,0.24897279739379882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,2,power_law_1.01,0.32736639976501464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,2,power_law_1.01,0.27212159633636473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,2,power_law_1.01,0.34401280879974366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,2,power_law_1.01,0.2834687948226929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,2,power_law_1.01,0.339897608757019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,2,power_law_1.01,0.35550079345703123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,2,power_law_1.01,0.2800256013870239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,2,power_law_1.01,0.35491199493408204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,2,power_law_1.01,0.28580479621887206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,2,power_law_1.01,0.03240959942340851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,2,power_law_1.01,0.3705663919448853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,2,power_law_1.01,0.30049920082092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,2,power_law_1.01,0.36319360733032224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,2,balanced,7.949653625488281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,2,power_law_1.01,0.30568320751190187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,2,power_law_1.01,0.3808831930160522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,2,power_law_1.01,0.042419201135635375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,2,power_law_1.01,0.311846399307251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,2,power_law_1.01,0.3859584093093872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,2,power_law_1.01,0.04917120039463043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,2,power_law_1.01,0.32072958946228025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,2,power_law_1.01,0.3864703893661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,2,power_law_1.01,0.07157760262489318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,2,power_law_1.01,0.08886399865150452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,2,power_law_1.01,0.33452160358428956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,2,power_law_1.01,0.3967616081237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,2,power_law_1.01,0.13112959861755372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,2,power_law_1.01,0.3475071907043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,2,power_law_1.01,0.42020478248596194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,2,power_law_1.01,0.14511359930038453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,2,power_law_1.01,0.35765120983123777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,2,power_law_1.01,0.15274879932403565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,2,power_law_1.01,0.46442880630493166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,2,power_law_1.01,0.39762558937072756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,2,power_law_1.01,0.15827840566635132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.01,0.49365758895874023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,2,power_law_1.01,0.4375487804412842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,2,power_law_1.01,0.16319359540939332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.01,0.5532159805297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,2,power_law_1.01,0.5339968204498291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,2,power_law_1.01,0.1684224009513855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.01,0.6425792217254639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,2,power_law_1.01,0.5532032012939453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,2,power_law_1.01,0.16817920207977294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.01,0.8361344337463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,2,power_law_1.01,0.6841216087341309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,2,power_law_1.01,0.18348159790039062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.01,0.9548288345336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,2,balanced,5.647658665974935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,2,power_law_1.01,0.18169599771499634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.01,1.2269311904907227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,2,power_law_1.01,0.8400128364562989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,2,balanced,9.28332265218099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,2,power_law_1.01,0.18988800048828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.01,1.5767423629760742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,2,power_law_1.01,1.1332608222961427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,2,power_law_1.01,0.1926911950111389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,2,power_law_1.01,1.5256704330444335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.01,2.2820159912109377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,2,power_law_1.01,0.22416000366210936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,2,power_law_1.01,0.2638911962509155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,2,power_law_1.01,2.0281087875366213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.01,0.24373118877410888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.01,2.8251903533935545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,2,power_law_1.01,2.6323711395263674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.01,0.2961087942123413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.01,0.3415359973907471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,2,power_law_1.01,3.369388961791992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,2,power_law_1.01,3.1540159225463866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.01,0.37580161094665526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.01,0.44971518516540526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,2,power_law_1.01,5.018764877319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,2,power_law_1.01,4.718265533447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.01,0.5623871803283691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.01,0.7053376197814941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,2,power_law_1.01,9.565325164794922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,2,power_law_1.01,10.39315185546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.01,0.9211199760437012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.01,1.2065216064453126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,2,power_law_1.01,1.464857578277588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,2,power_law_1.01,2.282271957397461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,2,power_law_1.01,4.246284866333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,2,power_law_1.2,0.10280959606170655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,2,power_law_1.2,0.08271359801292419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,2,power_law_1.2,0.11852799654006958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,2,power_law_1.2,0.1636031985282898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,2,power_law_1.2,0.23635199069976806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,2,power_law_1.2,0.4352704048156738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,2,power_law_1.2,0.4627200126647949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,2,power_law_1.2,0.48813438415527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,2,power_law_1.2,0.5061888217926025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,2,power_law_1.2,0.5302783966064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,2,power_law_1.2,0.5578176021575928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,2,power_law_1.2,0.5466879844665528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,2,power_law_1.2,0.5726848125457764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,2,power_law_1.2,0.587500810623169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,2,power_law_1.2,0.5982912063598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,2,power_law_1.2,0.6321599960327149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,2,power_law_1.2,0.6369088172912598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,2,power_law_1.2,0.7337152004241944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,2,power_law_1.2,0.7685376167297363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,2,power_law_1.2,0.894495964050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,2,power_law_1.2,0.9043519973754883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,2,power_law_1.2,1.061740779876709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,2,power_law_1.2,0.06934400200843811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,2,power_law_1.2,1.2641663551330566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,2,power_law_1.2,0.10335359573364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,2,power_law_1.2,1.5568063735961915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,2,power_law_1.2,0.05811200141906738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,2,power_law_1.2,0.09786880016326904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,2,power_law_1.2,2.1876991271972654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,2,power_law_1.2,0.08091520071029663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,2,power_law_1.2,0.12003840208053589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,2,power_law_1.2,0.11976319551467896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,2,power_law_1.2,2.6784383773803713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,2,power_law_1.2,0.16582399606704712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,2,power_law_1.2,0.152729594707489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,2,power_law_1.2,3.2803264617919923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,2,power_law_1.2,0.2588160037994385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,2,power_law_1.2,0.20428800582885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,2,power_law_1.2,0.2655359983444214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,2,power_law_1.2,0.3142591953277588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,2,power_law_1.2,4.067206573486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,2,power_law_1.2,0.283404803276062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,2,power_law_1.2,0.35853440761566163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,2,power_law_1.2,0.2842751979827881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,2,power_law_1.2,0.34983680248260496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,2,power_law_1.2,5.603359985351562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,2,power_law_1.2,0.29465599060058595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,2,power_law_1.2,0.3675904035568237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,2,power_law_1.2,0.3827903985977173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,2,power_law_1.2,0.308569598197937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,2,power_law_1.2,0.39201281070709226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,2,power_law_1.2,0.31059839725494387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,2,power_law_1.2,11.73784942626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,2,power_law_1.2,0.37620480060577394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,2,power_law_1.2,0.3231040000915527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,2,power_law_1.2,0.3857088088989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,2,power_law_1.2,0.3293312072753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,2,power_law_1.2,0.38782079219818116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,2,power_law_1.2,0.34911360740661623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,2,power_law_1.2,0.4085824012756348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,2,power_law_1.2,0.3737472057342529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,2,power_law_1.2,0.40940160751342775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,2,power_law_1.2,0.3835520029067993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,1,balanced,0.07044800122578938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,2,power_law_1.2,0.4355008125305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,2,power_law_1.2,0.4200575828552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,2,power_law_1.2,0.03247359991073608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,1,balanced,0.0977280040582021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,2,power_law_1.2,0.4667647838592529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,2,power_law_1.2,0.4644927978515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,2,power_law_1.2,0.03364480137825012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,2,power_law_1.2,0.04684799909591675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,2,power_law_1.2,0.5569215774536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,1,balanced,0.15033066272735596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.2,0.5149375915527343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,2,power_law_1.2,0.061766397953033444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,2,power_law_1.2,0.5805568218231201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,1,balanced,0.25572266181310016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.2,0.5815167903900147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,2,power_law_1.2,0.08639360070228577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,2,power_law_1.2,0.7492608070373535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,1,balanced,0.47469866275787354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,2,power_law_1.2,0.13528319597244262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,2,power_law_1.2,0.8541631698608398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.2,0.6669504165649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,1,balanced,0.0518506666024526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,2,power_law_1.2,0.1446720004081726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,1,balanced,0.5786773363749186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,2,power_law_1.2,1.1766976356506347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.2,0.8378815650939941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,2,power_law_1.2,0.152838397026062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,1,balanced,0.5816693305969238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,1,balanced,0.07303466896216075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,1,balanced,0.09831999739011128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,2,power_law_1.2,0.16117759943008422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.2,1.0137280464172362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,2,power_law_1.2,1.5857600212097167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,1,balanced,0.5884106556574503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,1,balanced,0.10362666845321655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,2,power_law_1.2,0.17200640439987183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,1,balanced,0.1601759990056356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,1,balanced,0.5912959973017374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.2,1.3029696464538574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,1,balanced,0.12577600280443826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,2,power_law_1.2,0.16768640279769897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,2,power_law_1.2,2.199532890319824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,1,balanced,0.2698773344357808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,1,balanced,0.5918826659520467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.2,1.5614975929260253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,2,power_law_1.2,0.18236160278320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,1,balanced,0.3234399954477946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,1,balanced,0.16889599959055582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,1,balanced,0.5983893473943075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,2,power_law_1.2,2.9048831939697264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,1,balanced,0.3283413251241048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,1,balanced,0.26314133405685425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,2,power_law_1.2,0.18856960535049438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,1,balanced,0.6046453317006429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.2,2.2375999450683595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,1,balanced,0.3333546717961629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,1,balanced,0.40089066823323566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,1,balanced,0.6066666841506958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,2,power_law_1.2,0.1817023992538452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,2,power_law_1.2,3.276377487182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,1,balanced,0.02460266649723053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,1,balanced,0.4702666600545247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,1,balanced,0.6142400105794271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.2,2.872159957885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,1,balanced,0.337226668993632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,2,power_law_1.2,0.18555519580841065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,1,balanced,0.4429173469543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,1,balanced,0.027456000447273254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,1,balanced,0.6294826666514078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,2,power_law_1.2,4.918291091918945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,1,balanced,0.33623464902242023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,2,power_law_1.2,3.4962432861328123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,2,power_law_1.2,0.20047359466552733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,1,balanced,0.04269866645336151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,1,balanced,0.44837868213653564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,1,balanced,0.6370240052541097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,2,power_law_1.2,9.524742126464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,1,balanced,0.34325865904490155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,1,balanced,0.05794133245944977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,2,power_law_1.2,0.2319040060043335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,1,balanced,0.4486986796061198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,2,power_law_1.2,5.474854278564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,1,balanced,0.6527573267618815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,1,balanced,0.0886400043964386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,2,power_law_1.2,0.2759936094284058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,1,balanced,0.44564799467722577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,1,balanced,0.3471359809239705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,1,balanced,0.1072213351726532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.2,0.2495487928390503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,1,balanced,0.741370677947998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,1,balanced,0.44185598691304523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,1,balanced,0.3500639994939168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,1,balanced,0.11360533038775127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,2,power_law_1.2,11.44476776123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,1,balanced,0.1135093371073405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.2,0.3104127883911133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,1,balanced,0.35788798332214355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,1,balanced,0.44728533426920575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,1,balanced,0.7144426504770914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,1,balanced,0.11632532874743144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,1,balanced,0.36818134784698486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.2,0.35350399017333983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,1,balanced,0.44785066445668537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,1,balanced,0.11967999736467998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,1,balanced,0.3797173500061035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,1,balanced,0.1281066636244456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,1,balanced,0.8746186892191569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,1,balanced,0.4799893299738566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.2,0.3861632108688354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,1,balanced,0.1360319952170054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,1,balanced,0.39717332522074383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,1,balanced,0.46745598316192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.2,0.4569215774536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,1,balanced,0.1390613317489624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,1,balanced,0.13372266292572021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,1,balanced,0.4741599957148234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,1,balanced,0.8507253328959147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,1,balanced,0.46005864938100177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.2,0.5633152008056641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,1,balanced,0.1421333352724711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,1,balanced,0.14892799655596414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.2,0.7230656147003174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,1,balanced,0.48607468605041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,1,balanced,0.46539199352264404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,1,balanced,0.1622719963391622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,1,balanced,1.2485013008117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.2,0.9544256210327149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,1,balanced,0.5220213333765665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,1,balanced,0.21662400166193643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,1,balanced,0.608677347501119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.2,1.1973952293395995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,1,balanced,0.5529386599858602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,1,balanced,0.21563732624053955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,1,balanced,0.6392906506856283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,1,balanced,1.1560533046722412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,2,power_law_1.2,1.5911359786987305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,1,balanced,0.28594134251276654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,1,balanced,0.6152480045954386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,2,power_law_1.2,2.2531904220581054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,1,balanced,0.9901813666025797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,1,balanced,0.28116800387700397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,1,balanced,0.6844533284505209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,2,power_law_1.2,4.3069313049316404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,1,balanced,1.6613653500874836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,1,balanced,0.3701653480529785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,1,balanced,0.9861226876576742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,1,balanced,1.0995946725209553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,1,balanced,0.3378453254699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,1,balanced,1.2082613309224446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,1,balanced,1.9140586853027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,1,balanced,1.4760266939798992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,1,balanced,0.5527146657307943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,1,balanced,1.7008639971415203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,1,balanced,1.7359840075174968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,1,balanced,2.936533292134603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,1,balanced,0.6129066546758016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,1,balanced,2.24292262395223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,1,balanced,2.608485380808512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,1,balanced,0.8747626940409342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,1,balanced,2.960559844970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,1,balanced,3.5764214197794595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,1,balanced,3.276122728983561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,1,balanced,1.1422719955444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,1,balanced,4.053215980529785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,1,balanced,4.011642773946126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,1,balanced,3.8324693044026694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,1,balanced,1.255568027496338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,1,power_law_1.01,0.06757760047912598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,1,power_law_1.01,0.09109119772911071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,1,power_law_1.01,0.11591039896011353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,1,balanced,4.758085250854492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,1,power_law_1.01,0.1832383990287781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,1,power_law_1.01,0.2556159973144531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,1,power_law_1.01,0.4449471950531006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,1,power_law_1.01,0.46874241828918456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,1,power_law_1.01,0.5065855979919434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,1,power_law_1.01,0.505951976776123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,1,power_law_1.01,0.5149695873260498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,1,power_law_1.01,0.5279871940612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,1,power_law_1.01,0.5450304031372071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,1,power_law_1.01,0.5581759929656982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,1,power_law_1.01,0.5860095977783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,1,balanced,6.642714818318685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,1,power_law_1.01,0.6111104011535644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,1,power_law_1.01,0.6211647987365723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,1,power_law_1.01,0.6504767894744873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,1,power_law_1.01,0.7237823963165283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,1,power_law_1.01,0.7380671977996827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,1,power_law_1.01,0.8641152381896973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,1,power_law_1.01,0.9749183654785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,1,power_law_1.01,1.2178943634033204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,1,balanced,6.086528142293294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,1,balanced,2.0705973307291665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,1,power_law_1.01,1.3003135681152345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,1,power_law_1.01,1.669932746887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,1,power_law_1.01,2.1582143783569334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,1,power_law_1.01,2.8887680053710936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,1,balanced,7.592213312784831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,1,power_law_1.01,3.6891326904296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,1,power_law_1.01,4.432275390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,1,power_law_1.01,6.655667114257812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,1,power_law_1.01,12.482937622070313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,1,power_law_1.01,0.09252480268478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,1,power_law_1.01,0.049747198820114136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,1,power_law_1.01,0.11660159826278686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,1,power_law_1.01,0.06584960222244263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,1,power_law_1.01,0.13990399837493897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,1,power_law_1.01,0.0818943977355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,1,power_law_1.01,0.20005760192871094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,1,power_law_1.01,0.12877440452575684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,1,power_law_1.01,0.24656639099121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,1,power_law_1.01,0.16716159582138063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,1,power_law_1.01,0.3745471954345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,1,power_law_1.01,0.2530816078186035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,1,power_law_1.01,0.26991360187530516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,1,power_law_1.01,0.38798720836639405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,1,power_law_1.01,0.2890048027038574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,1,power_law_1.01,0.29550719261169434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,1,power_law_1.01,0.3941119909286499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,1,power_law_1.01,0.30183680057525636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,1,power_law_1.01,0.38894081115722656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,1,power_law_1.01,0.30762879848480223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,1,power_law_1.01,0.4051328182220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,1,power_law_1.01,0.3220799922943115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,1,power_law_1.01,0.4186367988586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,1,power_law_1.01,0.3262592077255249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,1,power_law_1.01,0.43279361724853516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,1,power_law_1.01,0.3394176006317139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,1,power_law_1.01,0.4305600166320801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,1,power_law_1.01,0.3578176021575928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,1,power_law_1.01,0.4440256118774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,1,power_law_1.01,0.36784000396728517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,1,power_law_1.01,0.4627967834472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,1,power_law_1.01,0.404531192779541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,1,power_law_1.01,0.47454080581665037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,1,power_law_1.01,0.46153597831726073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,1,power_law_1.01,0.48883838653564454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,1,power_law_1.01,0.49694080352783204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,1,balanced,12.500165303548178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,1,power_law_1.01,0.5494016170501709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,1,power_law_1.01,0.6087168216705322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,1,power_law_1.01,0.7363200187683105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.01,0.6053760051727295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,1,power_law_1.01,0.9594304084777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.01,0.7160255908966064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,1,power_law_1.01,1.0812800407409668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,1,power_law_1.01,1.4363967895507812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,1,balanced,3.929658571879069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.01,0.8140671730041504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,1,balanced,11.664992014567057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,1,power_law_1.01,1.863270378112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.01,1.057817554473877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.01,1.2712127685546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,1,power_law_1.01,2.5547008514404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,1,power_law_1.01,0.02314240038394928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.01,1.7218303680419922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,1,power_law_1.01,3.19738883972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,1,power_law_1.01,0.032102400064468385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.01,2.1744384765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,1,power_law_1.01,3.9200065612792967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,1,balanced,14.836427052815756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,1,power_law_1.01,0.038540801405906676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,1,power_law_1.01,0.05545600056648255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.01,3.0677888870239256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,1,power_law_1.01,0.08138880133628845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,1,power_law_1.01,6.004313659667969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,1,power_law_1.01,0.1217344045639038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.01,3.945119857788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,1,power_law_1.01,0.14062080383300782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,1,power_law_1.01,0.1438976049423218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,1,power_law_1.01,4.834553527832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,1,power_law_1.01,11.642304229736329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,1,power_law_1.01,0.14202879667282103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,1,power_law_1.01,7.531501007080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,1,power_law_1.01,0.14375040531158448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,1,power_law_1.01,0.15448960065841674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,1,power_law_1.01,0.16554239988327027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,1,power_law_1.01,14.774406433105469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,1,power_law_1.01,0.17171839475631714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,1,power_law_1.01,0.16614400148391723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,1,power_law_1.01,0.1754047989845276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,1,power_law_1.01,0.17957119941711425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,1,power_law_1.01,0.1918720006942749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,1,power_law_1.01,0.2319488048553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.01,0.24476799964904786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.01,0.3146944046020508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.01,0.30392320156097413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.01,0.3994816064834595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.01,0.425932788848877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.01,0.539795207977295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.01,0.6565375804901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.01,0.8925056457519531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.01,1.1277312278747558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,1,power_law_1.01,1.3391615867614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,1,power_law_1.01,2.078246307373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,1,power_law_1.01,3.993996810913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,1,power_law_1.2,0.0676479995250702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,1,power_law_1.2,0.07405440211296081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,1,power_law_1.2,0.11027840375900269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,1,power_law_1.2,0.16740479469299316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,1,power_law_1.2,0.22972800731658935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,1,power_law_1.2,0.4404287815093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,1,power_law_1.2,0.48253440856933594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,1,power_law_1.2,0.5002495765686035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,1,power_law_1.2,0.5194687843322754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,1,power_law_1.2,0.5383999824523926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,1,power_law_1.2,0.5587007999420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,1,power_law_1.2,0.5758975982666016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,1,power_law_1.2,0.5854911804199219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,1,power_law_1.2,0.5956480026245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,1,power_law_1.2,0.625881576538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,1,power_law_1.2,0.6394624233245849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,1,power_law_1.2,0.6672959804534913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,1,power_law_1.2,0.7397183895111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,1,power_law_1.2,0.7665664196014405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,1,power_law_1.2,0.890220832824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,1,power_law_1.2,1.0061632156372071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,1,power_law_1.2,1.2530943870544433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,1,power_law_1.2,1.3389568328857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,1,power_law_1.2,1.7076288223266602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,1,power_law_1.2,2.2118911743164062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,1,power_law_1.2,2.915974426269531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,1,power_law_1.2,0.04965119957923889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,1,power_law_1.2,3.7284095764160154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,1,power_law_1.2,4.463667297363282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,1,power_law_1.2,0.057894402742385866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,1,power_law_1.2,0.07825919985771179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,1,power_law_1.2,0.11694079637527466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,1,power_law_1.2,6.660518646240234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,1,power_law_1.2,0.15073280334472655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,1,power_law_1.2,0.25560319423675537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,1,power_law_1.2,0.27694718837738036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,1,power_law_1.2,12.489004516601563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,1,power_law_1.2,0.2916991949081421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,1,power_law_1.2,0.3020224094390869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,1,power_law_1.2,0.3118144035339355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,1,power_law_1.2,0.3209536075592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,1,power_law_1.2,0.3319936037063599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,1,power_law_1.2,0.3374399900436401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,1,power_law_1.2,0.09265919923782348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,1,power_law_1.2,0.35481600761413573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,1,power_law_1.2,0.3704447984695435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,4,balanced,0.06861333549022675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,1,power_law_1.2,0.09882879853248597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,1,power_law_1.2,0.3831295967102051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,1,power_law_1.2,0.023475199937820435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,1,power_law_1.2,0.1342144012451172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,1,power_law_1.2,0.4137983798980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,4,balanced,0.09240000446637471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,1,power_law_1.2,0.16531840562820435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,1,power_law_1.2,0.47786879539489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,4,balanced,0.14475199580192566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,1,power_law_1.2,0.026335999369621277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,1,power_law_1.2,0.2155071973800659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,4,balanced,0.24398932854334512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,1,power_law_1.2,0.5103744029998779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,1,power_law_1.2,0.03932160139083862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,1,power_law_1.2,0.36645119190216063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,4,balanced,0.4411840041478475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,4,balanced,0.051856001218159996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,1,power_law_1.2,0.6203392028808594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,1,power_law_1.2,0.052185600996017455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,4,balanced,0.8375306924184164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,4,balanced,0.0703359991312027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,1,power_law_1.2,0.3793152093887329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,1,power_law_1.2,0.7613887786865234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,4,balanced,0.09506666660308838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,4,balanced,1.2235093116760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,4,balanced,0.05817066629727682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,1,power_law_1.2,0.9682880401611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,1,power_law_1.2,0.3964992046356201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,1,power_law_1.2,0.07494400143623352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,4,balanced,0.14801067113876343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,4,balanced,1.2268479665120442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,1,power_law_1.2,0.41224961280822753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,1,power_law_1.2,1.0992256164550782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,4,balanced,0.250490665435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,1,power_law_1.2,0.11832319498062134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,4,balanced,0.06422399977842967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,4,balanced,1.227178653081258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,4,balanced,0.4475253423055013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,1,power_law_1.2,0.41598081588745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,1,power_law_1.2,1.4491519927978516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,1,power_law_1.2,0.1347584009170532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,4,balanced,1.2245919704437256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,4,balanced,0.6438613335291544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,4,balanced,0.08444266517957051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,1,power_law_1.2,0.44009599685668943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,1,power_law_1.2,0.14414080381393432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,4,balanced,0.6458773215611776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,4,balanced,1.235968033472697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,1,power_law_1.2,1.8709119796752929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,1,power_law_1.2,0.43036160469055174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,4,balanced,0.12029866377512614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,4,balanced,0.6466079950332642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,1,power_law_1.2,0.14487680196762084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,4,balanced,1.2446666558583577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,1,power_law_1.2,0.4461376190185547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,4,balanced,0.6462613344192505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,4,balanced,0.1819093426068624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,1,power_law_1.2,2.577926445007324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,1,power_law_1.2,0.14755840301513673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,4,balanced,1.2466826438903809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,1,power_law_1.2,0.4647039890289307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,4,balanced,0.6512906551361084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,4,balanced,0.300927996635437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,1,power_law_1.2,0.15857280492782594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,4,balanced,1.2448906898498535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,1,power_law_1.2,0.4648064136505127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,4,balanced,0.6566986640294393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,4,balanced,0.42479999860127765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,1,power_law_1.2,3.216966247558594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,1,power_law_1.2,0.48207998275756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,1,power_law_1.2,0.16160639524459838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,4,balanced,0.657093326250712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,4,balanced,1.2575679620107014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,4,balanced,0.4261546532313029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,1,power_law_1.2,0.5010176181793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,1,power_law_1.2,0.17463040351867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,4,balanced,0.6621866623560587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,4,balanced,1.2655093669891357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,1,power_law_1.2,3.931903839111328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,1,power_law_1.2,0.569868803024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,1,power_law_1.2,0.17116160392761232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,4,balanced,0.4265333414077759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,4,balanced,0.6756640275319418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,4,balanced,1.267738660176595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.2,0.6191616058349609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,1,power_law_1.2,5.976614379882813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,1,power_law_1.2,0.17548160552978515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,4,balanced,0.42741334438323975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,4,balanced,0.6778826713562012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,4,balanced,1.2882826328277588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.2,0.7164415836334228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,1,power_law_1.2,11.612313842773437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,1,power_law_1.2,0.18152960538864135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,4,balanced,0.4321599801381429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,4,balanced,0.6891360282897949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,4,balanced,1.3090133666992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.2,0.8329407691955566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,4,power_law_1.01,0.14835200309753419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,1,power_law_1.2,0.203987193107605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,4,balanced,0.43566401799519855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,4,balanced,0.7145973046620687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,4,balanced,1.3398826917012532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.2,1.051027202606201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,4,power_law_1.01,0.24133760929107667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,1,power_law_1.2,0.24041600227355958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,4,balanced,0.43851733207702637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,4,balanced,0.7287680308024088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.2,1.2814847946166992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,4,power_law_1.01,0.46839680671691897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,4,balanced,1.3803359667460124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,4,balanced,0.4438186486562093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.2,0.24958720207214355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,4,balanced,0.7662026882171631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.2,1.717363166809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,4,power_law_1.01,0.20327680110931395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,4,balanced,0.4508053461710612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,4,power_law_1.01,0.36535680294036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.2,2.1621631622314452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.2,0.32164480686187746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,4,power_law_1.01,0.5202559947967529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,4,balanced,0.8090293407440186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,4,balanced,0.4551733334859212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,4,balanced,1.4420852661132812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.2,3.043833541870117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.2,0.32422399520874023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,4,power_law_1.01,0.6572415828704834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.2,0.40655999183654784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,4,power_law_1.01,0.7636799812316895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.2,3.921382522583008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,4,balanced,0.46463465690612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,4,balanced,0.888480027516683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.2,0.46195201873779296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,4,power_law_1.01,0.8410943984985352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,1,power_law_1.2,4.823878479003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,4,balanced,0.47898133595784503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,4,balanced,1.528154691060384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.2,0.5593664169311523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,4,power_law_1.01,0.8297599792480469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.2,0.6737792015075683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,1,power_law_1.2,7.5223747253417965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,4,balanced,0.49237334728240967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,4,balanced,0.9690506458282471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,4,power_law_1.01,0.8968128204345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.2,0.9012607574462891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,4,power_law_1.01,0.9575807571411132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,4,balanced,0.5253173510233561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.2,1.147987174987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,1,power_law_1.2,14.584562683105469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,4,power_law_1.01,0.9453887939453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,1,power_law_1.2,1.3935744285583496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,4,balanced,1.6406720479329426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,4,power_law_1.01,0.995148754119873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,4,balanced,1.1113173166910808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,4,power_law_1.01,0.9790528297424317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,4,balanced,0.5604373216629028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,1,power_law_1.2,2.1105600357055665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,4,power_law_1.01,1.0051199913024902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,4,power_law_1.01,1.1102656364440917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,4,power_law_1.01,1.1726143836975098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,1,power_law_1.2,4.020729446411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,4,balanced,1.8445760409037273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,4,balanced,0.6330560048421224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,4,power_law_1.01,1.263980770111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,4,balanced,1.3092319965362549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,4,power_law_1.01,1.4214143753051758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,4,power_law_1.01,1.479967975616455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,4,balanced,0.7081706523895264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,4,power_law_1.01,1.6688512802124023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,4,power_law_1.01,1.5781696319580079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,4,power_law_1.01,1.949625587463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,4,power_law_1.01,2.3345279693603516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,4,balanced,2.069354693094889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,4,balanced,1.5953013102213542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,4,balanced,0.8390293121337891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,4,power_law_1.01,2.906284713745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,4,power_law_1.01,3.438854217529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,4,power_law_1.01,4.151136016845703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,4,balanced,1.0678026676177979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,4,power_law_1.01,6.110835266113281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,4,power_law_1.01,10.88079376220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,4,balanced,2.1098666191101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,4,balanced,2.8709227244059243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,4,balanced,1.3262879848480225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,4,power_law_1.01,0.09312639832496643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,4,balanced,2.609269301096598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,4,power_law_1.01,0.1424512028694153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,4,balanced,3.588063875834147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,4,power_law_1.01,0.24277760982513427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,4,balanced,1.770645300547282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,4,power_law_1.01,0.15168639421463012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,4,power_law_1.01,0.19790079593658447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,4,power_law_1.01,0.27287681102752687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,4,power_law_1.01,0.3339904069900513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,4,power_law_1.01,0.4156032085418701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,4,power_law_1.01,0.45145602226257325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,4,power_law_1.01,0.4506688117980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,4,power_law_1.01,0.493177604675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,4,power_law_1.01,0.5137343883514405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,4,power_law_1.01,0.5190720081329345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,4,power_law_1.01,0.5280640125274658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,4,power_law_1.01,0.5341504096984864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,4,balanced,2.210223992665609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,4,power_law_1.01,0.564902400970459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,4,power_law_1.01,0.5896255970001221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,4,power_law_1.01,0.6383808135986329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,4,balanced,3.75053342183431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,4,power_law_1.01,0.7032000064849854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,4,power_law_1.01,0.8235263824462891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,4,power_law_1.01,0.8644288063049317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,4,balanced,4.610149383544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,4,power_law_1.01,0.9905280113220215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,4,power_law_1.01,1.0438783645629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,4,power_law_1.01,1.308396816253662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,4,power_law_1.01,1.5297344207763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,4,power_law_1.01,2.068844795227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,4,power_law_1.01,0.08888319730758668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,4,power_law_1.01,2.477868843078613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,4,power_law_1.01,0.1140671968460083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,4,power_law_1.01,3.1001407623291017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,4,power_law_1.01,0.17304960489273072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,4,power_law_1.01,0.11059199571609497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,4,power_law_1.01,4.5123649597167965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,4,power_law_1.01,0.15867520570755006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,4,power_law_1.01,0.2088320016860962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,4,power_law_1.01,0.22960000038146972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,4,power_law_1.01,8.260838317871094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,4,power_law_1.01,0.29351680278778075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,4,balanced,3.2751998901367188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,4,power_law_1.01,0.30473599433898924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,4,power_law_1.01,0.32071681022644044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,4,power_law_1.01,0.3303679943084717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,4,power_law_1.01,0.35034880638122556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,4,power_law_1.01,0.3471807956695557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,4,power_law_1.01,0.3531968116760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,4,power_law_1.01,0.3680448055267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,4,power_law_1.01,0.3745919942855835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,4,power_law_1.01,0.40049920082092283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,4,power_law_1.01,0.4380928039550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,4,power_law_1.01,0.45224318504333494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,4,power_law_1.01,0.512179183959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,4,power_law_1.01,0.5747776031494141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,4,power_law_1.01,0.7262527942657471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,4,power_law_1.01,0.7438079833984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,4,power_law_1.01,0.9425024032592774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,4,power_law_1.01,1.2454336166381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,4,power_law_1.01,1.6603263854980468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,4,power_law_1.01,2.0081600189208983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,4,power_law_1.01,2.4454336166381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,4,balanced,7.18122673034668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,4,power_law_1.01,3.765331268310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,4,balanced,8.935888290405273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,4,power_law_1.01,7.176870727539063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,4,power_law_1.2,0.1483839988708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,4,power_law_1.2,0.09427840113639832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,4,power_law_1.2,0.24086399078369142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,4,power_law_1.2,0.14391679763793946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,4,power_law_1.2,0.20399999618530273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,4,power_law_1.2,0.15667200088500977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,4,balanced,6.29319445292155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,4,power_law_1.2,0.22736001014709473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,4,power_law_1.2,0.3259648084640503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,4,power_law_1.2,0.11917439699172974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,4,power_law_1.2,0.452345609664917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,4,power_law_1.2,0.19293440580368043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,4,power_law_1.2,0.5191743850708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,4,power_law_1.2,0.2618688106536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,4,power_law_1.2,0.7468992233276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,4,power_law_1.2,0.31642880439758303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,4,power_law_1.2,0.7744703769683838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,4,power_law_1.2,0.39743359088897706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,4,power_law_1.2,0.833407974243164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,4,power_law_1.2,0.46681599617004393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,4,power_law_1.2,0.8505727767944335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,4,power_law_1.2,0.44249601364135743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,4,power_law_1.2,0.9165632247924804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,4,power_law_1.2,0.45676798820495607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,4,power_law_1.2,0.8942079544067383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,4,power_law_1.2,0.9787263870239258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,4,power_law_1.2,0.47481598854064944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,4,power_law_1.2,0.9976320266723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,4,power_law_1.2,0.0900223970413208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,4,power_law_1.2,0.5119872093200684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,4,power_law_1.2,0.5202688217163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,4,power_law_1.2,1.0284607887268067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,4,power_law_1.2,0.1171072006225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,4,power_law_1.2,0.5477312088012696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,4,power_law_1.2,0.12305920124053955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,4,power_law_1.2,1.1164480209350587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,4,power_law_1.2,0.5708672046661377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,4,power_law_1.2,0.11369600296020507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,4,power_law_1.2,1.2004863739013671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,4,power_law_1.2,0.16213760375976563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,4,power_law_1.2,0.5996799945831299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,4,power_law_1.2,1.35349760055542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,4,power_law_1.2,0.1998271942138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,4,power_law_1.2,0.6446335792541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,4,power_law_1.2,1.5255871772766114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,4,power_law_1.2,0.20913279056549072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,4,power_law_1.2,1.5813440322875976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,4,power_law_1.2,0.7357696056365967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,4,power_law_1.2,0.2776128053665161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,4,power_law_1.2,1.8119808197021485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,4,power_law_1.2,0.2841088056564331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,4,power_law_1.2,0.8405759811401368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,4,power_law_1.2,1.675846481323242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,4,power_law_1.2,0.29526400566101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,4,power_law_1.2,2.095609664916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,4,power_law_1.2,0.8751232147216796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,4,power_law_1.2,0.32045440673828124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,4,power_law_1.2,2.4529087066650392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,4,power_law_1.2,1.079148769378662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,4,power_law_1.2,0.33431038856506345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,4,power_law_1.2,3.1691072463989256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,4,power_law_1.2,1.096889591217041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,4,power_law_1.2,0.3383359909057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,4,power_law_1.2,3.9574913024902343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,4,power_law_1.2,1.3588671684265137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,4,power_law_1.2,0.35831680297851565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,4,power_law_1.2,1.6577152252197265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,4,power_law_1.2,0.3654592037200928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,4,power_law_1.2,4.61253776550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,4,power_law_1.2,0.3792320013046265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,4,power_law_1.2,2.076102447509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,4,power_law_1.2,6.1132865905761715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,4,power_law_1.2,0.3997760057449341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,4,power_law_1.2,2.678156852722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,4,power_law_1.2,0.4441984176635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,4,power_law_1.2,12.390144348144531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,4,power_law_1.2,0.46606078147888186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,4,power_law_1.2,3.30830078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,4,power_law_1.2,0.54519681930542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,4,power_law_1.2,4.738803100585938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,4,power_law_1.2,0.6228479862213134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,4,power_law_1.2,0.7599487781524659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,4,power_law_1.2,0.8350272178649902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,4,power_law_1.2,10.37088623046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,4,power_law_1.2,1.0290111541748046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,4,power_law_1.2,1.3492287635803222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,4,power_law_1.2,1.8356735229492187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,4,power_law_1.2,2.155353546142578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,4,power_law_1.2,3.0005184173583985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,4,power_law_1.2,4.020780944824219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,4,power_law_1.2,8.043974304199219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,2,balanced,0.0672159989674886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,2,balanced,0.09268800417582194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,2,balanced,0.1425920029481252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,2,balanced,0.2429973284403483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,2,balanced,0.44369598229726154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,2,balanced,0.8420213063557943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,2,balanced,1.244319995244344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,2,balanced,1.2329866886138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,2,balanced,1.2382720311482747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,2,balanced,1.2364799976348877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,2,balanced,1.236026684443156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,2,balanced,1.244879961013794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,2,balanced,1.2507359981536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,2,balanced,1.2541653315226238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,2,balanced,1.2770346800486247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,2,balanced,1.2691786289215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,2,balanced,1.290277322133382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,2,balanced,1.3106079896291096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,2,balanced,0.05398933092753092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,2,balanced,1.3405866622924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,2,balanced,0.07051733136177063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,2,balanced,0.06631466746330261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,2,balanced,0.09748799602190654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,2,balanced,0.07434133191903432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,2,balanced,1.3982346852620442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,2,balanced,0.1492906709512075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,2,balanced,0.09578667084376018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,2,balanced,0.2502773404121399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,2,balanced,0.13266133268674216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,2,balanced,0.44818135102589923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,2,balanced,0.20069332917531332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,2,balanced,1.4512960116068523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,2,balanced,0.6444693406422933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,2,balanced,0.32443734010060626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,2,balanced,0.6489973465601603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,2,balanced,0.45474668343861896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,2,balanced,0.4542773167292277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,2,balanced,0.6529333194096884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,2,balanced,0.45905065536499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,2,balanced,0.6533493200937907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,2,balanced,1.540437380472819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,2,balanced,0.46323732535044354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,2,balanced,0.6585386594136556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,2,balanced,0.4648373524347941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,2,balanced,0.671557346979777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,2,balanced,0.473200003306071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,2,balanced,0.667365312576294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,2,balanced,0.4776800076166789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,2,power_law_1.01,0.09183359742164612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,2,balanced,0.6781333287556967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,2,balanced,0.47948265075683594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,2,balanced,1.6368160247802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,2,balanced,0.48627734184265137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,2,power_law_1.01,0.14151040315628052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,2,balanced,0.6929919719696045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,2,balanced,0.49855999151865643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,2,power_law_1.01,0.24010241031646729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,2,power_law_1.01,0.06677119731903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,2,balanced,0.6957706610361735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,2,balanced,1.8344799677530925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,2,balanced,0.5158986647923788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,2,power_law_1.01,0.18764159679412842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,2,power_law_1.01,0.09492480158805847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,2,balanced,0.7131306330362955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,2,balanced,0.5395093361536661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,2,power_law_1.01,0.34513919353485106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,2,power_law_1.01,0.144159996509552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,2,balanced,0.745199998219808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,2,balanced,0.5695893367131551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,2,power_law_1.01,0.5161280155181884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,2,power_law_1.01,0.1115839958190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,2,balanced,2.1058506965637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,2,power_law_1.01,0.6061759948730469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,2,balanced,0.7691360314687093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,2,balanced,0.6310293277104696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,2,power_law_1.01,0.19835519790649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,2,power_law_1.01,0.7824128150939942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,2,power_law_1.01,0.28078720569610593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,2,balanced,0.8246453603108724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,2,power_law_1.01,0.8132736206054687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,2,balanced,0.6684746742248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,2,power_law_1.01,0.328108811378479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,2,power_law_1.01,0.8536255836486817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,2,power_law_1.01,0.4187903881072998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,2,balanced,0.8772213459014893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,2,power_law_1.01,0.8926655769348144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,2,power_law_1.01,0.4429632186889648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,2,balanced,0.77347731590271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,2,power_law_1.01,0.9341504096984863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,2,power_law_1.01,0.45661439895629885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,2,balanced,0.983674685160319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,2,power_law_1.01,0.9348416328430176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,2,power_law_1.01,0.4648320198059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,2,power_law_1.01,0.9710911750793457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,2,power_law_1.01,0.4948160171508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,2,balanced,0.8951199849446615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,2,balanced,2.4391627311706543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,2,power_law_1.01,1.0207039833068847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,2,balanced,1.0991253058115642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,2,power_law_1.01,0.5208191871643066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,2,power_law_1.01,1.0473088264465331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,2,power_law_1.01,0.5449984073638916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,2,power_law_1.01,1.075321578979492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,2,power_law_1.01,0.564524793624878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,2,balanced,1.0856373310089111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,2,power_law_1.01,0.07351679801940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,2,power_law_1.01,1.152947235107422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,2,power_law_1.01,0.5796031951904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,2,balanced,1.3223893642425537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,2,power_law_1.01,1.2600319862365723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,2,power_law_1.01,0.09653760194778442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,2,power_law_1.01,0.6124032020568848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,2,power_law_1.01,1.406278419494629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,2,power_law_1.01,0.6697855949401855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,2,power_law_1.01,0.1283519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,2,power_law_1.01,0.7117248058319092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,2,power_law_1.01,1.4609855651855468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,2,power_law_1.01,0.11550079584121704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,2,power_law_1.01,0.8262911796569824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,2,power_law_1.01,1.6410432815551759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,2,balanced,1.4330986340840657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,2,power_law_1.01,0.16615040302276612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,2,power_law_1.01,0.8957695960998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,2,balanced,1.6232479413350422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,2,power_law_1.01,0.22215681076049804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,2,power_law_1.01,1.692300796508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,2,balanced,3.282165209452311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,2,power_law_1.01,1.0773759841918946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,2,power_law_1.01,0.2508415937423706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,2,power_law_1.01,1.9544960021972657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,2,power_law_1.01,1.1446335792541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,2,power_law_1.01,0.31281280517578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,2,power_law_1.01,2.420876884460449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,2,power_law_1.01,0.3323263883590698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,2,power_law_1.01,1.4619711875915526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,2,power_law_1.01,0.3396352052688599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,2,power_law_1.01,3.0230207443237305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,2,power_law_1.01,1.7869312286376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,2,power_law_1.01,0.3464639902114868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,2,power_law_1.01,2.4246912002563477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,2,power_law_1.01,4.2453056335449215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,2,balanced,1.854522705078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,2,power_law_1.01,0.36136319637298586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,2,balanced,2.114544073740641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,2,power_law_1.01,2.974278450012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,2,power_law_1.01,0.3761023998260498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,2,power_law_1.2,0.09351040124893188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,2,power_law_1.01,4.73135986328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,2,power_law_1.01,0.389356803894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,2,power_law_1.01,3.6228286743164064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,2,power_law_1.2,0.1419711947441101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,2,power_law_1.01,0.4109248161315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,2,power_law_1.01,6.27757453918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,2,power_law_1.2,0.06886399984359741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,2,power_law_1.2,0.15128320455551147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,2,power_law_1.01,0.4238592147827148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,2,power_law_1.01,5.405785751342774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,2,power_law_1.01,0.44161281585693357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,2,power_law_1.2,0.1903040051460266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,2,power_law_1.2,0.096697598695755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,2,balanced,3.792895952860514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,2,power_law_1.01,0.48180480003356935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,2,power_law_1.2,0.32454400062561034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,2,power_law_1.01,11.939936065673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,2,power_law_1.2,0.09493119716644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,2,power_law_1.01,10.455379486083984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,2,power_law_1.01,0.5338111877441406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,2,power_law_1.2,0.4224703788757324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,2,power_law_1.2,0.11635839939117432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,2,power_law_1.01,0.6311359882354737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,2,power_law_1.2,0.5220863819122314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,2,balanced,2.4832213719685874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,2,power_law_1.2,0.18239359855651854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,2,power_law_1.01,0.6632319927215576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,2,balanced,2.744661331176758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,2,power_law_1.2,0.7770815849304199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,2,power_law_1.2,0.23955199718475342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,2,power_law_1.01,0.8311936378479003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,2,power_law_1.2,0.7722047805786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,2,power_law_1.2,0.29523839950561526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,2,power_law_1.01,0.9858431816101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,2,power_law_1.2,0.7985536098480225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,2,power_law_1.2,0.390502405166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,2,power_law_1.01,1.273465633392334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,2,power_law_1.2,0.8561152458190918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,2,power_law_1.2,0.4295231819152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,2,power_law_1.01,1.6030784606933595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,2,power_law_1.2,0.8677248001098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,2,power_law_1.2,0.43856000900268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,2,power_law_1.2,0.47027201652526857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,2,power_law_1.2,0.9242176055908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,2,power_law_1.01,2.198259162902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,2,power_law_1.2,0.4869376182556152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,2,power_law_1.2,0.9829055786132812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,2,power_law_1.01,2.5685056686401366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,2,power_law_1.2,1.0139264106750487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,2,power_law_1.2,0.5104640007019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,2,balanced,3.1226399739583335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,2,power_law_1.01,3.224787139892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,2,balanced,3.470063845316569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,2,power_law_1.2,1.0139583587646483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,2,power_law_1.2,0.5485055923461915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,2,power_law_1.01,4.798668670654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,2,power_law_1.2,0.5634943962097168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,2,power_law_1.2,1.1088576316833496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,2,balanced,5.584378560384114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,2,power_law_1.2,1.1840448379516602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,2,power_law_1.2,0.5769536018371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,2,power_law_1.01,9.289600372314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,2,power_law_1.2,1.2901439666748047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,2,power_law_1.2,0.6443007946014404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,2,power_law_1.2,1.4968192100524902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,2,power_law_1.2,0.6938432216644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,2,power_law_1.2,0.7349567890167237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,2,power_law_1.2,1.510489559173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,2,power_law_1.2,0.8360960006713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,2,power_law_1.2,1.7877056121826171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,2,power_law_1.2,0.9189567565917969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,2,power_law_1.2,1.7756223678588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,2,power_law_1.2,1.117459201812744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,2,power_law_1.2,2.1259328842163088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,2,power_law_1.2,1.2003583908081055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,2,power_law_1.2,2.5754432678222656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,2,power_law_1.2,1.5838399887084962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,2,balanced,5.045210520426433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,2,power_law_1.2,3.2296833038330077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,2,power_law_1.2,1.9389823913574218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,2,balanced,4.605546633402507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,2,power_law_1.2,2.4904640197753904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,2,power_law_1.2,4.404185485839844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,2,power_law_1.2,3.1123647689819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,2,power_law_1.2,5.1755329132080075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,2,power_law_1.2,3.797158432006836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,2,power_law_1.2,6.835654449462891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,2,power_law_1.2,5.571596908569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,2,power_law_1.2,12.67844467163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,2,power_law_1.2,11.190156555175781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,2,balanced,10.694170633951822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,2,balanced,9.652634938557943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,2,balanced,8.779786427815756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,1,balanced,0.07100800176461537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,1,balanced,0.09493333101272583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,2,power_law_1.2,0.07107200026512146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,1,balanced,0.14596266547838846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,1,balanced,0.2474720080693563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,2,power_law_1.2,0.0936959981918335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,1,balanced,0.4468746582667033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,2,power_law_1.2,0.09717119932174682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,1,balanced,0.8385759989420573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,2,power_law_1.2,0.1182528018951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,1,balanced,1.237674633661906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,2,power_law_1.2,0.15627520084381102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,1,balanced,1.2430933316548665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,2,power_law_1.2,0.20135679244995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,1,balanced,1.2464426358540852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,2,power_law_1.2,0.23573760986328124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,1,balanced,1.2480639616648357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,2,power_law_1.2,0.29868159294128416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,1,balanced,1.2551466623942058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,1,balanced,1.2686560153961182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,2,power_law_1.2,0.31441919803619384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,1,balanced,1.273045301437378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,2,power_law_1.2,0.3137408018112183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,1,balanced,1.2884373664855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,2,power_law_1.2,0.34259839057922364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,1,balanced,1.2989333470662434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,2,power_law_1.2,0.3652928113937378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,1,balanced,1.3031093279520671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,2,power_law_1.2,0.36135039329528806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,1,balanced,0.060506666700045265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,1,balanced,1.3226719697316487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,2,power_law_1.2,0.396230411529541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,1,balanced,0.07451733450094859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,1,balanced,1.3692213694254558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,1,balanced,0.10316266616185506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,2,power_law_1.2,0.4127488136291504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,1,balanced,0.15768000483512878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,2,power_law_1.2,0.4237823963165283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,1,balanced,0.266485333442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,1,balanced,1.414207935333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,2,power_law_1.2,0.44471039772033694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,1,balanced,0.4639146725336711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,1,balanced,0.674021323521932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,2,power_law_1.2,0.4846911907196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,1,balanced,1.5113174120585124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,1,balanced,0.6757493019104004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,2,power_law_1.2,0.5478335857391358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,1,balanced,0.6770133177439371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,2,power_law_1.2,0.6539328098297119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,1,balanced,1.5743306477864583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,1,balanced,0.6835947036743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,2,power_law_1.2,0.688159990310669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,1,balanced,0.69541335105896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,1,balanced,0.7011679808298746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,2,power_law_1.2,0.836128044128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,1,balanced,1.7337387402852376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,1,balanced,0.7058346271514893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,2,power_law_1.2,1.0268735885620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,1,balanced,0.7160267035166422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,2,power_law_1.2,1.341267204284668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,1,balanced,0.7225386301676432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,1,balanced,0.7364746729532877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,2,power_law_1.2,1.611622428894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,1,balanced,1.8621279398600261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,1,balanced,0.7546666463216146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,2,power_law_1.2,2.2986751556396485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,1,balanced,0.8037813504536947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,2,power_law_1.2,2.6782720565795897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,1,balanced,0.8424533208211263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,2,power_law_1.2,3.2620094299316404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,1,balanced,2.217679977416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,1,balanced,0.9277386665344238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,2,power_law_1.2,5.121299362182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,1,balanced,1.0103200276692708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,2,power_law_1.2,9.362592315673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,1,balanced,1.1767733097076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,1,balanced,2.6591572761535645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,1,balanced,1.4067254066467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,1,balanced,0.09694400429725647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,1,power_law_1.01,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,1,balanced,0.10542933146158855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,1,balanced,0.12782933314641318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,1,balanced,0.17605332533518472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,1,balanced,1.758080005645752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,1,power_law_1.01,0.09532160162925721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,1,balanced,3.3339306513468423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,1,balanced,0.263264000415802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,1,power_law_1.01,0.1429759979248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,1,balanced,0.43115198612213135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,1,power_law_1.01,0.17431039810180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,1,balanced,0.5960906744003296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,1,balanced,0.5867306788762411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,1,power_law_1.01,0.3265599966049194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,1,balanced,0.5920159816741943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,1,power_law_1.01,0.5107711791992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,1,balanced,0.5880906581878662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,1,balanced,0.5952746470769247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,1,power_law_1.01,0.5895423889160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,1,power_law_1.01,0.052070397138595584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,1,balanced,0.6014666557312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,1,balanced,2.2446773846944175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,1,power_law_1.01,0.7570367813110351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,1,balanced,0.606223980585734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,1,power_law_1.01,0.07090560197830201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,1,power_law_1.01,0.8172800064086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,1,balanced,0.6160159905751547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,1,power_law_1.01,0.09807360172271729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,1,power_law_1.01,0.8446720123291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,1,balanced,0.622053345044454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,1,balanced,0.630570650100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,1,power_law_1.01,0.12108800411224366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,1,power_law_1.01,0.879212760925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,1,balanced,0.6489866574605306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,1,power_law_1.01,0.19838720560073853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,1,power_law_1.01,0.9385984420776368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,1,balanced,0.7001866499582926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,1,power_law_1.01,0.298361611366272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,1,balanced,4.478512128194173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,1,power_law_1.01,0.9596223831176758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,1,power_law_1.01,0.3409600019454956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,1,balanced,0.73416535059611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,1,power_law_1.01,0.981446361541748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,1,power_law_1.01,0.43331198692321776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,1,power_law_1.01,1.0390080451965331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,1,balanced,0.8217493693033854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,1,balanced,2.923466682434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,1,power_law_1.01,0.4705023765563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,1,power_law_1.01,1.0420160293579102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,1,balanced,0.8938453197479248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,1,power_law_1.01,1.0994048118591309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,1,power_law_1.01,0.47728638648986815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,1,power_law_1.01,1.1796159744262695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,1,power_law_1.01,0.503052806854248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,1,power_law_1.01,1.2808704376220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,1,balanced,1.0664479732513428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,1,power_law_1.01,0.5344192028045655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,1,power_law_1.01,1.493273639678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,1,power_law_1.01,0.5503935813903809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,1,power_law_1.01,1.5300479888916017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,1,power_law_1.01,0.5614272117614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,1,balanced,1.270906686782837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,1,power_law_1.01,1.8045951843261718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,1,power_law_1.01,0.5819519996643067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,1,power_law_1.01,0.08539519906044006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,1,power_law_1.01,0.5982975959777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,1,balanced,5.355034510294597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,1,power_law_1.01,2.0930047988891602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,1,power_law_1.01,0.6559296131134034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,1,balanced,3.9966293970743814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,1,power_law_1.2,0.07082880139350892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,1,power_law_1.01,0.09416959881782531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,1,power_law_1.01,2.6032896041870117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,1,power_law_1.01,0.7077055931091308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,1,balanced,1.6377545992533367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,1,power_law_1.01,0.12095359563827515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,1,power_law_1.2,0.09473919868469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,1,power_law_1.01,0.7872000217437745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,1,power_law_1.01,2.831219291687012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,1,power_law_1.01,0.1440384030342102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,1,power_law_1.2,0.11372159719467163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,1,power_law_1.01,0.9266304016113281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,1,power_law_1.01,0.2064448118209839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,1,power_law_1.2,0.19505280256271362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,1,power_law_1.01,3.664595031738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,1,power_law_1.01,1.0046400070190429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,1,power_law_1.2,0.2935744047164917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,1,power_law_1.01,0.27832319736480715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,1,power_law_1.01,1.2546367645263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,1,power_law_1.2,0.4141632080078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,1,power_law_1.01,0.3208319902420044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,1,power_law_1.01,5.005292892456055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,1,power_law_1.2,0.5136640071868896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,1,balanced,2.157402674357096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,1,power_law_1.01,1.4325119972229003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,1,power_law_1.01,0.4106304168701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,1,power_law_1.2,0.7283584117889405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,1,power_law_1.01,5.91723518371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,1,power_law_1.01,1.8516992568969726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,1,balanced,5.045429229736328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,1,power_law_1.01,2.369683265686035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,1,power_law_1.2,0.8048959732055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,1,power_law_1.01,0.4193215847015381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,1,power_law_1.01,8.22537612915039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,1,power_law_1.2,0.8128512382507325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,1,power_law_1.01,0.42467198371887205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,1,power_law_1.01,3.1813888549804688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,1,power_law_1.2,0.8715519905090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,1,power_law_1.01,0.4575808048248291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,1,power_law_1.01,4.155398559570313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,1,power_law_1.2,0.9254528045654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,1,balanced,2.898101488749186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,1,power_law_1.01,0.46863999366760256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,1,power_law_1.01,15.411891174316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,1,power_law_1.2,0.9264575958251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,1,power_law_1.01,0.48183679580688477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,1,power_law_1.01,5.0247550964355465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,1,power_law_1.2,0.9918208122253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,1,power_law_1.01,0.5063744068145752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,1,balanced,7.8164323170979815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,1,power_law_1.2,1.0226112365722657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,1,power_law_1.01,0.5084415912628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,1,power_law_1.01,7.369868469238281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,1,power_law_1.2,1.0780351638793946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,1,power_law_1.01,0.528172779083252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,1,power_law_1.2,1.147225570678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,1,power_law_1.01,0.5555840015411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,1,power_law_1.2,1.2164544105529784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,1,power_law_1.01,13.957682800292968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,1,power_law_1.01,0.6155392169952393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,1,power_law_1.2,1.337279987335205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.01,0.676691198348999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,1,power_law_1.2,1.5597375869750976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.01,0.8003711700439453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,1,power_law_1.2,1.5691583633422852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,1,balanced,3.9512532552083335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.01,0.9088128089904786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,1,power_law_1.2,1.8884992599487305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.01,1.1334848403930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,1,power_law_1.2,2.238489532470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.01,1.2928319931030274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,1,power_law_1.2,2.7878719329833985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.01,1.7217023849487305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,1,power_law_1.2,3.0013759613037108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,1,power_law_1.01,2.2599679946899416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,1,power_law_1.2,3.829056167602539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,1,balanced,7.245434443155925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,1,power_law_1.01,3.0717952728271483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,1,power_law_1.2,5.226918411254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,1,power_law_1.01,4.04382095336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,1,power_law_1.2,6.1450752258300785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,1,power_law_1.01,4.847916793823242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,1,power_law_1.2,8.482694244384765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,1,power_law_1.01,7.429714965820312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,1,balanced,4.839818636576335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,1,power_law_1.2,15.685958862304688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,1,power_law_1.01,13.6761474609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,1,balanced,14.958260854085287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,1,balanced,7.388826370239258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,1,power_law_1.2,0.05219200253486633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,1,power_law_1.2,0.0716863989830017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,1,power_law_1.2,0.08056319952011108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,1,power_law_1.2,0.12282240390777588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,1,power_law_1.2,0.18309760093688965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,1,balanced,13.94693374633789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,1,power_law_1.2,0.25250558853149413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,1,power_law_1.2,0.30561280250549316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,1,power_law_1.2,0.42697601318359374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,1,power_law_1.2,0.44430718421936033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,1,power_law_1.2,0.46231679916381835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,1,power_law_1.2,0.49645438194274905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,1,power_law_1.2,0.506924819946289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,1,power_law_1.2,0.5224959850311279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,1,power_law_1.2,0.5476863861083985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,1,power_law_1.2,0.5773183822631835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,1,power_law_1.2,0.6175936222076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,1,power_law_1.2,0.6520639896392822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,1,power_law_1.2,0.732371187210083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,1,power_law_1.2,0.8066623687744141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,1,power_law_1.2,0.9609536170959473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,1,power_law_1.2,1.0459903717041015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,1,power_law_1.2,1.3059328079223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,1,power_law_1.2,1.482630443572998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,1,power_law_1.2,1.8973247528076171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,1,power_law_1.2,2.4382720947265626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,1,balanced,13.941572825113932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,1,power_law_1.2,3.2328895568847655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,1,power_law_1.2,4.218086242675781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,1,power_law_1.2,5.105913543701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,1,power_law_1.2,7.445471954345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,1,power_law_1.2,14.059577941894531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,1,power_law_1.2,0.08966400027275086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,1,power_law_1.2,0.0973695993423462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,1,power_law_1.2,0.11299200057983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,1,power_law_1.2,0.14728959798812866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,1,power_law_1.2,0.1947391986846924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,1,power_law_1.2,0.2566015958786011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,1,power_law_1.2,0.29782400131225584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,1,power_law_1.2,0.39701759815216064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,1,power_law_1.2,0.40254721641540525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,1,power_law_1.2,0.4166912078857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,1,power_law_1.2,0.44311041831970216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,1,power_law_1.2,0.4646463871002197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,8,balanced,0.03147733211517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,1,power_law_1.2,0.48535680770874023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,1,power_law_1.2,0.4908607959747314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,8,balanced,0.033376000821590424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,8,balanced,0.051141331593195595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,1,power_law_1.2,0.5144320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,8,balanced,0.07853866616884868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,1,power_law_1.2,0.5338304042816162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,8,balanced,0.12756266196568808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,8,balanced,0.12106666962305705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,1,power_law_1.2,0.5634687900543213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,8,balanced,0.12072533369064331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,8,balanced,0.12317867080370586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,1,power_law_1.2,0.6234816074371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,8,balanced,0.1220266620318095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.2,0.6988287925720215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,8,balanced,0.12201600273450215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,8,balanced,0.12397332986195882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.2,0.820531177520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,8,balanced,0.1234773298104604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.2,0.9392383575439454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,8,balanced,0.1258080005645752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.2,1.1678336143493653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,8,balanced,0.12593600153923035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.2,1.3299263954162597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,8,balanced,0.12945066889127096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.2,1.7412736892700196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,8,balanced,0.1311840017636617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,1,power_law_1.2,2.298361587524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,8,balanced,0.13475199540456137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,8,balanced,0.13942933082580566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,1,power_law_1.2,3.1075136184692385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,8,balanced,0.02977066735426585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,8,balanced,0.14443199833234152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,1,power_law_1.2,4.086368179321289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,8,balanced,0.03139200061559677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,8,balanced,0.03348266581694285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,8,balanced,0.15245866775512695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,8,balanced,0.056847999493281044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,1,power_law_1.2,4.9157054901123045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,8,balanced,0.07836266855398814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,8,balanced,0.07933866480986278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,8,balanced,0.16326933105786642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,8,balanced,0.08247999846935272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,1,power_law_1.2,7.473267364501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,8,balanced,0.08006399869918823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,8,balanced,0.08089600006739299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,8,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,8,balanced,0.08044266700744629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,8,balanced,0.19534399112065634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,8,balanced,0.0366239994764328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,8,balanced,0.08290666838486989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,1,power_law_1.2,14.128915405273437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,8,balanced,0.037578667203585304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,8,balanced,0.084197332461675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,8,balanced,0.04781866570313772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,8,balanced,0.06614933411280315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,8,balanced,0.08478400111198425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,8,balanced,0.06771733363469441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,8,balanced,0.08630399902661641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,8,balanced,0.0691840002934138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,8,balanced,0.21218132972717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,8,balanced,0.06961066524187724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,8,balanced,0.0957493285338084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,8,balanced,0.06821333368619283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,8,balanced,0.09574400385220845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,8,balanced,0.06944533189137776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,8,balanced,0.07049066821734111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,8,balanced,0.09544533491134644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,8,balanced,0.07220800220966339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,8,balanced,0.10181333621342976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,8,balanced,0.07146133482456207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,8,balanced,0.02327466756105423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,8,balanced,0.2627679904301961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,8,balanced,0.07477333148320515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,8,balanced,0.11157866319020589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,8,balanced,0.08057599763075511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,8,balanced,0.02526933451493581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,8,balanced,0.07980800171693166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,8,balanced,0.026704000929991405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,8,balanced,0.12657599647839865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,8,balanced,0.08389332890510559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,8,balanced,0.03182933231194814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,8,balanced,0.047824000318845115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,8,balanced,0.08898666501045227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,8,balanced,0.13822933038075766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,8,balanced,0.05017066498597463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,8,balanced,0.06251200040181477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,8,balanced,0.30710933605829877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,8,balanced,0.10442133744557698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,8,balanced,0.057445332407951355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,8,balanced,0.09008000294367473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,8,balanced,0.17322667439778647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,8,balanced,0.08490133285522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,8,balanced,0.11008000373840332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,8,balanced,0.07746666669845581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,8,balanced,0.09925867120424907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,8,balanced,0.1163200040658315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,8,balanced,0.10025599598884583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,8,balanced,0.19834667444229126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,8,balanced,0.09989333152770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,8,balanced,0.11392000317573547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,8,balanced,0.14642666776974997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,8,balanced,0.111653337876002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,8,balanced,0.3991893529891968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,8,balanced,0.10648000240325928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,8,balanced,0.2641119956970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,8,balanced,0.11521599690119426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,8,balanced,0.16811732451121011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,8,balanced,0.13058666388193765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,8,power_law_1.01,0.07639679908752442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,8,power_law_1.01,0.054764801263809205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,8,balanced,0.14536000291506448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,8,power_law_1.01,0.058387202024459836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,8,power_law_1.01,0.06808320283889771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,8,balanced,0.32469866673151654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,8,balanced,0.16435199975967407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,8,power_law_1.01,0.08004480004310607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,8,balanced,0.21892799933751425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,8,power_law_1.01,0.10400639772415161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,8,power_law_1.01,0.11409280300140381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,8,power_law_1.01,0.1164736032485962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,8,balanced,0.19130667050679526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,8,balanced,0.509066661198934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,8,power_law_1.01,0.1141055941581726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,8,power_law_1.01,0.11850240230560302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,8,power_law_1.01,0.11742080450057983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,8,balanced,0.2686079939206441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,8,power_law_1.01,0.12124160528182984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,8,balanced,0.22007467349370322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,8,power_law_1.01,0.12458879947662353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,8,balanced,0.4466453393300374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,8,power_law_1.01,0.1325824022293091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,8,power_law_1.01,0.13761919736862183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,8,power_law_1.01,0.13845759630203247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,8,power_law_1.01,0.1397055983543396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,8,power_law_1.01,0.15865600109100342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,8,balanced,0.3412693341573079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,8,power_law_1.01,0.16787840127944947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,8,power_law_1.01,0.19409279823303222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,8,power_law_1.01,0.2102976083755493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,8,power_law_1.01,0.24491519927978517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,8,balanced,0.37135998407999676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,8,power_law_1.01,0.30371201038360596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,8,power_law_1.01,0.395686411857605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,8,balanced,0.6077599922815958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,8,power_law_1.01,0.05253120064735413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,8,power_law_1.01,0.46929278373718264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,8,balanced,0.3988320032755534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,8,power_law_1.01,0.6269248008728028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,8,power_law_1.01,0.03934719860553741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,8,balanced,0.5616639852523804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,8,power_law_1.01,0.8701824188232422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,8,power_law_1.01,0.04069119989871979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,8,power_law_1.01,0.052179199457168576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,8,power_law_1.01,0.9896320343017578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,8,power_law_1.01,0.057107198238372806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,8,power_law_1.01,0.07429760098457336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,8,power_law_1.01,1.6281152725219727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,8,power_law_1.01,0.07631360292434693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,8,power_law_1.01,0.07663999795913697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,8,power_law_1.01,3.421478271484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,8,power_law_1.01,0.07753599882125854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,8,power_law_1.01,0.07652480006217957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,8,power_law_1.01,0.08067200183868409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,8,power_law_1.01,0.04552960097789764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,8,balanced,0.4673279921213786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,8,power_law_1.01,0.082777601480484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,8,balanced,0.576906681060791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,8,power_law_1.01,0.08364160060882568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,8,power_law_1.01,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,8,power_law_1.01,0.09181439876556396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,8,power_law_1.01,0.09722880125045777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,8,power_law_1.01,0.03904640078544617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,8,power_law_1.01,0.10458240509033204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,8,power_law_1.01,0.05021439790725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,8,power_law_1.01,0.10083199739456176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,8,power_law_1.01,0.051609599590301515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,8,power_law_1.01,0.11345920562744141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,8,power_law_1.01,0.06493440270423889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,8,power_law_1.01,0.13116159439086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,8,power_law_1.01,0.06416640281677247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,8,balanced,0.68449600537618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,8,power_law_1.01,0.15708800554275512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,8,power_law_1.01,0.06599040031433105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,8,power_law_1.01,0.1674623966217041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,8,power_law_1.01,0.06734079718589783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,8,power_law_1.01,0.22320001125335692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,8,balanced,0.905461311340332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,8,power_law_1.01,0.06779519915580749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,8,balanced,0.7598293622334799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,8,power_law_1.01,0.2704384088516235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,8,power_law_1.01,0.06880000233650208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,8,power_law_1.01,0.374835205078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,8,power_law_1.01,0.07023360133171082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,8,balanced,0.5671306848526001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,8,power_law_1.01,0.07146880030632019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,8,power_law_1.01,0.46292481422424314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,8,power_law_1.01,0.07464320063591004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,8,power_law_1.01,0.7045055866241455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,8,power_law_1.01,0.08157439827919007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,8,power_law_1.01,0.910489559173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,8,power_law_1.01,0.08438400030136109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,8,power_law_1.01,1.0877568244934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,8,power_law_1.01,0.09102079868316651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,8,power_law_1.01,0.10451840162277222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,8,power_law_1.01,1.5671744346618652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.01,0.11887999773025512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.01,0.13945599794387817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,8,power_law_1.01,3.640006256103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.01,0.14913920164108277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.01,0.18938239812850952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.01,0.24071040153503417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.01,0.3046463966369629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.01,0.3909759998321533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.01,0.5635456085205078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,8,balanced,0.9459306399027506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.01,0.6897984027862549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,8,power_law_1.01,0.8752191543579102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,8,balanced,1.0461653073628743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,8,power_law_1.01,1.3275391578674316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,8,power_law_1.01,2.4929471969604493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,8,balanced,0.8633759816487631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,8,balanced,1.4958826700846355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,8,balanced,1.734816074371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,8,power_law_1.01,0.02625280022621155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,8,power_law_1.01,0.02484479993581772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,8,power_law_1.01,0.025939199328422546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,8,power_law_1.01,0.034143999218940735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,8,power_law_1.01,0.031718400120735166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,8,power_law_1.01,0.046009600162506104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,8,power_law_1.01,0.05324159860610962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,8,power_law_1.01,0.050470399856567386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,8,power_law_1.01,0.06863999962806702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,8,power_law_1.01,0.06872320175170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,8,power_law_1.01,0.06558079719543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,8,power_law_1.01,0.08543999791145325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,8,power_law_1.01,0.08555520176887513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,8,power_law_1.01,0.08403840065002441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,8,balanced,2.0243093172709146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,8,power_law_1.01,0.09800959825515747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,8,power_law_1.01,0.09805439710617066
TRTLLM,1.2.0rc5,NVIDIA B200,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 B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,8,power_law_1.01,0.08419200181961059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.01,0.0880511999130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.01,0.10760960578918458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.01,0.10958720445632934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.01,0.14342399835586547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,8,balanced,1.6577919324239094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.01,0.1762112021446228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.01,0.2337280035018921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.01,0.31208961009979247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.01,0.3928960084915161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.01,0.560537576675415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,8,power_law_1.01,0.6380224227905273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,8,power_law_1.01,1.0792384147644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,8,power_law_1.01,2.058304023742676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,8,balanced,2.9590934117635093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,8,power_law_1.2,0.07352960109710693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,8,power_law_1.2,0.04043520092964172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,8,power_law_1.2,0.04970879852771759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,8,power_law_1.2,0.06395519971847534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,8,power_law_1.2,0.07758079767227173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,8,power_law_1.2,0.10763520002365112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,8,power_law_1.2,0.11586560010910034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,8,power_law_1.2,0.11269760131835938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,8,power_law_1.2,0.1108288049697876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,8,power_law_1.2,0.12135039567947388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,8,power_law_1.2,0.12387839555740357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,8,power_law_1.2,0.12534400224685668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,8,power_law_1.2,0.12093440294265748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,8,power_law_1.2,0.13167359828948974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,8,power_law_1.2,0.14416639804840087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,8,power_law_1.2,0.15179519653320311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,8,power_law_1.2,0.14939520359039307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,8,power_law_1.2,0.159443199634552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,8,power_law_1.2,0.17182079553604127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,8,power_law_1.2,0.20896000862121583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,8,power_law_1.2,0.2284032106399536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,8,power_law_1.2,0.28394880294799807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,8,power_law_1.2,0.05175679922103882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,8,power_law_1.2,0.35434880256652834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,8,power_law_1.2,0.49811840057373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,8,power_law_1.2,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,8,power_law_1.2,0.6780992031097413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,8,power_law_1.2,0.04073599874973297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,8,power_law_1.2,0.051046401262283325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,8,power_law_1.2,0.8339072227478027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,8,power_law_1.2,0.05413119792938233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,8,power_law_1.2,1.0869888305664062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,8,power_law_1.2,0.07482240200042725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,8,power_law_1.2,1.529971218109131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,8,power_law_1.2,0.07660800218582153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,8,power_law_1.2,0.07701759934425353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,8,power_law_1.2,2.2164608001708985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,8,power_law_1.2,0.07683200240135193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,8,power_law_1.2,0.07918720245361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,8,power_law_1.2,0.0825984001159668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,8,power_law_1.2,4.728025436401367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,8,power_law_1.2,0.0838591992855072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,8,power_law_1.2,0.044684800505638125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,8,power_law_1.2,0.08819199800491333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,8,power_law_1.2,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,8,power_law_1.2,0.09084799885749817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,8,power_law_1.2,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,8,power_law_1.2,0.10298880338668823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,8,power_law_1.2,0.04586879909038544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,8,power_law_1.2,0.11203199625015259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,8,power_law_1.2,0.04995200037956238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,8,power_law_1.2,0.06331520080566407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,8,power_law_1.2,0.10670720338821411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,8,power_law_1.2,0.06512640118598938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,8,power_law_1.2,0.11809920072555542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,8,power_law_1.2,0.06554239988327026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,8,power_law_1.2,0.13890559673309327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,8,power_law_1.2,0.06842880249023438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,8,power_law_1.2,0.1667904019355774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,8,power_law_1.2,0.0670144021511078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,8,power_law_1.2,0.18227200508117675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,8,power_law_1.2,0.0695743978023529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,8,power_law_1.2,0.06925439834594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,8,power_law_1.2,0.24581758975982665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,8,power_law_1.2,0.07243520021438599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,8,power_law_1.2,0.31561601161956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,8,power_law_1.2,0.07620480060577392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,8,power_law_1.2,0.3835839986801147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,8,power_law_1.2,0.08204799890518188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,8,power_law_1.2,0.5856575965881348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,8,power_law_1.2,0.08610560297966004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,8,power_law_1.2,0.7983551979064941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,8,power_law_1.2,0.08999040126800537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,8,power_law_1.2,0.10428160429000854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,8,power_law_1.2,0.9381823539733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.2,0.12331520318984986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,8,power_law_1.2,1.3292927742004395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.2,0.14669439792633057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.2,0.1627135992050171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,8,power_law_1.2,2.1933759689331054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.2,0.1979200005531311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.2,0.2656192064285278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,8,power_law_1.2,4.044313430786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.2,0.35290238857269285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.2,0.44216318130493165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.2,0.613420820236206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.2,0.9761919975280762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,8,power_law_1.2,1.1667712211608887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,8,power_law_1.2,0.026009601354599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,8,power_law_1.2,1.836511993408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,8,power_law_1.2,0.021587200462818146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,8,power_law_1.2,0.02563199996948242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,8,power_law_1.2,3.687673568725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,8,power_law_1.2,0.02565760016441345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,8,power_law_1.2,0.031065601110458373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,8,power_law_1.2,0.045158401131629944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,8,power_law_1.2,0.052153599262237546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,4,balanced,0.0317546675602595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,8,power_law_1.2,0.05015680193901062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,8,power_law_1.2,0.07483519911766053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,8,power_law_1.2,0.0705344021320343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,4,balanced,0.03414933383464813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,8,power_law_1.2,0.0639743983745575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,4,balanced,0.054058666030565895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,8,power_law_1.2,0.08499839901924133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,4,balanced,0.07666133344173431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,8,power_law_1.2,0.08326399922370911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,4,balanced,0.12780267000198364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,8,power_law_1.2,0.08375679850578308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,4,balanced,0.12089600165685017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,8,power_law_1.2,0.10050560235977173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,4,balanced,0.12266666690508525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,8,power_law_1.2,0.09667199850082397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,4,balanced,0.03154666721820831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,4,balanced,0.12327466408411662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,8,power_law_1.2,0.09130240082740784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,8,power_law_1.2,0.08045439720153809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,4,balanced,0.12145066261291504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,4,balanced,0.0332640012105306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.2,0.09137279987335205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,4,balanced,0.1237333317597707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,4,balanced,0.03573333223660787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.2,0.10648959875106812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,4,balanced,0.1237493356068929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,4,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.2,0.12071679830551148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,4,balanced,0.08108266691366832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,4,balanced,0.12486400206883748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.2,0.15690879821777343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,4,balanced,0.08547733227411906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.2,0.1910272002220154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,4,balanced,0.12557866175969443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,4,balanced,0.08582933743794759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.2,0.268883204460144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,4,balanced,0.1274186670780182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,4,balanced,0.04170133173465729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,4,balanced,0.08403199911117554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.2,0.32065279483795167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,4,balanced,0.08525333801905315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,4,balanced,0.13291200002034506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,4,balanced,0.048154667019844055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,4,balanced,0.08481599887212117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.2,0.4807487964630127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,4,balanced,0.1339359978834788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,4,balanced,0.05806933343410492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,4,balanced,0.08617066343625386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.2,0.5898111820220947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,4,balanced,0.02367999901374181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,4,balanced,0.0806826651096344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,4,balanced,0.09021866321563721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,4,balanced,0.1376479963461558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,4,balanced,0.11687466502189636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,4,balanced,0.08974933624267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,8,power_law_1.2,0.7208447933197022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,4,balanced,0.10956799983978271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,4,balanced,0.02439466615517934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,4,balanced,0.0934399962425232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,4,balanced,0.14574933052062988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,4,balanced,0.10643733541170756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,8,power_law_1.2,1.3335040092468262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,4,balanced,0.02383466561635335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,4,balanced,0.10358400146166484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,4,balanced,0.10646399855613708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,4,balanced,0.030618667602539062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,4,balanced,0.10377599795659383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,4,balanced,0.10947733124097188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,4,balanced,0.04750399788220724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,4,balanced,0.1537813345591227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,8,power_law_1.2,2.3372159957885743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,4,balanced,0.11018133163452148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,4,balanced,0.05012266834576925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,4,balanced,0.10633066296577454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,4,balanced,0.1033066709836324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,4,balanced,0.16986133654912314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,4,balanced,0.056314667065938316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,4,balanced,0.11103999614715576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,4,balanced,0.1167680025100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,4,balanced,0.06197333335876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,4,balanced,0.1118239959081014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,4,balanced,0.06392000118891399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,4,balanced,0.1830986738204956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,4,balanced,0.12892799576123556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,4,balanced,0.1018933355808258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,4,balanced,0.07016000151634216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,4,balanced,0.12271466851234436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,4,balanced,0.05852266649405161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,4,balanced,0.11386133233706157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,4,balanced,0.061978667974472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,4,balanced,0.15133866667747498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,4,balanced,0.22740266720453897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,4,balanced,0.06595733265082042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,4,balanced,0.11410133043924968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,4,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,4,balanced,0.11859200398127238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,4,balanced,0.06670400003592174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,4,balanced,0.16422399878501892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,4,balanced,0.1253493328889211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,4,balanced,0.07050666709740956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,4,balanced,0.08508267005284627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,4,balanced,0.25034133593241376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,4,balanced,0.13919466733932495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,4,balanced,0.10041067004203796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,4,balanced,0.21314134200414023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,4,balanced,0.15651733676592508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,4,balanced,0.15131200353304544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,4,balanced,0.17974932988484701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,4,balanced,0.33921066919962567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,4,balanced,0.24834666649500528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,4,balanced,0.22339733441670737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,4,balanced,0.14843199650446573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,4,balanced,0.25100799401601154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,4,balanced,0.16234667102495828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,4,balanced,0.3512959877649943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,4,balanced,0.40084266662597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,4,balanced,0.3395306666692098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,4,balanced,0.17525333166122437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,4,balanced,0.43298133214314777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,4,balanced,0.2813546657562256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,4,balanced,0.4277706543604533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,4,balanced,0.5370186567306519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,4,balanced,0.3111199935277303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,4,balanced,0.6100533405939738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,4,balanced,0.6065546671549479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,4,balanced,0.6816853682200114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,4,balanced,0.44650665918986004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,4,balanced,0.7893919944763184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,4,balanced,0.7832159996032715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,4,balanced,0.5799200137456259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,4,balanced,0.8178026676177979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,4,power_law_1.01,0.054604798555374146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,4,power_law_1.01,0.03986560106277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,4,power_law_1.01,0.04888319969177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,4,power_law_1.01,0.06656000018119812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,4,power_law_1.01,0.0791487991809845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,4,balanced,0.9665706952412924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,4,power_law_1.01,0.10565119981765747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,4,balanced,0.952618678410848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,4,power_law_1.01,0.11246720552444459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,4,power_law_1.01,0.1089792013168335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,4,power_law_1.01,0.11389440298080444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,4,power_law_1.01,0.11281280517578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,4,power_law_1.01,0.11438080072402954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,4,power_law_1.01,0.12019840478897095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,4,power_law_1.01,0.12336000204086303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,4,power_law_1.01,0.12494720220565796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,4,balanced,0.7199040253957113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,4,power_law_1.01,0.1332800030708313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,4,power_law_1.01,0.03609600067138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,4,power_law_1.01,0.13971199989318847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,4,power_law_1.01,0.14229120016098024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,4,power_law_1.01,0.0355648010969162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,4,power_law_1.01,0.1518272042274475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,4,power_law_1.01,0.0377920001745224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,4,power_law_1.01,0.1593984007835388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,4,power_law_1.01,0.04645119905471802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,4,power_law_1.01,0.1941248059272766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,4,power_law_1.01,0.05690240263938904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,4,power_law_1.01,0.22055039405822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,4,power_law_1.01,0.07684479951858521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,4,power_law_1.01,0.26911358833312987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,4,power_law_1.01,0.07827200293540955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,4,balanced,1.2584853172302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,4,power_law_1.01,0.30812160968780516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,4,power_law_1.01,0.08067839741706848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,4,power_law_1.01,0.08105599880218506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,4,power_law_1.01,0.42656641006469725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,4,power_law_1.01,0.08254079818725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,4,power_law_1.01,0.5297088146209716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,4,power_law_1.01,0.08443520069122315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,4,power_law_1.01,0.6779520034790039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,4,power_law_1.01,0.08878719806671143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,4,power_law_1.01,0.8895104408264161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,4,power_law_1.01,0.09146239757537841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,4,power_law_1.01,0.09502080082893372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,4,power_law_1.01,1.1065024375915526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,4,power_law_1.01,0.056806397438049314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,4,power_law_1.01,0.10703999996185302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,4,balanced,1.4951945940653484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,4,power_law_1.01,1.586956787109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,4,power_law_1.01,0.10951039791107178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,4,balanced,1.47652800877889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,4,power_law_1.01,0.05082240104675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,4,power_law_1.01,0.10837759971618652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,4,power_law_1.01,0.0516543984413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,4,power_law_1.01,0.1271615982055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,4,power_law_1.01,3.2238590240478517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,4,power_law_1.01,0.06903039813041686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,4,power_law_1.01,0.1405311942100525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,4,power_law_1.01,0.07415680289268493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,4,power_law_1.01,0.09635840058326721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,4,power_law_1.01,0.1746559977531433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,4,power_law_1.01,0.10204160213470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,4,power_law_1.01,0.18815360069274903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,4,power_law_1.01,0.1007807970046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,4,balanced,1.1446186701456706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,4,power_law_1.01,0.24944639205932617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,4,power_law_1.01,0.10165760517120362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,4,power_law_1.01,0.3000511884689331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,4,power_law_1.01,0.10826239585876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,4,power_law_1.01,0.10455039739608765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,4,power_law_1.01,0.42070398330688474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,4,power_law_1.01,0.10787839889526367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,4,power_law_1.01,0.5363776206970214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,4,power_law_1.01,0.10812159776687622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,4,power_law_1.01,0.10901119709014892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,4,power_law_1.01,0.7616447925567627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,4,power_law_1.01,0.11515519618988038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,4,power_law_1.01,1.003212833404541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,4,power_law_1.01,0.11640959978103638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,4,power_law_1.01,0.1208191990852356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,4,power_law_1.01,1.147974395751953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,4,power_law_1.01,0.13708800077438354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.01,0.1536512017250061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,4,power_law_1.01,1.977881622314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.01,0.1893183946609497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.01,0.2103424072265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,4,power_law_1.01,3.784748840332031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.01,0.2679487943649292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.01,0.3137727975845337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.01,0.42792959213256837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.01,0.5139776229858398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.01,0.7622911930084229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.01,1.040396785736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,4,power_law_1.01,1.2723135948181152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,4,power_law_1.01,1.81646728515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,4,power_law_1.01,3.5793598175048826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,4,power_law_1.01,0.0225600004196167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,4,power_law_1.01,0.02083200067281723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,4,power_law_1.01,0.02513279914855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,4,power_law_1.01,0.027628800272941588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,4,power_law_1.01,0.030502399802207945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,4,power_law_1.01,0.04368639886379242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,4,balanced,2.4377973874409995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,4,power_law_1.01,0.04656639993190766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,4,power_law_1.01,0.04997119903564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,4,power_law_1.01,0.05201280117034912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,4,power_law_1.01,0.05691519975662231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,4,power_law_1.01,0.049593600630760196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,4,power_law_1.01,0.052671998739242554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,4,power_law_1.01,0.055820798873901366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,4,balanced,2.8731466929117837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,4,power_law_1.01,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,4,balanced,2.8965492248535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,4,power_law_1.01,0.05541120171546936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,4,power_law_1.01,0.059494400024414064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,4,power_law_1.01,0.0729856014251709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,4,power_law_1.01,0.0900160014629364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,4,balanced,2.256656010945638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.01,0.12979840040206908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.01,0.15719679594039918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.01,0.10818560123443603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.01,0.130239999294281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.01,0.1537600040435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.01,0.20733439922332764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.01,0.25662078857421877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.01,0.37233920097351075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.01,0.45192317962646483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,4,power_law_1.01,0.5683775901794433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,4,power_law_1.01,0.8879424095153808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,4,power_law_1.01,1.7194879531860352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,4,power_law_1.2,0.05454720258712768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,4,power_law_1.2,0.03982079923152924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,4,power_law_1.2,0.05018879771232605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,4,power_law_1.2,0.055961602926254274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,4,power_law_1.2,0.06901760101318359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,4,power_law_1.2,0.10002559423446655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,4,power_law_1.2,0.10958720445632934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,4,power_law_1.2,0.11352319717407226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,4,power_law_1.2,0.11815680265426635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,4,power_law_1.2,0.1211840033531189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,4,power_law_1.2,0.11960959434509277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,4,power_law_1.2,0.12167680263519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,4,power_law_1.2,0.12082560062408447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,4,power_law_1.2,0.13125760555267335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,4,power_law_1.2,0.1396607995033264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,4,power_law_1.2,0.14502400159835815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,4,power_law_1.2,0.14443520307540894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,4,power_law_1.2,0.16301440000534057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,4,power_law_1.2,0.17541760206222534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,4,power_law_1.2,0.03550719916820526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,4,power_law_1.2,0.2068608045578003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,4,power_law_1.2,0.22942719459533692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,4,power_law_1.2,0.034406399726867674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,4,power_law_1.2,0.3023423910140991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,4,power_law_1.2,0.038412800431251524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,4,power_law_1.2,0.34615681171417234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,4,power_law_1.2,0.04591360092163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,4,power_law_1.2,0.47331838607788085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,4,power_law_1.2,0.05318400263786316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,4,power_law_1.2,0.07502080202102661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,4,power_law_1.2,0.5688000202178956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,4,power_law_1.2,0.07972480058670044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,4,power_law_1.2,0.8175488471984863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,4,power_law_1.2,0.07928959727287292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,4,power_law_1.2,1.0540224075317384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,4,power_law_1.2,0.08110079765319825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,4,power_law_1.2,0.08432000279426574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,4,power_law_1.2,1.298476791381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,4,power_law_1.2,0.08611199855804444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,4,power_law_1.2,0.09131519794464112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,4,power_law_1.2,2.101728057861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,4,power_law_1.2,0.057017600536346434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,4,power_law_1.2,0.09327999949455261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,4,power_law_1.2,0.09928960204124451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,4,power_law_1.2,3.7884735107421874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,4,power_law_1.2,0.04609920084476471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,4,power_law_1.2,0.1080191969871521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,4,power_law_1.2,0.054553598165512085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,4,power_law_1.2,0.11063679456710815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,4,power_law_1.2,0.06556159853935242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,4,power_law_1.2,0.1135103940963745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,4,power_law_1.2,0.070278400182724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,4,power_law_1.2,0.12700799703598023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,4,power_law_1.2,0.09587200284004212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,4,power_law_1.2,0.14991999864578248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,4,power_law_1.2,0.10268160104751586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,4,power_law_1.2,0.1928256034851074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,4,power_law_1.2,0.10131839513778687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,4,power_law_1.2,0.21546239852905275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,4,power_law_1.2,0.10588799715042115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,2,balanced,0.03145066648721695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,4,power_law_1.2,0.27355520725250243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,4,power_law_1.2,0.1067199945449829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,4,power_law_1.2,0.022419199347496033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,4,power_law_1.2,0.3508352041244507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,4,power_law_1.2,0.10080000162124633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,2,balanced,0.03535466641187668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,4,power_law_1.2,0.4559296131134033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,4,power_law_1.2,0.11099519729614257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,4,power_law_1.2,0.02128639966249466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,4,power_law_1.2,0.5627903938293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,2,balanced,0.05301333467165629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,4,power_law_1.2,0.11167360544204712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,4,power_law_1.2,0.022592000663280487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,2,balanced,0.07978666822115581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,4,power_law_1.2,0.11463680267333984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,4,power_law_1.2,0.9192831993103028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,4,power_law_1.2,0.02508159875869751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,2,balanced,0.13010133306185404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,4,power_law_1.2,0.11576319932937622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,4,power_law_1.2,0.030457600951194763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,4,power_law_1.2,1.1140159606933593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,2,balanced,0.12542933225631714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,4,power_law_1.2,0.04345600008964538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,4,power_law_1.2,0.12226560115814208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,2,balanced,0.12570133805274963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,4,power_law_1.2,0.046438398957252505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,4,power_law_1.2,1.415993595123291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,4,power_law_1.2,0.12416640520095826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,2,balanced,0.12923199931780496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,4,power_law_1.2,0.04965119957923889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,4,power_law_1.2,0.1402176022529602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,2,balanced,0.12772799531618753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,4,power_law_1.2,0.05383679866790771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,4,power_law_1.2,2.565407943725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,2,balanced,0.13004266222318014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.2,0.1589311957359314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,4,power_law_1.2,0.05691519975662231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,2,balanced,0.04990399877230326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,2,balanced,0.03153600047032038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,4,power_law_1.2,0.051622402667999265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,2,balanced,0.12878400087356567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.2,0.17864320278167725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,4,power_law_1.2,0.053376001119613645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,2,balanced,0.05665599803129832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,2,balanced,0.13260266184806824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,2,balanced,0.034314667185147606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.2,0.21603200435638428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,4,power_law_1.2,4.662080001831055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,4,power_law_1.2,0.05619199872016907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,2,balanced,0.06957866748174031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,2,balanced,0.1346453328927358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.2,0.26803839206695557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,2,balanced,0.039493332306543984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,4,power_law_1.2,0.0546239972114563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,2,balanced,0.0923413336277008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,2,balanced,0.13564800222714743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,2,balanced,0.06039466460545858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,2,balanced,0.1365653375784556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,4,power_law_1.2,0.05648639798164368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.2,0.34954879283905027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,2,balanced,0.08687466382980347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,2,balanced,0.1423786679903666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,2,balanced,0.1330826679865519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,2,balanced,0.021151999632517498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,4,power_law_1.2,0.06053119897842407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.2,0.46090240478515626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,2,balanced,0.08866666754086812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,2,balanced,0.13095466295878092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,4,power_law_1.2,0.07502719759941101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,2,balanced,0.14432000120480856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,2,balanced,0.09115200241406758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,2,balanced,0.13362666964530945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.2,0.577184009552002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,2,balanced,0.021114667256673176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,4,power_law_1.2,0.09068800210952759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,2,balanced,0.09091732899347942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,2,balanced,0.1285653313000997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,2,balanced,0.15201066931088766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.2,0.1295680046081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,2,balanced,0.023445333043734234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,2,balanced,0.13572266697883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.2,0.8674880027770996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,2,balanced,0.09310932954152425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.2,0.1640128016471863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,2,balanced,0.1260426640510559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,2,balanced,0.09670933087666829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,2,balanced,0.16620266437530518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,2,balanced,0.031189332405726116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.2,0.11484800577163697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.2,1.169580841064453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,2,balanced,0.13019200166066489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,2,balanced,0.09762666622797649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.2,0.13697919845581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,2,balanced,0.13568533460299173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,2,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,2,balanced,0.10310399532318115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,2,balanced,0.17827200889587402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,4,power_law_1.2,1.2171648025512696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.2,0.16079360246658325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,2,balanced,0.1316373348236084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,2,balanced,0.04967466493447622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,2,balanced,0.10158933202425639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.2,0.2188800096511841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,2,balanced,0.14454933007558188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,2,balanced,0.10607999563217163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,2,balanced,0.05177066723505656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,2,balanced,0.20730666319529215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,4,power_law_1.2,2.616831970214844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.2,0.2692863941192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,2,balanced,0.13919466733932495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,2,balanced,0.1181066632270813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,2,balanced,0.05395199855168661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.2,0.3704832077026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,2,balanced,0.14482667048772177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,2,balanced,0.11959999799728394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,2,balanced,0.2325013279914856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,2,balanced,0.058789332707722984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,4,power_law_1.2,4.4821632385253904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.2,0.5006207942962646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,2,balanced,0.16204266746838888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,2,balanced,0.12284266948699951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,2,balanced,0.06127466758092245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,4,power_law_1.2,0.6091072082519531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,2,balanced,0.1774346629778544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,2,balanced,0.29682666063308716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,2,balanced,0.05602133274078369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,4,power_law_1.2,0.9548352241516114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,2,balanced,0.0625546673933665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,2,balanced,0.13684266805648804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,2,balanced,0.20265066623687744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,2,balanced,0.06469866633415222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,4,power_law_1.2,2.008390426635742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,2,balanced,0.060965334375699363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,2,balanced,0.15309866269429526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,2,balanced,0.23100799322128296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,2,balanced,0.33956801891326904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,2,balanced,0.06524266799290974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,2,balanced,0.06850666801134746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,2,balanced,0.19113600254058838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,2,balanced,0.34567467371622723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,2,balanced,0.07393066585063934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,2,balanced,0.08736532926559448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,2,balanced,0.20816532770792642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,2,balanced,0.10237333178520203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,2,balanced,0.4759519894917806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,2,balanced,0.40433065096537274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,2,balanced,0.29119465748469037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,2,balanced,0.14099733034769693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,2,balanced,0.15385599931081137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,2,balanced,0.5626986821492513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,2,balanced,0.344650665918986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,2,balanced,0.5853653351465861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,2,balanced,0.16406400005022684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,2,balanced,0.1755680044492086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,2,balanced,0.49380799134572345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,2,balanced,0.7160800298055013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,2,balanced,0.8119626839955648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,2,balanced,0.28445865710576373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,2,balanced,0.6255786816279093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,2,balanced,1.0251519680023193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,2,balanced,0.3033493359883626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,2,balanced,1.061743974685669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,2,balanced,0.9032533168792725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,2,balanced,0.43238401412963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,2,balanced,1.334880034128825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,2,balanced,1.1737600167592366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,2,balanced,1.300714651743571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,2,balanced,0.5664800008138021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,2,balanced,1.645477294921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,2,balanced,1.4381386439005535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,2,balanced,0.7048160235087076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,2,balanced,1.934442679087321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,2,power_law_1.01,0.033766400814056394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,2,power_law_1.01,0.03767040073871612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,2,power_law_1.01,0.0492031991481781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,2,power_law_1.01,0.06448000073432922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,2,balanced,2.5758719444274902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,2,power_law_1.01,0.07877119779586791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,2,power_law_1.01,0.11015039682388306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,2,power_law_1.01,0.11460479497909545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,2,power_law_1.01,0.11964800357818603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,2,power_law_1.01,0.1182528018951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,2,balanced,2.245946725209554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,2,power_law_1.01,0.12092159986495972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,2,power_law_1.01,0.12479360103607177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,2,power_law_1.01,0.12773760557174682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,2,power_law_1.01,0.13066879510879517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,2,power_law_1.01,0.13523199558258056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,2,power_law_1.01,0.1422592043876648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,2,power_law_1.01,0.15018240213394166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,2,balanced,1.1267999807993572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,2,power_law_1.01,0.15701760053634645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,2,power_law_1.01,0.1737215995788574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,2,power_law_1.01,0.19559680223464965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,2,power_law_1.01,0.23185279369354247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,2,power_law_1.01,0.24940159320831298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,2,power_law_1.01,0.3200704097747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,2,power_law_1.01,0.39416320323944093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,2,power_law_1.01,0.5396671772003174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,2,power_law_1.01,0.6545407772064209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,2,power_law_1.01,0.9206144332885742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,2,power_law_1.01,1.1806912422180176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,2,power_law_1.01,0.03261440098285675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,2,power_law_1.01,1.5083776473999024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,2,power_law_1.01,2.1710527420043944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,2,power_law_1.01,0.035148799419403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,2,power_law_1.01,0.03935360014438629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,2,power_law_1.01,0.050995200872421265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,2,power_law_1.01,4.206195068359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,2,power_law_1.01,0.05847039818763733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,2,power_law_1.01,0.08043519854545593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,2,power_law_1.01,0.08106880187988282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,2,power_law_1.01,0.08584319949150085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,2,power_law_1.01,0.0548799991607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,2,power_law_1.01,0.08975359797477722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,2,power_law_1.01,0.09165440201759338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,2,power_law_1.01,0.050969600677490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,2,power_law_1.01,0.0952127993106842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,2,power_law_1.01,0.06042240262031555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,2,power_law_1.01,0.10032000541687011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,2,power_law_1.01,0.08220800161361694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,2,balanced,3.738506635030111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,2,power_law_1.01,0.10064640045166015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,2,power_law_1.01,0.09123200178146362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,2,power_law_1.01,0.10648319721221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,2,power_law_1.01,0.11799039840698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,2,power_law_1.01,0.11356159448623657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,2,power_law_1.01,0.12126719951629639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,2,power_law_1.01,0.118777596950531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,2,power_law_1.01,0.1217087984085083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,2,power_law_1.01,0.12854399681091308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,2,power_law_1.01,0.12465920448303222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,2,power_law_1.01,0.14616960287094116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,2,power_law_1.01,0.12753280401229858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,2,power_law_1.01,0.17266559600830078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,2,power_law_1.01,0.1301632046699524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,2,power_law_1.01,0.019782400131225585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,2,power_law_1.01,0.13078399896621704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,2,power_law_1.01,0.01916159987449646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,2,power_law_1.01,0.210150408744812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,2,power_law_1.01,0.022361600399017335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,2,power_law_1.01,0.13201279640197755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,2,power_law_1.01,0.02766079902648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,2,power_law_1.01,0.13674240112304686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,2,power_law_1.01,0.23502719402313232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,2,balanced,5.038949330647786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,2,power_law_1.01,0.03203200101852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,2,power_law_1.01,0.13809280395507811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,2,power_law_1.01,0.31648640632629393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,2,power_law_1.01,0.0434112012386322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,2,power_law_1.01,0.1422976016998291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,2,power_law_1.01,0.3786623954772949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,2,power_law_1.01,0.04429439902305603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,2,power_law_1.01,0.15483520030975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,2,power_law_1.01,0.048563200235366824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,2,power_law_1.01,0.5299007892608643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,2,balanced,4.360869407653809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,2,power_law_1.01,0.1771008014678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,2,power_law_1.01,0.04981119930744171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,2,power_law_1.01,0.6715648174285889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,2,power_law_1.01,0.05392640233039856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.01,0.19775999784469606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,2,power_law_1.01,1.006809616088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.01,0.2336127996444702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.01,0.27939839363098146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,2,power_law_1.01,0.05120000243186951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,2,power_law_1.01,1.3263936042785645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,2,balanced,2.219242731730143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.01,0.36537599563598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.01,0.47399678230285647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,2,power_law_1.01,1.546835231781006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,2,power_law_1.01,0.05408639907836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.01,0.6333312034606934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.01,0.8572416305541992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,2,power_law_1.01,0.05761920213699341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,2,power_law_1.01,2.4703039169311523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.01,1.1840959548950196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,2,power_law_1.01,0.05551999807357788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,2,power_law_1.01,4.674054336547852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.01,1.5142784118652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,2,power_law_1.01,0.05898879766464234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,2,power_law_1.01,0.06165760159492493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,2,power_law_1.01,1.770649528503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,2,power_law_1.01,0.07296000123023987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,2,power_law_1.01,0.08349440097808838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,2,power_law_1.01,2.791321563720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.01,0.09616000056266785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.01,0.1207808017730713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,2,power_law_1.01,5.5224769592285154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.01,0.1264256000518799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.01,0.14515199661254882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.01,0.1700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.01,0.22391040325164796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.01,0.26878719329833983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.01,0.36821119785308837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.01,0.47801599502563474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,2,power_law_1.01,0.5633600234985352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,2,power_law_1.01,0.9036800384521484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,2,power_law_1.01,1.7068159103393554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,2,power_law_1.2,0.03167999982833862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,2,power_law_1.2,0.03390080034732819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,2,power_law_1.2,0.03222399950027466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,2,power_law_1.2,0.03733760118484497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,2,power_law_1.2,0.03632639944553375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,2,power_law_1.2,0.0466048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,2,power_law_1.2,0.04657280147075653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,2,power_law_1.2,0.054041600227355956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,2,power_law_1.2,0.05960320234298706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,2,power_law_1.2,0.08126720190048217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,2,power_law_1.2,0.07012479901313781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,2,power_law_1.2,0.08168960213661194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,2,power_law_1.2,0.11136000156402588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,2,power_law_1.2,0.08481919765472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,2,power_law_1.2,0.08783360123634339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,2,power_law_1.2,0.11485439538955688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,2,power_law_1.2,0.09134079813957215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,2,power_law_1.2,0.09483519792556763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,2,power_law_1.2,0.1205888032913208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,2,power_law_1.2,0.0992959976196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,2,power_law_1.2,0.1226304054260254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,2,power_law_1.2,0.1004032015800476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,2,power_law_1.2,0.1211967945098877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,2,power_law_1.2,0.10712319612503052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,2,power_law_1.2,0.1275712013244629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,2,power_law_1.2,0.11489280462265014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,2,power_law_1.2,0.12929279804229737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,2,power_law_1.2,0.12092800140380859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,2,power_law_1.2,0.13372160196304322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,2,power_law_1.2,0.12867200374603271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,2,power_law_1.2,0.14043519496917725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,2,power_law_1.2,0.14801280498504638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,2,power_law_1.2,0.14730240106582643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,2,power_law_1.2,0.1732800006866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,2,power_law_1.2,0.15149439573287965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,2,power_law_1.2,0.21326079368591308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,2,power_law_1.2,0.15997439622879028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,2,power_law_1.2,0.2446079969406128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,2,power_law_1.2,0.05418239831924439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,1,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,2,power_law_1.2,0.17496320009231567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,2,power_law_1.2,0.33308799266815187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,2,power_law_1.2,0.019968000054359437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,2,power_law_1.2,0.04940800070762634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,2,power_law_1.2,0.19853440523147584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,2,power_law_1.2,0.40303997993469237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,1,balanced,0.03976533313592275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,2,power_law_1.2,0.05639680027961731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,2,power_law_1.2,0.23644158840179444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,2,power_law_1.2,0.5859583854675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,2,power_law_1.2,0.020134399831295013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,2,power_law_1.2,0.07306879758834839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,1,balanced,0.056874667604764305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,2,power_law_1.2,0.02222079932689667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,2,power_law_1.2,0.2545727968215942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,2,power_law_1.2,0.7202367782592773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,2,power_law_1.2,0.09260159730911255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,1,balanced,0.08443733056386311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,2,power_law_1.2,0.02685439884662628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,2,power_law_1.2,0.36175360679626467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,2,power_law_1.2,0.1193727970123291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,1,balanced,0.14087999860445657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,2,power_law_1.2,1.0011008262634278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,2,power_law_1.2,0.030163198709487915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,1,balanced,0.13571199774742126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,2,power_law_1.2,0.11987839937210083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,2,power_law_1.2,0.4236735820770264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,1,balanced,0.03629866739114126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,2,power_law_1.2,0.04257279932498932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,2,power_law_1.2,1.3956480026245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,1,balanced,0.13800000150998434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,2,power_law_1.2,0.12563199996948243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,2,power_law_1.2,0.5751359939575196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,2,power_law_1.2,0.04482559859752655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,1,balanced,0.14094932874043783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,2,power_law_1.2,0.12626559734344484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,2,power_law_1.2,0.04921599924564361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,1,balanced,0.03977066775163015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,2,power_law_1.2,1.7251775741577149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,2,power_law_1.2,0.6865920066833496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,1,balanced,0.13851733009020487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,2,power_law_1.2,0.12899199724197388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,2,power_law_1.2,0.0499455988407135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,1,balanced,0.1449066698551178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,2,power_law_1.2,0.12942080497741698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,1,balanced,0.04558933277924856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,2,power_law_1.2,0.9764479637145996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,2,power_law_1.2,0.053913599252700804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,2,power_law_1.2,2.5653696060180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,1,balanced,0.14149866501490274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,2,power_law_1.2,0.13074560165405275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,1,balanced,0.07026133437951405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,2,power_law_1.2,0.05173119902610779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,2,power_law_1.2,1.3260095596313477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,1,balanced,0.14722133676211038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,2,power_law_1.2,0.1345471978187561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,1,balanced,0.09941866993904114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,2,power_law_1.2,0.05441280007362366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,1,balanced,0.15263467033704123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,2,power_law_1.2,0.13902080059051514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,1,balanced,0.10219732920328777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,2,power_law_1.2,0.05744640231132507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,2,power_law_1.2,5.442918395996093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,2,power_law_1.2,1.754707145690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,2,power_law_1.2,0.14275200366973878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,1,balanced,0.16035733620325723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,1,power_law_1.01,0.03272959887981415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,2,power_law_1.2,0.05532159805297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,1,balanced,0.10534399747848511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,2,power_law_1.2,0.1487488031387329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,1,balanced,0.16547200083732605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,2,power_law_1.2,2.469817543029785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,2,power_law_1.2,0.05898879766464234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,1,balanced,0.1053653359413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,1,power_law_1.01,0.03648639917373657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,2,power_law_1.2,0.154585599899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,2,power_law_1.2,0.061715197563171384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,2,power_law_1.2,0.1844928026199341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,1,balanced,0.17430933316548666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,1,balanced,0.10738666852315266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,1,power_law_1.01,0.04870400130748749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,2,power_law_1.2,0.07431679964065552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.2,0.19820799827575683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,2,power_law_1.2,4.960940933227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,1,balanced,0.10931733250617981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,1,balanced,0.18132799863815308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,2,power_law_1.2,0.08250240087509156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.2,0.23960959911346436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,1,power_law_1.01,0.06376960277557372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,1,balanced,0.1127893328666687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.2,0.09590399861335755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.2,0.28933119773864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,1,power_law_1.01,0.07689599990844727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,1,balanced,0.11725333333015442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,1,balanced,0.2102186679840088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.2,0.1260800004005432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.2,0.38204801082611084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,1,power_law_1.01,0.11630079746246338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,1,balanced,0.11983999609947205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.2,0.12854399681091308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.2,0.4453375816345215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,1,balanced,0.23099732398986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.2,0.14818559885025023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.2,0.643500804901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,1,balanced,0.12592533230781555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,1,power_law_1.01,0.12174719572067261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,1,power_law_1.01,0.03224320113658905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.2,0.16698880195617677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.2,0.8314623832702637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,1,balanced,0.13921067118644714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,1,power_law_1.01,0.12405120134353638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.2,0.22869760990142823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,1,balanced,0.2819360097249349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.2,1.1773440361022949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,1,balanced,0.1441973348458608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,1,power_law_1.01,0.13128960132598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,1,power_law_1.01,0.03511039912700653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.2,0.2768831968307495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.2,1.5599295616149902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,1,balanced,0.153546671072642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,1,balanced,0.31670933961868286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,1,power_law_1.01,0.04142720103263855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.2,0.40015358924865724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,1,power_law_1.01,0.1322175979614258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,2,power_law_1.2,2.019468879699707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,1,power_law_1.01,0.05467519760131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,1,balanced,0.17993066708246866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.2,0.46316161155700686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,2,power_law_1.2,3.181510353088379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,1,power_law_1.01,0.06762239933013917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,1,power_law_1.01,0.13886719942092896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,1,power_law_1.01,0.08899840116500854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,2,power_law_1.2,0.5761472225189209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,1,balanced,0.44445868333180744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,1,balanced,0.21628799041112265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,2,power_law_1.2,5.664479827880859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,1,power_law_1.01,0.09418240189552307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,2,power_law_1.2,0.9195136070251465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,1,power_law_1.01,0.1449087977409363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,1,power_law_1.01,0.0979200005531311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,1,balanced,0.2733706633249919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,1,power_law_1.01,0.10104320049285889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,2,power_law_1.2,1.8514944076538087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,1,power_law_1.01,0.15045759677886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,1,power_law_1.01,0.10542720556259155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,1,balanced,0.5169599850972494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,1,balanced,0.3118666609128316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,1,power_law_1.01,0.11093120574951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,1,power_law_1.01,0.15632640123367308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,1,power_law_1.01,0.11802239418029785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,1,power_law_1.01,0.16686079502105713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,1,power_law_1.01,0.12419840097427368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,1,power_law_1.01,0.12426879405975341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,1,balanced,0.46327467759450275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,1,power_law_1.01,0.17178879976272582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,1,power_law_1.01,0.1331712007522583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,1,power_law_1.01,0.14181120395660402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,1,balanced,0.7453493277231852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,1,power_law_1.01,0.18670719861984253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,1,power_law_1.01,0.1555840015411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,1,power_law_1.01,0.18778879642486573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,1,power_law_1.01,0.21745920181274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,1,balanced,0.5586453278859457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,1,power_law_1.01,0.22287359237670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,1,power_law_1.01,0.23992960453033446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,1,power_law_1.01,0.2855679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,1,power_law_1.01,0.2958400011062622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,1,balanced,0.9157493114471436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,1,power_law_1.01,0.3404927968978882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,1,power_law_1.01,0.3439039945602417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,1,balanced,0.8188640276590983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,1,power_law_1.01,0.4649343967437744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,1,power_law_1.01,0.4481152057647705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,1,power_law_1.01,0.5887551784515381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,1,power_law_1.01,0.5653439998626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,1,power_law_1.01,0.8254528045654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,1,power_law_1.01,0.7550015926361084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,1,power_law_1.01,1.0613439559936524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,1,power_law_1.01,0.9493887901306153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,1,power_law_1.01,1.347270393371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,1,power_law_1.01,1.5321408271789552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,1,balanced,1.0273173650105794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,1,power_law_1.01,1.792473602294922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,1,power_law_1.01,1.9857152938842773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,1,balanced,1.3438560167948406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,1,power_law_1.01,2.1947328567504885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,1,power_law_1.01,2.442464065551758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,1,power_law_1.01,3.2712638854980467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,1,power_law_1.01,3.803692626953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,1,power_law_1.01,6.361843109130859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,1,power_law_1.01,7.51041259765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,1,balanced,1.5149599711100261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,1,balanced,1.7251520156860352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,1,balanced,1.9949493408203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,1,balanced,2.130298614501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,1,balanced,2.4810773531595864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,1,balanced,3.316650708516439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,1,balanced,3.8505172729492188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,1,power_law_1.2,0.03158400058746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,1,power_law_1.2,0.03619840145111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,1,power_law_1.2,0.04347519874572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,1,power_law_1.2,0.061791998147964475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,1,power_law_1.2,0.07230719923973083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,1,power_law_1.2,0.1128767967224121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,1,power_law_1.2,0.12043520212173461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,1,power_law_1.2,0.1287551999092102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,1,power_law_1.2,0.13404799699783326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,1,power_law_1.2,0.1336959958076477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,1,power_law_1.2,0.13971840143203734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,1,power_law_1.2,0.14663039445877074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,1,power_law_1.2,0.1521664023399353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,1,power_law_1.2,0.03268480002880096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,1,power_law_1.2,0.1595136046409607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,1,power_law_1.2,0.1704192042350769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,1,power_law_1.2,0.17420799732208253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,1,power_law_1.2,0.03676159977912903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,1,power_law_1.2,0.19123200178146363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,1,power_law_1.2,0.04213759899139404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,1,power_law_1.2,0.2212928056716919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,1,power_law_1.2,0.05175039768218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,1,power_law_1.2,0.241158390045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,1,power_law_1.2,0.06403200030326843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,1,power_law_1.2,0.2975680112838745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,1,power_law_1.2,0.09099519848823548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,1,power_law_1.2,0.3458175897598267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,1,power_law_1.2,0.09744639992713929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,1,power_law_1.2,0.4509439945220947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,1,power_law_1.2,0.09916800260543823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,1,power_law_1.2,0.5685376167297364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,1,power_law_1.2,0.1032639980316162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,1,power_law_1.2,0.7573440074920654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,1,power_law_1.2,0.10647679567337036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,1,power_law_1.2,0.11291520595550537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,1,power_law_1.2,0.9500096321105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,1,balanced,6.365887959798177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,1,power_law_1.2,0.11882879734039306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,1,power_law_1.2,1.3427264213562011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,1,power_law_1.2,0.12547839879989625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,1,power_law_1.2,1.7981887817382813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,1,power_law_1.2,0.1254464030265808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,1,power_law_1.2,0.13419519662857055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,8,balanced,0.04814933240413666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,1,power_law_1.2,2.212224006652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,1,power_law_1.2,0.1430400013923645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,8,balanced,0.0529013325770696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,1,power_law_1.2,0.1573248028755188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,8,balanced,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,1,power_law_1.2,3.2559486389160157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,1,power_law_1.2,0.1896512031555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,1,balanced,7.514261245727539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,8,balanced,0.06930133203665416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,8,balanced,0.07228266696135204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,1,power_law_1.2,0.2263808012008667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,8,balanced,0.10292800267537434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,8,balanced,0.0981066624323527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,8,balanced,0.15833066900571188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,1,power_law_1.2,0.2883968114852905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,1,power_law_1.2,6.343987274169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,8,balanced,0.15171200037002563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,8,balanced,0.1849493384361267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,1,power_law_1.2,0.34328958988189695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,8,balanced,0.189626673857371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,8,balanced,0.26030399401982623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,1,power_law_1.2,0.4648767948150635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,8,balanced,0.18631466229756674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,8,balanced,0.31407999992370605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,1,power_law_1.2,0.5873727798461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,8,balanced,0.19035200277964273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,8,balanced,0.3174080053965251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,1,power_law_1.2,0.8269503593444825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,8,balanced,0.18757333358128866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,1,power_law_1.2,1.0678784370422363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,8,balanced,0.319050669670105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,8,balanced,0.19004799922307333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,8,balanced,0.3190186619758606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,1,power_law_1.2,1.5264320373535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,8,balanced,0.19377599159876505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,8,balanced,0.3147573272387187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,1,power_law_1.2,1.9916736602783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,8,balanced,0.19126933813095093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,8,balanced,0.31647467613220215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,1,power_law_1.2,2.451353645324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,8,balanced,0.31946667035420734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,8,balanced,0.19663467009862265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,8,balanced,0.32204266389211017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,1,power_law_1.2,3.7995136260986326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,8,balanced,0.20217599471410116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,8,balanced,0.32185065746307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,8,balanced,0.3282613356908162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,8,balanced,0.20337067047754923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,1,power_law_1.2,7.500691223144531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,8,balanced,0.32705066601435345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,8,balanced,0.2084853251775106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,8,balanced,0.24215465784072876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,8,balanced,0.3352479934692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,8,balanced,0.35684800148010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,8,balanced,0.2306613326072693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,8,balanced,0.3521493275960286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,8,balanced,0.28091200192769367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,8,balanced,0.05109333495299021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,8,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,8,balanced,0.4421226580937703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,8,balanced,0.27354133129119873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,8,balanced,0.04539733131726583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,8,balanced,0.05188799897829691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,8,balanced,0.053264002005259194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,8,balanced,0.06228800117969513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,8,balanced,0.06589866677920024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,8,balanced,0.0865119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,8,balanced,0.38939201831817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,8,balanced,0.08745599786440532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,8,balanced,0.12170666456222534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,8,balanced,0.4066559871037801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,8,balanced,0.10363733768463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,8,balanced,0.13714667161305746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,8,balanced,0.13857600092887878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,8,balanced,0.12249599893887837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,8,balanced,0.6638133525848389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,8,balanced,0.14018666744232178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,8,balanced,0.38124267260233563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,8,balanced,0.13983999689420065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,8,balanced,0.11692266662915547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,8,balanced,0.13992533087730408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,8,balanced,0.1225493351618449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,8,balanced,0.4890933434168498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,8,balanced,0.14317867159843445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,8,balanced,0.5506879885991415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,8,balanced,0.14406399925549826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,8,balanced,0.17831466595331827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,8,balanced,0.1443839967250824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,8,balanced,0.16305599610010782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,8,balanced,0.14813866217931113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,8,balanced,0.17454399665196738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,8,balanced,0.22396800915400186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,8,balanced,0.15190399686495462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,8,balanced,0.7392693360646566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,8,balanced,0.23068799575169882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,8,balanced,0.15292267004648843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,8,balanced,0.5956053336461385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,8,balanced,0.2580373287200928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,8,balanced,0.15778666734695435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,8,balanced,0.2342133323351542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,8,balanced,0.16885866721471152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,8,balanced,0.22450133164723715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,8,balanced,0.17616534233093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,8,balanced,0.24089600642522177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,8,balanced,0.7259679635365804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,8,balanced,0.2638026674588521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,8,balanced,0.2053013245264689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,8,balanced,0.3095039923985799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,8,balanced,0.9225546518961588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,8,balanced,0.21369065841039023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,8,balanced,0.3433493375778198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,8,balanced,0.2733813325564067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,8,balanced,0.41973865032196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,8,balanced,1.2802293300628662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,8,balanced,0.3053973317146301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,8,balanced,0.47539734840393066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,8,balanced,1.0763786633809407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,8,balanced,0.43909867604573566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,8,balanced,0.7330133120218912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,8,balanced,1.3249279657999675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,8,balanced,0.5050346851348877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,8,balanced,0.9000159899393717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,8,balanced,1.2194240093231201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,8,balanced,0.7066986560821533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,8,balanced,1.3269120057423909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,8,balanced,1.4083946545918782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,8,balanced,0.8803520202636719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,8,balanced,1.7268640200297039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,8,balanced,1.953536033630371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,8,power_law_1.01,0.1569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,8,power_law_1.01,0.19934719800949097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,8,power_law_1.01,0.10646400451660157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,8,power_law_1.01,0.19187840223312377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,8,balanced,2.35483725865682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,8,power_law_1.01,0.18240640163421631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,8,power_law_1.01,0.2613696098327637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,8,balanced,1.0431466897328694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,8,power_law_1.01,0.27984640598297117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,8,power_law_1.01,0.28080000877380373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,8,balanced,2.0117012659708657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,8,power_law_1.01,0.2934528112411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,8,power_law_1.01,0.29691519737243655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,8,power_law_1.01,0.2987584114074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,8,power_law_1.01,0.28575360774993896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,8,power_law_1.01,0.3123584032058716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,8,power_law_1.01,0.31886720657348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,8,power_law_1.01,0.3289088010787964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,8,power_law_1.01,0.35311360359191896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,8,power_law_1.01,0.36833279132843016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,8,power_law_1.01,0.0880511999130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,8,power_law_1.01,0.41509761810302737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,8,power_law_1.01,0.09440640211105347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,8,power_law_1.01,0.43172478675842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,8,power_law_1.01,0.09681919813156128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,8,power_law_1.01,0.5284287929534912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,8,power_law_1.01,0.06370559930801392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,8,power_law_1.01,0.11327999830245972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,8,power_law_1.01,0.5841856002807617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,8,power_law_1.01,0.08743039965629577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,8,power_law_1.01,0.06873599886894226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,8,power_law_1.01,0.09282559752464295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,8,power_law_1.01,0.731987190246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,8,power_law_1.01,0.08655999898910523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,8,power_law_1.01,0.1217919945716858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,8,power_law_1.01,0.11560959815979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,8,power_law_1.01,0.7696063995361329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,8,power_law_1.01,0.1274880051612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,8,power_law_1.01,0.15471999645233153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,8,balanced,1.5991627375284831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,8,power_law_1.01,0.125382399559021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,8,power_law_1.01,0.9209792137145996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,8,power_law_1.01,0.16525440216064452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,8,power_law_1.01,0.1339903950691223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,8,balanced,3.317296028137207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,8,power_law_1.01,1.27456636428833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,8,power_law_1.01,0.16678400039672853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,8,power_law_1.01,0.13191039562225343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,8,power_law_1.01,0.16885759830474853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,8,power_law_1.01,1.612384033203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,8,power_law_1.01,0.17072000503540039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,8,power_law_1.01,0.13501440286636351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,8,power_law_1.01,0.17813119888305665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,8,power_law_1.01,0.13659520149230958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,8,power_law_1.01,2.099411201477051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,8,power_law_1.01,0.18444160223007203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,8,power_law_1.01,0.053958398103713986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,8,power_law_1.01,0.14025599956512452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,8,power_law_1.01,0.18636800050735475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,8,power_law_1.01,2.3809535980224608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,8,power_law_1.01,0.14232959747314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,8,power_law_1.01,0.19445760250091554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,8,power_law_1.01,0.05589119791984558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,8,power_law_1.01,0.1487936019897461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,8,power_law_1.01,0.20051200389862062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,8,power_law_1.01,3.3022529602050783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,8,power_law_1.01,0.04712960124015808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,8,power_law_1.01,0.15300480127334595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,8,power_law_1.01,0.21565439701080322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,8,power_law_1.01,0.055743998289108275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,8,power_law_1.01,0.23191680908203124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,8,power_law_1.01,0.16536320447921754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,8,power_law_1.01,0.0648256003856659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,8,power_law_1.01,0.2663935899734497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,8,power_law_1.01,0.18394240140914916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,8,power_law_1.01,7.356768035888672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,8,power_law_1.01,0.08646399974822998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,8,power_law_1.01,0.26382079124450686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.01,0.20278398990631102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,8,power_law_1.01,0.33755519390106203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,8,power_law_1.01,0.10599039793014527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,8,balanced,3.7045974731445312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.01,0.2502079963684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,8,power_law_1.01,0.3933824062347412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,8,power_law_1.01,0.0958079993724823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.01,0.25322880744934084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,8,power_law_1.01,0.4688896179199219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,8,power_law_1.01,0.10121599435806275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.01,0.3452159881591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,8,power_law_1.01,0.501523208618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,8,power_law_1.01,0.1364351987838745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.01,0.4621056079864502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,8,power_law_1.01,0.6485568046569824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,8,power_law_1.01,0.14123519659042358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.01,0.5529215812683106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,8,power_law_1.01,0.8600447654724122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,8,power_law_1.01,0.12743040323257446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,8,balanced,4.4759521484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.01,0.6896383762359619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,8,power_law_1.01,1.1987263679504394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,8,power_law_1.01,0.19266560077667236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.01,1.0154175758361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,8,power_law_1.01,1.555820846557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,8,power_law_1.01,0.19900799989700318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.01,1.2119168281555175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,8,power_law_1.01,0.20355839729309083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,8,power_law_1.01,1.842336082458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,8,power_law_1.01,0.20472319126129152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,8,power_law_1.01,1.6449855804443358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,8,power_law_1.01,0.20321919918060302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,8,power_law_1.01,2.523852729797363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,8,power_law_1.01,2.6436479568481444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,8,power_law_1.01,0.18748159408569337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.01,0.1867136001586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,8,power_law_1.01,4.610329437255859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,8,power_law_1.01,5.161798477172852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.01,0.2065279960632324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.01,0.2313920021057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.01,0.2883008003234863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.01,0.340883207321167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.01,0.4795328140258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.01,0.5614272117614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.01,0.8496959686279297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.01,1.1248064041137695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,8,power_law_1.01,1.314406394958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,8,power_law_1.01,2.059942436218262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,8,balanced,3.0557918548583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,8,power_law_1.01,4.167091369628906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,8,balanced,6.372351964314778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,8,power_law_1.2,0.1568127989768982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,8,power_law_1.2,0.09785599708557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,8,power_law_1.2,0.08531200289726257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,8,power_law_1.2,0.0920960009098053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,8,power_law_1.2,0.055936002731323244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,8,power_law_1.2,0.10650880336761474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,8,power_law_1.2,0.06102399826049805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,8,power_law_1.2,0.1381183981895447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,8,power_law_1.2,0.0756928026676178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,8,power_law_1.2,0.17827839851379396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,8,power_law_1.2,0.09237759709358215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,8,power_law_1.2,0.0663424015045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,8,power_law_1.2,0.2694463968276978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,8,power_law_1.2,0.11281280517578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,8,power_law_1.2,0.08328319787979126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,8,power_law_1.2,0.16275839805603026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,8,power_law_1.2,0.2691839933395386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,8,power_law_1.2,0.09205120205879211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,8,power_law_1.2,0.05390080213546753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,8,power_law_1.2,0.16955519914627076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,8,power_law_1.2,0.29006080627441405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,8,power_law_1.2,0.12008960247039795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,4,balanced,0.05004266897837321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,8,power_law_1.2,0.1679487943649292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,8,power_law_1.2,0.12598400115966796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,8,power_law_1.2,0.30691840648651125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,8,power_law_1.2,0.044460800290107724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,4,balanced,0.07067733506361644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,8,power_law_1.2,0.17244160175323486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,8,power_law_1.2,0.12579200267791749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,8,power_law_1.2,0.3102272033691406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,4,balanced,0.09526399771372478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,8,power_law_1.2,0.04478079974651337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,4,balanced,0.15756799777348837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,8,power_law_1.2,0.05011839866638183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,8,power_law_1.2,0.313100790977478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,8,power_law_1.2,0.17775360345840455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,8,power_law_1.2,0.13241599798202514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,4,balanced,0.2591093381245931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,8,power_law_1.2,0.06165760159492493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,4,balanced,0.3091306686401367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,8,power_law_1.2,0.3131200075149536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,8,power_law_1.2,0.18322559595108032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,8,power_law_1.2,0.13119360208511352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,8,power_law_1.2,0.08881919980049133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,4,balanced,0.31069332361221313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,8,power_law_1.2,0.18996479511260986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,8,power_law_1.2,0.3177023887634277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,8,power_law_1.2,0.10060160160064698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,4,balanced,0.3146559993426005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,8,power_law_1.2,0.13774720430374146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,8,power_law_1.2,0.09695360064506531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,4,balanced,0.3152160048484802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,8,power_law_1.2,0.19054720401763917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,8,power_law_1.2,0.33651199340820315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,8,power_law_1.2,0.14030079841613768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,8,power_law_1.2,0.10188800096511841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,4,balanced,0.31809600194295246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,4,balanced,0.04598933458328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,8,power_law_1.2,0.20305280685424804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,8,power_law_1.2,0.35726079940795896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,4,balanced,0.3171893358230591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,8,power_law_1.2,0.1374400019645691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,8,power_law_1.2,0.1419648051261902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,4,balanced,0.05007466673851013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,4,balanced,0.319050669670105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,8,power_law_1.2,0.13999999761581422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,8,power_law_1.2,0.21882240772247313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,8,power_law_1.2,0.3613248109817505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,4,balanced,0.07228266696135204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,8,power_law_1.2,0.14413440227508545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,4,balanced,0.3197173277537028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,8,power_law_1.2,0.12094080448150635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,4,balanced,0.10333866874376933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,8,power_law_1.2,0.22540159225463868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,8,power_law_1.2,0.3687488079071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,4,balanced,0.1585973302523295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,4,balanced,0.3239786624908447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,8,power_law_1.2,0.15190399885177613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,8,power_law_1.2,0.19759999513626098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,4,balanced,0.1821920077006022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,4,balanced,0.18394132455190024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,8,power_law_1.2,0.24725120067596434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,4,balanced,0.3327839970588684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,8,power_law_1.2,0.44023680686950684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,8,power_law_1.2,0.15351680517196656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,8,power_law_1.2,0.20288639068603515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,4,balanced,0.18597867091496786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,4,balanced,0.06657066444555919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,4,balanced,0.027237333357334137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,8,power_law_1.2,0.3055999994277954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,4,balanced,0.1886826753616333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,8,power_law_1.2,0.21438078880310057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,4,balanced,0.3323520024617513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,8,power_law_1.2,0.1647744059562683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,8,power_law_1.2,0.45941758155822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,4,balanced,0.1880106727282206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,4,balanced,0.07653866708278656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,4,balanced,0.035349334279696144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,8,power_law_1.2,0.20266880989074706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,8,power_law_1.2,0.3056447982788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,8,power_law_1.2,0.18873599767684937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,4,balanced,0.19170665740966797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,8,power_law_1.2,0.5837759971618652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,4,balanced,0.3435200055440267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,4,balanced,0.09698667128880818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,4,balanced,0.04621866842110952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,8,power_law_1.2,0.20298240184783936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,4,balanced,0.19385600090026855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.2,0.20571520328521728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,8,power_law_1.2,0.34571518898010256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,4,balanced,0.14174933234850565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,4,balanced,0.06137600044409434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,8,power_law_1.2,0.6865791797637939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,4,balanced,0.19527999560038248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,8,power_law_1.2,0.18855040073394774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.2,0.26056320667266847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,4,balanced,0.2214133342107137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,4,balanced,0.35283199946085614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,8,power_law_1.2,0.41130881309509276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,4,balanced,0.08732799688975017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,4,balanced,0.19901333252588907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,8,power_law_1.2,0.8416383743286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.2,0.20707199573516846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.2,0.2830399990081787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,4,balanced,0.24974934260050455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,8,power_law_1.2,0.4955776214599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,4,balanced,0.36533331871032715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,4,balanced,0.23348265886306763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,4,balanced,0.20588266849517822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,8,power_law_1.2,0.5537407875061036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,4,balanced,0.10475732882817586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.2,0.23013761043548583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,4,balanced,0.23055466016133627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.2,0.3537280082702637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,8,power_law_1.2,0.8484736442565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,4,balanced,0.45286933581034344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,4,balanced,0.20914133389790854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,8,power_law_1.2,0.7885503768920898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,4,balanced,0.11282133062680562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,4,balanced,0.2244266668955485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.2,0.23494400978088378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.2,0.4735424041748047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,4,balanced,0.11075733105341594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,8,power_law_1.2,1.0684351921081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,4,balanced,0.22028799851735434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,8,power_law_1.2,0.9901375770568848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,4,balanced,0.21673067410786948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,4,balanced,0.1149066686630249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.2,0.30732800960540774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.2,0.6031231880187988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,4,balanced,0.40814932187398273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,8,power_law_1.2,1.3186047554016114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,4,balanced,0.22785067558288574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,8,power_law_1.2,1.5249088287353516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,4,balanced,0.23066665728886923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,4,balanced,0.11994666854540507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,4,balanced,0.22478399674097696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,8,power_law_1.2,1.9150400161743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.2,0.385913610458374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.2,0.857472038269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,4,balanced,0.13174933195114136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,8,power_law_1.2,1.7907264709472657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,4,balanced,0.6627680063247681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,4,balanced,0.24296534061431885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,4,balanced,0.2265440026919047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,8,power_law_1.2,2.685536003112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.2,0.5055615901947021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,4,balanced,0.14266666769981384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.2,1.0762623786926269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,8,power_law_1.2,2.1820735931396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,4,balanced,0.23783999681472778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,8,power_law_1.2,3.3145023345947267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,4,balanced,0.30129067103068036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,4,balanced,0.23814932505289713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,4,balanced,0.5278826554616293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,4,balanced,0.15425067146619162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.2,0.6872511863708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.2,1.3514112472534179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,8,power_law_1.2,4.252729415893555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,4,balanced,0.23094934225082397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,8,power_law_1.2,3.5270206451416017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,4,balanced,0.13944000005722046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,4,balanced,0.3064799904823303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.2,0.9666879653930665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,4,balanced,0.24228266874949136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,8,power_law_1.2,1.8698175430297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,4,balanced,0.15033599734306335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,4,balanced,0.16236799955368042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,4,balanced,0.2490239938100179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.2,1.213907241821289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,8,power_law_1.2,6.5939582824707035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,8,power_law_1.2,9.152909088134766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,8,power_law_1.2,2.809459114074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,4,balanced,0.17758933703104654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,8,power_law_1.2,1.4844415664672852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,4,balanced,0.24302399158477783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,4,balanced,0.43438398838043213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,4,balanced,0.7615306377410889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,4,balanced,0.17983466386795044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,8,power_law_1.2,5.907123184204101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,4,balanced,0.26103999217351276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,8,power_law_1.2,2.601587104797363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,4,balanced,0.25312532981236774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,4,balanced,0.4440000057220459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,4,balanced,0.2797546585400899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,4,balanced,0.2678133249282837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,8,power_law_1.2,5.086316680908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,4,balanced,0.29212266206741333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,4,balanced,0.4479946692784627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,4,balanced,0.7875520388285319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,4,balanced,0.6229333480199178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,4,balanced,0.3360746701558431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,4,balanced,0.47118933995564777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,4,balanced,0.36233067512512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,4,balanced,0.7424320379892985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,4,balanced,0.6565599838892618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,4,balanced,1.272111972173055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,4,balanced,0.5555786689122518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,4,balanced,0.8646293481191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,4,balanced,0.6470400094985962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,4,balanced,1.0530186494191487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,4,balanced,1.4239892959594727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,4,balanced,1.0659200350443523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,4,balanced,0.9502773284912109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,4,balanced,1.3486240704854329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,4,balanced,1.4684693018595378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,4,balanced,1.5502773920694988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,4,balanced,1.2413067022959392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,4,balanced,1.5762346585591633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,4,balanced,1.6449653307596843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,4,balanced,1.3921546936035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,4,balanced,2.555487950642904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,4,balanced,2.48745059967041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,4,balanced,2.590640068054199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,4,balanced,2.3087894121805825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,4,balanced,4.940224011739095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,4,balanced,4.760239919026692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,4,balanced,5.132207870483398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,4,power_law_1.01,0.10104320049285889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,4,power_law_1.01,0.13752319812774658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,4,balanced,4.393146514892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,4,power_law_1.01,0.08499199748039246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,4,power_law_1.01,0.13237119913101197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,4,power_law_1.01,0.1787392020225525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,4,power_law_1.01,0.251859188079834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,4,power_law_1.01,0.06910079717636108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,4,power_law_1.01,0.2731839895248413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,4,power_law_1.01,0.09822720289230347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,4,power_law_1.01,0.033983999490737916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,4,power_law_1.01,0.28024959564208984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,4,power_law_1.01,0.07913600206375122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,4,power_law_1.01,0.27818880081176756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,4,power_law_1.01,0.2894464015960693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,4,power_law_1.01,0.06119040250778198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,4,power_law_1.01,0.11729919910430908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,4,power_law_1.01,0.04126720130443573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,4,power_law_1.01,0.09136639833450318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,4,power_law_1.01,0.08799359798431397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,4,power_law_1.01,0.2909888029098511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,4,power_law_1.01,0.04145280122756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,4,power_law_1.01,0.05478399991989136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,4,power_law_1.01,0.2926784038543701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,4,power_law_1.01,0.11199359893798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,4,power_law_1.01,0.06033279895782471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,4,power_law_1.01,0.15359359979629517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,4,power_law_1.01,0.12462719678878784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,4,power_law_1.01,0.30899200439453123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,4,power_law_1.01,0.3176127910614014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,4,power_law_1.01,0.15806080102920533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,4,power_law_1.01,0.08419839739799499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,4,power_law_1.01,0.32677760124206545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,4,power_law_1.01,0.16136959791183472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,4,power_law_1.01,0.13902080059051514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,4,power_law_1.01,0.0868607997894287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,4,power_law_1.01,0.34225919246673586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,4,power_law_1.01,0.16814719438552855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,4,power_law_1.01,0.0906175971031189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,4,power_law_1.01,0.3432960033416748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,4,power_law_1.01,0.17454080581665038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,4,power_law_1.01,0.2037440061569214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,4,power_law_1.01,0.09524480104446412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,4,power_law_1.01,0.3754431962966919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,4,power_law_1.01,0.1778175950050354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,4,power_law_1.01,0.3949376106262207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,4,power_law_1.01,0.09809920191764832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,4,power_law_1.01,0.20993280410766602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,4,power_law_1.01,0.17984000444412232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,4,power_law_1.01,0.10046080350875855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,4,power_law_1.01,0.47501440048217775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,4,power_law_1.01,0.21597440242767335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,4,power_law_1.01,0.18272639513015748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,4,power_law_1.01,0.11064319610595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,4,power_law_1.01,0.5424767971038819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,4,power_law_1.01,0.18881920576095582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,4,power_law_1.01,0.11673599481582642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,4,power_law_1.01,0.21835520267486572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,4,power_law_1.01,0.6709504127502441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,4,power_law_1.01,0.204038405418396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,4,power_law_1.01,0.11155200004577637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,4,power_law_1.01,0.21409280300140382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,4,power_law_1.01,0.7091328144073487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,4,power_law_1.01,0.11898239850997924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,4,power_law_1.01,0.2191999912261963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,4,power_law_1.01,0.22428159713745116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,4,power_law_1.01,0.8648127555847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,4,power_law_1.01,0.1284160017967224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,4,power_law_1.01,0.22967040538787842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,4,power_law_1.01,1.0902527809143066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,4,power_law_1.01,0.2147200107574463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,4,power_law_1.01,0.12193280458450317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,4,power_law_1.01,0.2214143991470337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,4,power_law_1.01,1.559712028503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,4,power_law_1.01,0.22677760124206542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,4,power_law_1.01,0.1401919960975647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,4,power_law_1.01,0.2376960039138794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,4,power_law_1.01,1.8420543670654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,4,power_law_1.01,0.2388672113418579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,4,power_law_1.01,0.2784960031509399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.01,0.1856384038925171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,4,power_law_1.01,2.2570880889892577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,4,power_law_1.01,0.23452799320220946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.01,0.18921600580215453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,4,power_law_1.01,0.2500799894332886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,4,power_law_1.01,0.3231040000915527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.01,0.21413118839263917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,4,power_law_1.01,3.4384254455566405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,4,power_law_1.01,0.272377610206604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.01,0.26017279624938966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,4,power_law_1.01,0.37499520778656004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.01,0.29752960205078127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.01,0.29117441177368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.01,0.34092159271240235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,4,power_law_1.01,6.208768081665039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,4,power_law_1.01,0.49889278411865234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.01,0.37392640113830566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.01,0.3954432010650635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.01,0.48215041160583494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,4,power_law_1.01,0.5353151798248291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.01,0.4778751850128174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.01,0.6615615844726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.01,0.6384128093719482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,4,power_law_1.01,0.698636817932129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.01,0.8782015800476074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,4,power_law_1.01,0.861241626739502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.01,0.7419583797454834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,4,power_law_1.01,1.0810815811157226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,4,power_law_1.01,1.2723072052001954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.01,0.9714431762695312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.01,1.3968064308166503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,4,power_law_1.01,1.740991973876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,4,power_law_1.01,1.7499135971069335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.01,1.7366399765014648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,4,power_law_1.01,2.024838447570801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,4,power_law_1.01,2.0967679977416993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,4,power_law_1.01,3.193350410461426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,4,power_law_1.01,3.203327941894531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,4,power_law_1.01,3.254067230224609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,4,power_law_1.01,5.953004837036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,4,power_law_1.01,6.698342132568359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,4,power_law_1.2,0.09923200011253357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,4,power_law_1.2,0.07351679801940918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,4,power_law_1.2,0.08368639945983887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,4,power_law_1.2,0.12125439643859863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,4,power_law_1.2,0.15498239994049073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,4,power_law_1.2,0.24734079837799072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,4,power_law_1.2,0.27400319576263427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,4,power_law_1.2,0.2805887937545776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,4,power_law_1.2,0.2873408079147339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,4,power_law_1.2,0.286028790473938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,4,power_law_1.2,0.3010175943374634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,4,power_law_1.2,0.30899839401245116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,4,power_law_1.2,0.3141632080078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,4,power_law_1.2,0.3210752010345459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,4,power_law_1.2,0.32487680912017824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,4,power_law_1.2,0.35674879550933836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,4,power_law_1.2,0.06977279782295227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,4,power_law_1.2,0.3473344087600708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,4,power_law_1.2,0.09908480048179627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,4,power_law_1.2,0.39324159622192384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,4,power_law_1.2,0.06025599837303162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,4,power_law_1.2,0.4392064094543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,4,power_law_1.2,0.08190720081329346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,4,power_law_1.2,0.07186560034751892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,4,power_law_1.2,0.5080575942993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,4,power_law_1.2,0.09119359850883484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,4,power_law_1.2,0.08504319787025452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,4,power_law_1.2,0.11034879684448243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,4,power_law_1.2,0.5702591896057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,4,power_law_1.2,0.10801279544830322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,4,power_law_1.2,0.13724160194396973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,4,power_law_1.2,0.7849599838256835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,4,power_law_1.2,0.1963711977005005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,4,power_law_1.2,0.14910720586776732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,4,power_law_1.2,0.7164159774780273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,4,power_law_1.2,0.20507519245147704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,4,power_law_1.2,0.1686336040496826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,4,power_law_1.2,0.9155584335327148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,4,power_law_1.2,0.1666815996170044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,4,power_law_1.2,0.21484160423278809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,4,power_law_1.2,1.2518848419189452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,4,power_law_1.2,0.1696320056915283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,4,power_law_1.2,0.22209279537200927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,4,power_law_1.2,0.1720896005630493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,4,power_law_1.2,0.03720319867134094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,4,power_law_1.2,1.6188543319702149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,4,power_law_1.2,0.22473599910736083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,4,power_law_1.2,0.1787775993347168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,4,power_law_1.2,0.22922880649566652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,4,power_law_1.2,2.2337343215942385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,4,power_law_1.2,0.025440001487731935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,4,power_law_1.2,0.1843840003013611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,4,power_law_1.2,0.23428480625152587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,4,power_law_1.2,0.040320000052452086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,4,power_law_1.2,0.2293247938156128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,4,power_law_1.2,0.18866560459136963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,4,power_law_1.2,2.7143743515014647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,4,power_law_1.2,0.23279359340667724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,4,power_law_1.2,0.050393599271774295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,4,power_law_1.2,0.24736640453338624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,4,power_law_1.2,0.058976000547409056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,4,power_law_1.2,0.1861567974090576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,4,power_law_1.2,0.2501503944396973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,4,power_law_1.2,3.9704513549804688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,4,power_law_1.2,0.0828927993774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,4,power_law_1.2,0.2112895965576172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,4,power_law_1.2,0.25830399990081787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,4,power_law_1.2,0.09056640267372132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,4,power_law_1.2,0.21619200706481934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,4,power_law_1.2,0.26975998878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,4,power_law_1.2,0.08855680227279664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,4,power_law_1.2,0.21971840858459474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,4,power_law_1.2,7.777734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.2,0.3119935989379883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,4,power_law_1.2,0.09774720072746276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,4,power_law_1.2,0.25269761085510256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,4,power_law_1.2,0.1004480004310608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.2,0.3433471918106079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,4,power_law_1.2,0.2786367893218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,4,power_law_1.2,0.1051967978477478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.2,0.424345588684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,4,power_law_1.2,0.34503679275512694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,4,power_law_1.2,0.11210240125656128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,4,power_law_1.2,0.39836161136627196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.2,0.50316162109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,4,power_law_1.2,0.11891839504241944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,4,power_law_1.2,0.506496000289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.2,0.583897590637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,4,power_law_1.2,0.11393280029296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,4,power_law_1.2,0.555788803100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.2,0.7502592086791993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,4,power_law_1.2,0.12226560115814208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,4,power_law_1.2,0.8038592338562012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.2,1.048307228088379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,4,power_law_1.2,0.12878719568252564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,4,power_law_1.2,0.9943360328674317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.2,1.3677311897277833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,4,power_law_1.2,0.12462079524993896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,4,power_law_1.2,1.3595392227172851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.2,2.0588735580444335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,4,power_law_1.2,0.14711040258407593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,4,power_law_1.2,1.7915071487426757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.2,0.1843392014503479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,4,power_law_1.2,2.3538816452026365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,4,power_law_1.2,2.1182783126831053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.2,0.19717760086059571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,4,power_law_1.2,3.4226303100585938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,4,power_law_1.2,3.602873611450195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.2,0.22398080825805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.2,0.2560447931289673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,2,balanced,0.04985066751639048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.2,0.3170687913894653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,4,power_law_1.2,7.304768371582031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,4,power_law_1.2,7.703238677978516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.2,0.4139391899108887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,2,balanced,0.07364266614119212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,2,balanced,0.10045866171518962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.2,0.508742380142212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,2,balanced,0.15410133202870688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,2,balanced,0.26202666759490967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.2,0.7090816020965576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,2,balanced,0.31427733103434247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,2,balanced,0.31496532758076984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,2,balanced,0.3175999919573466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.2,0.933779239654541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,2,balanced,0.31946667035420734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,2,balanced,0.3190240065256755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,4,power_law_1.2,1.1871871948242188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,2,balanced,0.32263465722401935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,2,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,2,balanced,0.3258826732635498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,2,balanced,0.05606933434804281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,2,balanced,0.3284533421198527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,4,power_law_1.2,1.81080322265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,2,balanced,0.05227200190226237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,2,balanced,0.33099732796351117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,2,balanced,0.05997333427270254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,2,balanced,0.07260799904664357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,2,balanced,0.3426133394241333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,2,balanced,0.07297066847483318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,2,balanced,0.10644267002741496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,4,power_law_1.2,3.6634559631347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,2,balanced,0.1023466686407725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,2,balanced,0.1630346675713857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,2,balanced,0.3438560167948405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,2,balanced,0.1879840095837911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,2,balanced,0.14388799667358398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,2,balanced,0.19115199645360312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,2,balanced,0.16524266203244528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,2,balanced,0.35685332616170246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,2,balanced,0.1941493352254232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,2,balanced,0.16739734013875326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,2,balanced,0.1704053282737732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,2,balanced,0.1955946683883667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,2,balanced,0.37892266114552814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,2,balanced,0.17122666041056314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,2,balanced,0.19619733095169067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,2,balanced,0.025349333882331848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,2,balanced,0.17247466246287027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,2,balanced,0.20012799898783365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,2,balanced,0.3899093468983968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,2,balanced,0.17595734198888144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,2,balanced,0.20362667242685953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,2,balanced,0.028677334388097126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,2,balanced,0.17692800362904867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,2,balanced,0.203984002272288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,2,balanced,0.04215999941031138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,2,balanced,0.4788586695988973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,2,balanced,0.17876267433166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,2,balanced,0.20773865779240927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,2,balanced,0.059706668059031166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,2,balanced,0.1865546703338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,2,balanced,0.08724799752235413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,2,balanced,0.21741332610448202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,2,balanced,0.10542399684588115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,2,balanced,0.4721493323644002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,2,balanced,0.22405334313710532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,2,balanced,0.19233600298563638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,2,balanced,0.11105066537857056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,2,balanced,0.11504532893498738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,2,balanced,0.23516800006230673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,2,balanced,0.1987733244895935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,2,balanced,0.12171199917793274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,2,balanced,0.20975999037424722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,2,balanced,0.11459733049074809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,2,balanced,0.25934932629267377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,2,balanced,0.6693973541259766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,2,balanced,0.11989333232243855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,2,balanced,0.2353760004043579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,2,balanced,0.27719465891520184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,2,balanced,0.1255413293838501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,2,balanced,0.12744533022244772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,2,balanced,0.2570613423983256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,2,balanced,0.3498400052388509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,2,balanced,0.6439946492513021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,2,balanced,0.1316159963607788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,2,balanced,0.1318773329257965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,2,balanced,0.33371198177337646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,2,balanced,0.37912531693776447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,2,balanced,0.1720106601715088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,2,balanced,0.1709973414738973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,2,balanced,0.34831468264261883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,2,balanced,0.2593226631482442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,2,balanced,0.5546773274739584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,2,balanced,0.8958346843719482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,2,balanced,0.29291200637817383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,2,balanced,0.5024640162785848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,2,balanced,0.27775466442108154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,2,balanced,0.592682679494222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,2,balanced,0.5462239980697632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,2,balanced,1.0405973593393962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,2,balanced,0.2897760073343913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,2,balanced,0.8390666643778483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,2,balanced,0.3193173408508301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,2,balanced,0.7985066572825114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,2,balanced,0.33672531445821124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,2,balanced,1.6823679606119792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,2,balanced,1.0034026304880779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,2,balanced,0.9821653366088867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,2,balanced,0.5394239823023478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,2,balanced,0.6115520000457764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,2,balanced,1.5553919474283855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,2,balanced,1.4357013702392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,2,balanced,1.9273440043131511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,2,balanced,0.875279982884725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,2,balanced,1.863525390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,2,balanced,1.8238239288330078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,2,balanced,2.2179412841796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,2,balanced,1.1471786499023438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,2,balanced,2.2146666844685874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,2,balanced,2.1821066538492837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,2,power_law_1.01,0.07082239985466003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,2,balanced,1.2600693702697754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,2,power_law_1.01,0.08648319840431214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,2,power_law_1.01,0.08205440044403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,2,power_law_1.01,0.12238719463348388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,2,power_law_1.01,0.16903680562973022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,2,power_law_1.01,0.2473088026046753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,2,power_law_1.01,0.26873600482940674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,2,power_law_1.01,0.27806720733642576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,2,balanced,3.443087895711263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,2,power_law_1.01,0.28051199913024905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,2,power_law_1.01,0.285862398147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,2,power_law_1.01,0.28913280963897703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,2,power_law_1.01,0.299564790725708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,2,power_law_1.01,0.3097471952438354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,2,power_law_1.01,0.31985280513763426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,2,power_law_1.01,0.33520638942718506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,2,power_law_1.01,0.3415359973907471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,2,power_law_1.01,0.05017600059509277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,2,power_law_1.01,0.3395711898803711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,2,power_law_1.01,0.3897088050842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,2,power_law_1.01,0.0671999990940094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,2,balanced,3.474426587422689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,2,balanced,3.396047910054525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,2,power_law_1.01,0.41695361137390136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,2,power_law_1.01,0.064300799369812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,2,power_law_1.01,0.48750720024108884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,2,power_law_1.01,0.0905023992061615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,2,power_law_1.01,0.5712448120117187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,2,power_law_1.01,0.10997120141983033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,2,power_law_1.01,0.15884799957275392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,2,power_law_1.01,0.718067216873169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,2,power_law_1.01,0.1687872052192688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,2,power_law_1.01,0.7555712223052978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,2,power_law_1.01,0.17242879867553712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,2,power_law_1.01,0.9630720138549804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,2,power_law_1.01,0.17745920419692993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,2,power_law_1.01,1.338700771331787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,2,power_law_1.01,0.1779711961746216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,2,balanced,2.0696372985839844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,2,power_law_1.01,1.7325120925903321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,2,power_law_1.01,0.05841919779777527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,2,power_law_1.01,0.18575359582901002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,2,power_law_1.01,2.2585792541503906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,2,power_law_1.01,0.06472319960594178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,2,power_law_1.01,0.19170559644699098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,2,power_law_1.01,2.622208023071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,2,power_law_1.01,0.19633920192718507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,2,power_law_1.01,0.06542080044746398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,2,power_law_1.01,0.20733439922332764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,2,power_law_1.01,0.09196799993515015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,2,power_law_1.01,3.852511978149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,2,power_law_1.01,0.2156287908554077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,2,power_law_1.01,0.1068160057067871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,2,power_law_1.01,0.22847359180450438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,2,power_law_1.01,0.14295680522918702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,2,power_law_1.01,0.23632640838623048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,2,power_law_1.01,0.14852479696273804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,2,power_law_1.01,7.337126159667969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,2,power_law_1.01,0.2697088003158569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,2,power_law_1.01,0.1522047996520996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,2,power_law_1.01,0.2977855920791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,2,power_law_1.01,0.15646079778671265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,2,power_law_1.01,0.37487359046936036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,2,power_law_1.01,0.16255999803543092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,2,power_law_1.01,0.023161600530147552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,2,power_law_1.01,0.16506240367889405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,2,power_law_1.01,0.4363584041595459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,2,power_law_1.01,0.026624000072479247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,2,power_law_1.01,0.16656639575958251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,2,power_law_1.01,0.5912320137023925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,2,power_law_1.01,0.17537920475006102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,2,power_law_1.01,0.032742398977279666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,2,power_law_1.01,0.6859072208404541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,2,power_law_1.01,0.18272639513015748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,2,power_law_1.01,0.04857600033283234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,2,power_law_1.01,0.8598912239074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,2,power_law_1.01,0.19544960260391236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,2,power_law_1.01,0.060499197244644164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,2,power_law_1.01,1.215494441986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,2,power_law_1.01,0.20798079967498778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,2,power_law_1.01,0.08046079874038696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,2,power_law_1.01,0.21548159122467042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,2,power_law_1.01,1.6950975418090821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,2,power_law_1.01,0.09060479998588562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,2,power_law_1.01,0.24294400215148926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,2,power_law_1.01,0.09513599872589111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.01,0.2693824052810669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,2,power_law_1.01,2.061996841430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.01,0.3355776071548462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,2,power_law_1.01,0.09751679897308349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,2,power_law_1.01,2.48288631439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,2,power_law_1.01,0.09251199960708618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.01,0.3926336050033569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,2,power_law_1.01,0.10144640207290649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,2,balanced,6.7711842854817705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.01,0.5219647884368896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,2,power_law_1.01,3.5867454528808596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,2,power_law_1.01,0.10773119926452637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.01,0.6013567924499512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,2,power_law_1.01,0.10882560014724732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.01,0.8150272369384766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,2,power_law_1.01,0.10886399745941162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,2,power_law_1.01,6.998713684082031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.01,1.067302417755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,2,power_law_1.01,0.11464320421218872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.01,1.6351423263549805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,2,power_law_1.01,0.12363519668579101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,2,power_law_1.01,0.1268224000930786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.01,1.9728384017944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,2,power_law_1.01,0.23724799156188964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,2,balanced,6.587690353393555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,2,power_law_1.01,2.3795520782470705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,2,balanced,6.711413065592448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.01,0.2511039972305298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.01,0.2100287914276123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,2,power_law_1.01,3.7465152740478516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.01,0.22686080932617186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.01,0.25432960987091063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.01,0.29235200881958007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,2,power_law_1.01,7.47357406616211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.01,0.38045439720153806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,2,balanced,3.9372425079345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.01,0.4757120132446289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.01,0.6593408107757568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.01,0.8564736366271972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,2,power_law_1.01,1.0112256050109862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,2,power_law_1.01,1.545145606994629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,2,power_law_1.01,2.998905563354492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,2,power_law_1.2,0.07096319794654846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,2,power_law_1.2,0.05902079939842224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,2,power_law_1.2,0.0828927993774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,2,power_law_1.2,0.05181440114974976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,2,power_law_1.2,0.10904959440231324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,2,power_law_1.2,0.14858239889144897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,2,power_law_1.2,0.24821760654449462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,2,power_law_1.2,0.050355201959609984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,2,power_law_1.2,0.262227201461792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,2,power_law_1.2,0.06348159909248352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,2,power_law_1.2,0.2760447978973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,2,power_law_1.2,0.08476160168647766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,2,power_law_1.2,0.29580800533294677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,2,power_law_1.2,0.10363520383834839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,2,power_law_1.2,0.3040895938873291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,2,power_law_1.2,0.15537279844284058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,2,power_law_1.2,0.30320000648498535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,2,power_law_1.2,0.1639680027961731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,2,power_law_1.2,0.30875520706176757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,2,power_law_1.2,0.17102080583572388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,2,power_law_1.2,0.3208256006240845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,2,power_law_1.2,0.1758784055709839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,2,power_law_1.2,0.32712318897247317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,2,power_law_1.2,0.18362879753112793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,2,power_law_1.2,0.34442241191864015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,2,power_law_1.2,0.19011839628219604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,2,power_law_1.2,0.34295039176940917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,2,power_law_1.2,0.19608960151672364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,2,power_law_1.2,0.3509119987487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,2,power_law_1.2,0.1962623953819275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,2,power_law_1.2,0.39901440143585204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,2,power_law_1.2,0.20938239097595215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,2,power_law_1.2,0.055052798986434934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,2,power_law_1.2,0.43364481925964354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,2,power_law_1.2,0.2210239887237549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,2,power_law_1.2,0.49683837890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,2,power_law_1.2,0.05627520084381103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,2,power_law_1.2,0.22746880054473878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,2,power_law_1.2,0.5897664070129395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,2,power_law_1.2,0.06439039707183838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,2,power_law_1.2,0.24506239891052245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,2,power_law_1.2,0.023475199937820435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,2,power_law_1.2,0.7407872200012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,1,balanced,0.05261866748332977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,2,power_law_1.2,0.08295680284500122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,2,power_law_1.2,0.2751744031906128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,2,power_law_1.2,0.7433407783508301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,2,power_law_1.2,0.024281600117683412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,2,power_law_1.2,0.09526399970054626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,2,power_law_1.2,0.3045120000839233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,2,power_law_1.2,1.0119808197021485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,1,balanced,0.07119999825954437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,2,power_law_1.2,0.13551360368728638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,2,power_law_1.2,0.030355200171470642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,1,balanced,0.10014399886131287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,2,power_law_1.2,0.37285120487213136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,2,power_law_1.2,1.4385087966918946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,2,power_law_1.2,0.14249600172042848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,2,power_law_1.2,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,1,balanced,0.15798933307329813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,2,power_law_1.2,0.46405758857727053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,2,power_law_1.2,0.05541120171546936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,2,power_law_1.2,0.15308159589767456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,2,power_law_1.2,1.9642368316650392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,1,balanced,0.2667413353919983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,2,power_law_1.2,0.08080000281333924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,2,power_law_1.2,0.15912959575653077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,2,power_law_1.2,0.5832960128784179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,1,balanced,0.32396266857783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,2,power_law_1.2,2.2869247436523437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,2,power_law_1.2,0.08858879804611205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,1,balanced,0.3258453408877055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,2,power_law_1.2,0.1596735954284668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,2,power_law_1.2,0.7071296215057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,2,power_law_1.2,0.09475839734077454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,1,balanced,0.32825066645940143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,2,power_law_1.2,0.16525440216064452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,2,power_law_1.2,2.8869632720947265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,2,power_law_1.2,0.941107177734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,1,balanced,0.3296053409576416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,2,power_law_1.2,0.10252159833908081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,2,power_law_1.2,0.17063679695129394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,1,balanced,0.04975466430187225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,2,power_law_1.2,1.2294143676757812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,1,balanced,0.331989328066508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,2,power_law_1.2,0.09756799936294555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,2,power_law_1.2,0.18007680177688598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,2,power_law_1.2,4.269619369506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,1,balanced,0.3381173213322957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,1,power_law_1.01,0.05018240213394165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,2,power_law_1.2,0.10362880229949951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,2,power_law_1.2,1.6773439407348634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,2,power_law_1.2,0.18611840009689332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,1,balanced,0.05801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,1,balanced,0.33954131603240967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,2,power_law_1.2,0.10792319774627686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,2,power_law_1.2,0.19644800424575806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,1,balanced,0.07881600161393483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,1,power_law_1.01,0.045177599787712096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,1,balanced,0.34519465764363605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,2,power_law_1.2,2.233132743835449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,1,power_law_1.01,0.06403840184211732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,2,power_law_1.2,7.398892974853515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,1,balanced,0.11351466178894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,2,power_law_1.2,0.10618239641189575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,1,balanced,0.3540053367614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,2,power_law_1.2,0.2118272066116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,1,power_law_1.01,0.08076800107955932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,1,balanced,0.17874133586883545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,2,power_law_1.2,2.674220848083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,2,power_law_1.2,0.10793600082397461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,1,balanced,0.36533868312835693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,2,power_law_1.2,0.21860480308532715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,1,power_law_1.01,0.05219200253486633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,1,power_law_1.01,0.12420480251312256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,1,balanced,0.20933334032694498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,2,power_law_1.2,0.2507967948913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,1,balanced,0.3704426685969035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,1,power_law_1.01,0.1622015953063965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,2,power_law_1.2,0.11592320203781128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,1,balanced,0.2156053384145101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,1,power_law_1.01,0.06476160287857055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,2,power_law_1.2,4.071142578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.2,0.2866175889968872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,1,balanced,0.3858453432718913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,1,power_law_1.01,0.24800000190734864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,2,power_law_1.2,0.1268224000930786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.2,0.339520001411438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,1,power_law_1.01,0.2705343961715698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,1,power_law_1.01,0.09070079922676086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,1,balanced,0.21556266148885092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,1,balanced,0.41858665148417157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,2,power_law_1.2,0.1242751955986023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,1,power_law_1.01,0.2815040111541748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.2,0.40885119438171386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,1,power_law_1.01,0.11632000207901001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,1,balanced,0.21893332401911417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,2,power_law_1.2,7.872409820556641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,1,power_law_1.01,0.29062399864196775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,1,balanced,0.22097599506378174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,1,power_law_1.01,0.17216639518737792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,2,power_law_1.2,0.23564159870147705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,1,balanced,0.4530133406321208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.2,0.5276864051818848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,1,power_law_1.01,0.2929728031158447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,1,balanced,0.22740266720453897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,1,power_law_1.01,0.1844480037689209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,1,power_law_1.01,0.30612480640411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.2,0.2531712055206299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.2,0.629030418395996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,1,balanced,0.232095996538798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,1,power_law_1.01,0.19431040287017823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,1,balanced,0.5616586605707804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,1,power_law_1.01,0.31550719738006594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,1,balanced,0.23663999636967978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,1,power_law_1.01,0.19892480373382568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.2,0.21799039840698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.2,0.8231488227844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,1,power_law_1.01,0.20022399425506593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,1,power_law_1.01,0.3272320032119751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,1,balanced,0.24521066745122275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.2,0.2355263948440552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.2,1.0796607971191405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,1,power_law_1.01,0.3319295883178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,1,balanced,0.5916586716969808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,1,power_law_1.01,0.21155200004577637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,1,balanced,0.25257599353790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.2,0.2668479919433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.2,1.6307199478149415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,1,power_law_1.01,0.34627199172973633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,1,power_law_1.01,0.2180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,1,balanced,0.8910133043924967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.2,0.30979199409484864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,1,balanced,0.2664480010668437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.2,2.0481599807739257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,1,power_law_1.01,0.3564863920211792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,1,power_law_1.01,0.23160960674285888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.2,0.4259136199951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,1,balanced,0.8855679829915365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.2,0.49891839027404783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,1,power_law_1.01,0.23729920387268066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,1,power_law_1.01,0.3828223943710327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,1,balanced,0.2837120095888774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,2,power_law_1.2,2.5004608154296877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,1,balanced,1.4565173784891765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.2,0.6591360092163085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,1,power_law_1.01,0.42641921043395997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,1,power_law_1.01,0.25049600601196287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,1,balanced,0.3171573281288147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,2,power_law_1.2,4.031667327880859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,1,power_law_1.01,0.47351679801940916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,1,power_law_1.01,0.2582079887390137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.2,0.8871808052062988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,1,balanced,0.3481546640396118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,2,power_law_1.2,7.4649406433105465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,1,power_law_1.01,0.5716095924377441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,1,power_law_1.01,0.28387839794158937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,1,power_law_1.01,0.325382399559021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,1,power_law_1.01,0.6795135974884033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,2,power_law_1.2,1.026905632019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,1,balanced,0.4771466652552287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,1,power_law_1.01,0.37499520778656004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,1,balanced,1.5482667287190754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,1,power_law_1.01,0.8627840042114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,2,power_law_1.2,1.5825792312622071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,1,power_law_1.01,0.46725120544433596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,1,balanced,0.5038293202718099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,1,power_law_1.01,1.0749759674072266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,1,power_law_1.01,1.3984640121459961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,1,power_law_1.01,0.553056001663208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,1,power_law_1.01,1.8036224365234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,2,power_law_1.2,3.147763252258301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,1,power_law_1.01,0.7295487880706787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,1,balanced,0.7454133033752441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,1,power_law_1.01,2.4345983505249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,1,power_law_1.01,0.9609919548034668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,1,power_law_1.01,2.96363525390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,1,power_law_1.01,1.2973888397216797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,1,power_law_1.01,3.5834049224853515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,1,balanced,2.4946239789326987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,1,power_law_1.01,1.6724863052368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,1,balanced,0.8470719655354818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,1,power_law_1.01,5.39728012084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,1,power_law_1.01,2.284390449523926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,1,power_law_1.01,2.8883520126342774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,1,power_law_1.01,3.5532352447509767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,1,power_law_1.01,10.259846496582032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,1,balanced,1.3289546966552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,1,power_law_1.01,5.493280029296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,1,power_law_1.01,10.766802978515624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,1,balanced,1.5420479774475098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,1,balanced,2.8911733627319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,1,balanced,2.33079465230306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,1,balanced,3.3589760462443032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,1,balanced,2.936496098836263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,1,balanced,3.49509334564209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,1,balanced,5.460133234659831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,1,power_law_1.2,0.04965760111808777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,1,power_law_1.2,0.06156799793243408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,1,power_law_1.2,0.08154240250587463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,1,power_law_1.2,0.11210880279541016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,1,power_law_1.2,0.14885760545730592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,1,power_law_1.2,0.24753921031951903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,1,balanced,5.5270131429036455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,1,power_law_1.2,0.27068800926208497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,1,power_law_1.2,0.2867072105407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,1,power_law_1.2,0.2936448097229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,1,power_law_1.2,0.3059200048446655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,1,power_law_1.2,0.31242239475250244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,1,power_law_1.2,0.3288127899169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,1,power_law_1.2,0.3337599992752075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,1,power_law_1.2,0.3441087961196899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,1,power_law_1.2,0.0472896009683609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,1,power_law_1.2,0.35900158882141114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,1,power_law_1.2,0.053260797262191774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,1,power_law_1.2,0.36643199920654296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,1,power_law_1.2,0.06645119786262513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,1,power_law_1.2,0.3908479928970337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,1,power_law_1.2,0.08799359798431397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,1,power_law_1.2,0.43601279258728026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,1,power_law_1.2,0.11136000156402588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,1,power_law_1.2,0.4832064151763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,1,power_law_1.2,0.1648576021194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,1,power_law_1.2,0.5744128227233887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,1,power_law_1.2,0.18553600311279297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,1,power_law_1.2,0.6941440105438232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,1,power_law_1.2,0.19242240190505983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,1,power_law_1.2,0.8777728080749512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,1,power_law_1.2,0.2009727954864502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,1,power_law_1.2,1.0998592376708984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,1,power_law_1.2,0.20723841190338135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,1,power_law_1.2,1.4252351760864257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,1,power_law_1.2,0.22001919746398926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,1,power_law_1.2,0.2245311975479126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,1,power_law_1.2,1.8311296463012696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,1,power_law_1.2,0.2359679937362671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,1,power_law_1.2,2.4471103668212892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,1,power_law_1.2,0.24223999977111815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,1,power_law_1.2,2.98525447845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,1,power_law_1.2,0.2557440042495728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,1,power_law_1.2,0.2668479919433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,1,power_law_1.2,3.6100479125976563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,1,power_law_1.2,0.28894720077514646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,1,balanced,10.358751932779947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,1,power_law_1.2,0.33491199016571044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,1,power_law_1.2,5.441798400878906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,1,power_law_1.2,0.3826303958892822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,8,balanced,0.04990399877230326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,1,power_law_1.2,0.4770751953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,1,power_law_1.2,0.5570112228393554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,1,power_law_1.2,10.22131805419922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,8,balanced,0.06676800052324931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,1,power_law_1.2,0.732588815689087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,8,balanced,0.09202667077382405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,1,power_law_1.2,0.9687680244445801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,8,balanced,0.14537066221237183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,1,power_law_1.2,1.3001407623291015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,8,balanced,0.24503467480341592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,8,balanced,0.44432000319163006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,1,power_law_1.2,1.6742656707763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,8,balanced,0.6446133454640707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,1,power_law_1.2,2.302751922607422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,8,balanced,0.6380693515141805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,8,balanced,0.6382026672363281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,1,power_law_1.2,2.8843711853027343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,8,balanced,0.6438133319218954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,8,balanced,0.6488053401311239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,1,power_law_1.2,3.560038375854492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,8,balanced,0.6508106787999471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,1,balanced,10.705963134765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,1,power_law_1.2,5.499110412597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,8,balanced,0.659279982248942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,8,balanced,0.6499093373616537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,8,balanced,0.6612319946289062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,1,power_law_1.2,10.772319793701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,8,balanced,0.6641866763432821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,8,balanced,0.6671093304951986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,8,balanced,0.6811573505401611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,8,balanced,0.688922643661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,8,balanced,0.7131040096282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,8,balanced,0.04587199787298838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,8,balanced,0.054144000013669334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,8,balanced,0.7435466448465983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,8,balanced,0.049829334020614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,8,balanced,0.06841066479682922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,8,balanced,0.0993333359559377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,8,balanced,0.05585599939028422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,8,balanced,0.06215466558933258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,8,balanced,0.1532639960447947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,8,balanced,0.7945333321889242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,8,balanced,0.08662933111190796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,8,balanced,0.12150933345158894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,8,balanced,0.2480693260828654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,8,balanced,0.17808000246683756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,8,balanced,0.35201601187388104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,8,balanced,0.24117867151896158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,8,balanced,0.34969600041707355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,8,balanced,0.24127999941507974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,8,balanced,0.3507946729660034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,8,balanced,0.2420533299446106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,8,balanced,0.8395573298136393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,8,power_law_1.01,0.14691200256347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,8,balanced,0.352453351020813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,8,balanced,0.24213866392771402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,8,balanced,0.35462931791941327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,8,power_law_1.01,0.24216320514678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,8,balanced,0.35755733648935956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,8,balanced,0.24677334229151407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,8,power_law_1.01,0.4351935863494873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,8,balanced,0.3604160149892171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,8,balanced,0.24684266249338785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,8,balanced,0.3616906801859538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,8,power_law_1.01,0.14068479537963868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,8,balanced,0.247488001982371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,8,balanced,0.370037317276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,8,balanced,0.9126613140106201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,8,power_law_1.01,0.22832000255584717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,8,balanced,0.2525706688563029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,8,balanced,0.3731360038121541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,8,balanced,0.2584639986356099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,8,balanced,0.3817226489384969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,8,power_law_1.01,0.30730879306793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,8,balanced,0.26154667139053345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,8,balanced,0.3953333298365275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,8,power_law_1.01,0.36839680671691893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,8,balanced,0.2688159942626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,8,balanced,0.4083520174026489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,8,power_law_1.01,0.45233922004699706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,8,balanced,0.28038400411605835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,8,balanced,1.0585973262786865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,8,power_law_1.01,0.4905407905578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,8,balanced,0.2914399902025859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,8,balanced,0.44490134716033936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,8,power_law_1.01,0.48721919059753416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,8,power_law_1.01,0.491648006439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,8,balanced,0.3156320055325826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,8,balanced,0.4697546561559041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,8,power_law_1.01,0.49781122207641604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,8,power_law_1.01,0.5030464172363281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,8,balanced,0.33885331948598224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,8,power_law_1.01,0.5334847927093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,8,balanced,0.5250720183054606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,8,balanced,0.39236799875895184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,8,balanced,1.2163466612497966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,8,power_law_1.01,0.5685696125030517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,8,power_law_1.01,0.5516863822937011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,8,balanced,0.5858506759007772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,8,power_law_1.01,0.5715136051177978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,8,power_law_1.01,0.6102464199066162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,8,balanced,0.44044800599416095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,8,power_law_1.01,0.6894144058227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,8,power_law_1.01,0.7817215919494629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,8,power_law_1.01,0.7909823894500733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,8,balanced,0.6952160199483236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,8,power_law_1.01,0.9826623916625976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,8,balanced,0.5424160162607828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,8,power_law_1.01,1.0922752380371095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,8,power_law_1.01,1.4744128227233886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,8,balanced,1.656218687693278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,8,power_law_1.01,1.384275245666504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,8,power_law_1.01,1.728550338745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,8,balanced,0.8430079619089762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,8,power_law_1.01,2.3300544738769533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,8,balanced,0.6798240343729655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,8,power_law_1.01,2.6930368423461912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,8,power_law_1.01,3.5934463500976563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,8,power_law_1.01,6.92015380859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,8,balanced,1.0585493246714275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,8,balanced,0.8647092978159586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,8,balanced,1.9185706774393718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,8,balanced,1.3814880053202312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,8,balanced,1.1264959971110027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,8,balanced,2.6721601486206055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,8,balanced,1.7009600003560383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,8,balanced,1.388389269510905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,8,power_law_1.01,0.0909056007862091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,8,power_law_1.01,0.14220160245895386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,8,power_law_1.01,0.2393280029296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,8,power_law_1.01,0.09962239861488342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,8,power_law_1.01,0.08469120264053345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,8,power_law_1.01,0.1283519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,8,power_law_1.01,0.1819584012031555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,8,power_law_1.01,0.11222399473190307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,8,power_law_1.01,0.20929279327392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,8,power_law_1.01,0.16959999799728392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,8,power_law_1.01,0.2529792070388794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,8,power_law_1.01,0.0821120023727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,8,power_law_1.01,0.2678976058959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,8,power_law_1.01,0.265068793296814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,8,power_law_1.01,0.10688639879226684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,8,power_law_1.01,0.27873280048370364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,8,power_law_1.01,0.28672640323638915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,8,power_law_1.01,0.13647359609603882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,8,power_law_1.01,0.2865535974502563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,8,power_law_1.01,0.14959360361099244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,8,power_law_1.2,0.14878079891204835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,8,power_law_1.01,0.3039360046386719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,8,power_law_1.01,0.17666560411453247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,8,power_law_1.01,0.3196608066558838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,8,power_law_1.01,0.1820479989051819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,8,power_law_1.2,0.24161920547485352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,8,power_law_1.01,0.3166912078857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,8,power_law_1.01,0.1850368022918701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,8,balanced,2.4809014002482095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,8,power_law_1.2,0.2451200008392334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,8,power_law_1.01,0.19685120582580568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,8,power_law_1.01,0.3592832088470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,8,power_law_1.2,0.14661120176315307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,8,power_law_1.01,0.36522879600524905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,8,power_law_1.01,0.2036992073059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,8,power_law_1.2,0.2281791925430298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,8,balanced,2.070144017537435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,8,power_law_1.01,0.40285439491271974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,8,power_law_1.01,0.20826239585876466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,8,power_law_1.2,0.2508928060531616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,8,power_law_1.01,0.48514561653137206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,8,power_law_1.01,0.20896639823913574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,8,power_law_1.2,0.30743680000305174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,8,power_law_1.2,0.09325439929962158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,8,power_law_1.01,0.49697279930114746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,8,power_law_1.01,0.22760319709777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,8,power_law_1.2,0.42288641929626464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,8,power_law_1.01,0.23063039779663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,8,power_law_1.01,0.6048384189605713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,8,power_law_1.2,0.14381439685821534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,8,power_law_1.2,0.4337791919708252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,8,power_law_1.01,0.2344575881958008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,8,power_law_1.01,0.7154176235198975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,8,power_law_1.2,0.11171200275421142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,8,power_law_1.2,0.45425920486450194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,8,power_law_1.01,0.25754880905151367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,8,power_law_1.01,0.9072319984436035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,8,power_law_1.01,0.2761663913726807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,8,power_law_1.2,0.4463488101959229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,8,power_law_1.2,0.09875839948654175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,8,power_law_1.01,0.30850560665130616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,8,power_law_1.2,0.5034111976623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,8,power_law_1.2,0.1322111964225769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,8,power_law_1.01,1.0217344284057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,8,power_law_1.01,0.3532288074493408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,8,power_law_1.2,0.479366397857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,8,power_law_1.2,0.1655295968055725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,8,power_law_1.01,1.338963222503662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,8,power_law_1.01,0.4380159854888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,8,power_law_1.2,0.4990719795227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,8,power_law_1.2,0.20965759754180907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,8,power_law_1.01,1.6467456817626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,8,power_law_1.01,0.49049601554870603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,8,power_law_1.2,0.5610239982604981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,8,power_law_1.2,0.2503040075302124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,8,power_law_1.01,0.6429952144622803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,8,power_law_1.2,0.5620863914489747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,8,power_law_1.01,2.090380859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,8,power_law_1.2,0.23719680309295654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,8,power_law_1.01,0.790886402130127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,8,power_law_1.2,0.6089536190032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,8,balanced,5.160010655721028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,8,power_law_1.2,0.25696001052856443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,8,power_law_1.01,1.0738559722900392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,8,power_law_1.2,0.6160895824432373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,8,power_law_1.01,3.222265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,8,power_law_1.2,0.25056641101837157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,8,power_law_1.01,1.322662353515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,8,power_law_1.2,0.7060671806335449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,8,power_law_1.2,0.27451519966125487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,8,power_law_1.01,1.6807424545288085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,8,power_law_1.2,0.8304896354675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,8,power_law_1.01,5.89502067565918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,8,power_law_1.2,0.28792319297790525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,8,power_law_1.01,2.4874048233032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,8,power_law_1.2,0.8340288162231445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,8,power_law_1.2,0.3028543949127197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,8,power_law_1.2,1.0436672210693358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,8,power_law_1.01,4.902444839477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,8,power_law_1.2,0.3132607936859131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,8,power_law_1.2,1.1737471580505372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,8,power_law_1.2,0.32135040760040284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,8,power_law_1.2,1.5615679740905761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,8,power_law_1.2,0.3332288026809692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,8,power_law_1.2,1.4325759887695313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,8,power_law_1.2,0.37989120483398436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,8,power_law_1.2,2.0052799224853515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,8,power_law_1.2,0.4276927947998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,8,power_law_1.2,2.8799039840698244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,8,power_law_1.2,0.5120895862579345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,8,power_law_1.2,3.336966323852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,8,power_law_1.2,0.5377024173736572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,8,power_law_1.2,0.6687935829162598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,8,power_law_1.2,4.102995300292969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,8,power_law_1.2,0.7197504043579102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,8,power_law_1.2,0.9660096168518066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,8,power_law_1.2,8.32200927734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,8,balanced,4.759077390034993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,8,power_law_1.2,1.0843968391418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,8,power_law_1.2,1.4550399780273438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,8,power_law_1.2,1.9023168563842774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,8,balanced,3.981701215108236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,8,power_law_1.2,2.3854272842407225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,8,power_law_1.2,3.6237823486328127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,8,power_law_1.2,6.408774566650391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,8,power_law_1.2,0.08451840281486511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,8,power_law_1.2,0.11251840591430665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,8,power_law_1.2,0.09101439714431762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,8,power_law_1.2,0.08712319731712341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,8,power_law_1.2,0.1016767978668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,8,power_law_1.2,0.1267135977745056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,8,power_law_1.2,0.13893760442733766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,8,power_law_1.2,0.17656960487365722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,8,power_law_1.2,0.1767680048942566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,8,power_law_1.2,0.1882367968559265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,8,power_law_1.2,0.19211519956588746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,8,power_law_1.2,0.20055680274963378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,8,power_law_1.2,0.20504961013793946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,8,power_law_1.2,0.2050368070602417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,8,power_law_1.2,0.2221695899963379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,8,power_law_1.2,0.2240895986557007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,8,power_law_1.2,0.24320640563964843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,8,power_law_1.2,0.2687551975250244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,8,power_law_1.2,0.2828480005264282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,8,power_law_1.2,0.33501439094543456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,8,power_law_1.2,0.36710400581359864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,8,power_law_1.2,0.5009727954864502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,8,power_law_1.2,0.5141119956970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,8,power_law_1.2,0.6960959911346436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,8,power_law_1.2,0.9158592224121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,8,power_law_1.2,1.186905574798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,8,power_law_1.2,1.4826175689697265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,8,power_law_1.2,1.9816383361816405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,8,power_law_1.2,2.748748779296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,8,power_law_1.2,5.467142486572266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,4,balanced,0.05186133086681366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,4,balanced,0.06634133557478587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,4,balanced,0.09294933080673218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,4,balanced,0.14662933349609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,4,balanced,0.24727465709050497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,4,balanced,0.4450720151265462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,4,balanced,0.6438506841659546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,4,balanced,0.6410080194473267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,4,balanced,0.64083198706309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,4,balanced,0.6413546800613403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,4,balanced,0.6499573389689127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,4,balanced,0.6530453364054362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,4,balanced,0.6537760098775228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,4,balanced,0.05403733253479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,4,balanced,0.6499520142873129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,4,balanced,0.04799999793370565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,4,balanced,0.6622666517893473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,4,balanced,0.05650666852792104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,4,balanced,0.0641599992911021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,4,balanced,0.6662826538085938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,4,balanced,0.04990933338801066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,4,balanced,0.09067199627558391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,4,balanced,0.06958933174610138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,4,balanced,0.6731893221537272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,4,balanced,0.12809600432713827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,4,balanced,0.10126399993896484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,4,balanced,0.18727999925613403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,4,balanced,0.15180266896883646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,4,balanced,0.6892639795939127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,4,balanced,0.25142399470011395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,4,balanced,0.2505066593488057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,4,balanced,0.2563680013020833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,4,balanced,0.3489493529001872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,4,balanced,0.2579200069109599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,4,balanced,0.7071093718210856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,4,balanced,0.35339200496673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,4,power_law_1.01,0.09393919706344604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,4,balanced,0.25646400451660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,4,balanced,0.3564800024032593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,4,balanced,0.26098666588465375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,4,balanced,0.3542453447977702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,4,balanced,0.7416480382283529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,4,power_law_1.01,0.14199680089950562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,4,balanced,0.2648319999376933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,4,balanced,0.35472532113393146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,4,balanced,0.7752319971720377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,4,power_law_1.01,0.0713536024093628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,4,balanced,0.2691199978192647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,4,balanced,0.3619680007298787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,4,power_law_1.01,0.24498560428619384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,4,balanced,0.27011199792226154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,4,balanced,0.3625760078430176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,4,power_law_1.01,0.1309183955192566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,4,power_law_1.01,0.09566720128059387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,4,balanced,0.3694346745808919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,4,balanced,0.2785653273264567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,4,power_law_1.01,0.19808640480041503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,4,balanced,0.827077309290568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,4,balanced,0.3797546625137329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,4,balanced,0.2831520040829976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,4,power_law_1.01,0.2789695978164673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,4,power_law_1.01,0.14545919895172119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,4,power_law_1.01,0.06992639899253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,4,power_law_1.01,0.345196795463562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,4,balanced,0.3794186512629191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,4,balanced,0.29471999406814575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,4,power_law_1.01,0.08833280205726624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,4,power_law_1.01,0.4200704097747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,4,balanced,0.3927146593729655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,4,balanced,0.3102026581764221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,4,power_law_1.01,0.08763520121574402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,4,power_law_1.01,0.13519359827041627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,4,power_law_1.01,0.44324479103088377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,4,balanced,0.8932639757792155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,4,balanced,0.4123733441034953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,4,balanced,0.3254026571909587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,4,power_law_1.01,0.11976959705352783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,4,power_law_1.01,0.1762943983078003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,4,power_law_1.01,0.4484543800354004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,4,power_law_1.01,0.19845759868621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,4,power_law_1.01,0.0887167990207672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,4,balanced,0.4288426637649536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,4,power_law_1.01,0.4957632064819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,4,balanced,0.3625119924545288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,4,power_law_1.01,0.24412159919738768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,4,power_law_1.01,0.11386239528656006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,4,power_law_1.01,0.48903679847717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,4,power_law_1.01,0.4845759868621826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,4,balanced,0.46744000911712646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,4,power_law_1.01,0.26317439079284666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,4,power_law_1.01,0.14191360473632814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,4,balanced,0.39902933438618976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,4,balanced,1.0016319751739502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,4,power_law_1.01,0.526310396194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,4,power_law_1.2,0.09454079866409301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,4,power_law_1.01,0.2598207950592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,4,power_law_1.01,0.16312320232391359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,4,power_law_1.01,0.5288000106811523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,4,balanced,0.5000319878260294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,4,power_law_1.01,0.18899199962615967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,4,power_law_1.01,0.26855039596557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,4,power_law_1.2,0.1419327974319458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,4,power_law_1.01,0.5672448158264161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,4,power_law_1.01,0.18954880237579347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,4,balanced,0.47416532039642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,4,power_law_1.01,0.2955199956893921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,4,power_law_1.2,0.12881280183792115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,4,power_law_1.01,0.5897727966308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,4,power_law_1.01,0.198854398727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,4,balanced,0.5742986599604288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,4,balanced,1.1779786745707195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,4,balanced,0.5317866802215576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,4,power_law_1.2,0.13308160305023192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,4,power_law_1.01,0.6192319869995118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,4,power_law_1.01,0.21261439323425294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,4,power_law_1.01,0.28314239978790284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,4,power_law_1.2,0.2014591932296753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,4,power_law_1.01,0.682367992401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,4,power_law_1.01,0.21143679618835448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,4,power_law_1.01,0.31439359188079835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,4,power_law_1.2,0.2667520046234131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,4,power_law_1.01,0.7725247859954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,4,power_law_1.01,0.22450559139251708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,4,balanced,0.6594133377075195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,4,power_law_1.01,0.32457599639892576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,4,power_law_1.2,0.30167040824890134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,4,power_law_1.01,0.797376012802124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,4,power_law_1.01,0.22906880378723143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,4,power_law_1.2,0.424729585647583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,4,power_law_1.01,0.31189761161804197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,4,power_law_1.01,0.8969856262207031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,4,balanced,0.6526986757914225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,4,power_law_1.01,0.2459712028503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,4,power_law_1.2,0.4350399971008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,4,power_law_1.01,0.34184319972991944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,4,power_law_1.01,1.028115177154541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,4,power_law_1.01,0.2549439907073975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,4,power_law_1.2,0.4406720161437988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,4,power_law_1.01,0.3803391933441162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,4,power_law_1.01,1.2852928161621093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,4,power_law_1.2,0.4547904014587402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,4,power_law_1.01,0.26268160343170166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,4,balanced,0.8067946434020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,4,power_law_1.01,0.3966207981109619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,4,balanced,1.38974396387736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,4,power_law_1.2,0.46506237983703613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,4,power_law_1.01,1.3749823570251465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,4,power_law_1.01,0.3012480020523071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,4,power_law_1.01,0.4688447952270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,4,power_law_1.2,0.507257604598999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,4,power_law_1.01,1.7663808822631837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,4,power_law_1.01,0.3265984058380127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,4,power_law_1.01,0.5222015857696534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,4,power_law_1.2,0.501587200164795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,4,balanced,0.8590559959411621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,4,power_law_1.01,0.36521599292755125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,4,power_law_1.01,2.375040054321289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,4,power_law_1.01,0.6446911811828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,4,power_law_1.2,0.5207359790802002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,4,power_law_1.01,0.4246528148651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,4,power_law_1.01,0.7287744045257568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,4,power_law_1.2,0.5382207870483399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,4,power_law_1.01,2.713849639892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,4,power_law_1.01,0.5363647937774658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,4,balanced,0.9931786855061849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,4,power_law_1.2,0.6058815956115723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,4,power_law_1.01,0.9922623634338379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,4,power_law_1.01,0.5912255764007568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,4,power_law_1.01,3.8806270599365233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,4,power_law_1.2,0.6404352188110352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,4,power_law_1.01,1.1097599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,4,power_law_1.01,0.7898367881774903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,4,power_law_1.2,0.6816576004028321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,4,power_law_1.01,1.5455360412597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,4,power_law_1.2,0.8080320358276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,4,power_law_1.01,0.9820799827575684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,4,power_law_1.01,7.4687553405761715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,4,power_law_1.2,0.8410112380981445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,4,power_law_1.01,1.9068159103393554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,4,power_law_1.01,1.3230655670166016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,4,balanced,1.1178826491038005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,4,power_law_1.2,0.9600000381469727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,4,balanced,1.842565377553304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,4,power_law_1.01,2.3472320556640627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,4,power_law_1.01,1.6367040634155274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,4,power_law_1.2,1.119820785522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,4,power_law_1.01,2.0462272644042967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,4,power_law_1.2,1.4005824089050294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,4,power_law_1.01,3.5876033782958983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,4,balanced,1.3145493666330974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,4,power_law_1.2,1.461068820953369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,4,power_law_1.01,3.0545984268188477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,4,power_law_1.2,1.9720256805419922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,4,power_law_1.01,6.781830596923828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,4,power_law_1.2,2.665260887145996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,4,power_law_1.01,5.85428466796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,4,power_law_1.2,3.207340621948242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,4,power_law_1.2,4.369203186035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,4,balanced,1.4746507008870442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,4,power_law_1.2,0.06937599778175355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,4,power_law_1.2,8.062604522705078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,4,balanced,2.1514293352762857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,4,balanced,1.682794729868571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,4,power_law_1.2,0.09457280039787293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,4,power_law_1.2,0.09180160164833069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,4,power_law_1.2,0.0906112015247345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,4,power_law_1.2,0.12218879461288452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,4,power_law_1.2,0.15816960334777833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,4,power_law_1.2,0.17825280427932738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,4,power_law_1.2,0.23523199558258057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,4,power_law_1.2,0.24620800018310546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,4,balanced,1.83515199025472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,4,power_law_1.2,0.26124160289764403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,4,power_law_1.2,0.2708031892776489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,4,balanced,2.1084960301717124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,4,power_law_1.2,0.2754175901412964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,4,power_law_1.2,0.2830336093902588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,4,power_law_1.2,0.31315200328826903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,4,power_law_1.2,0.32469758987426756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,4,power_law_1.2,0.3339967966079712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,4,power_law_1.2,0.35646719932556153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,4,power_law_1.2,0.38594560623168944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,4,balanced,3.168346722920736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,4,power_law_1.2,0.4188096046447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,4,power_law_1.2,0.47405438423156737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,4,power_law_1.2,0.523417615890503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,4,power_law_1.2,0.6913536071777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,4,power_law_1.2,0.8335616111755371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,4,power_law_1.2,1.0532352447509765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,4,balanced,2.793018658955892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,4,power_law_1.2,1.1916735649108887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,4,power_law_1.2,1.6614463806152344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,4,balanced,3.109978675842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,4,power_law_1.2,2.0419456481933596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,4,power_law_1.2,2.50447998046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,4,power_law_1.2,3.633881759643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,4,power_law_1.2,7.541241455078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,4,balanced,6.052053451538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,4,balanced,5.303567886352539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,4,balanced,6.034645080566406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,2,balanced,0.05403733253479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,4,power_law_1.2,0.06947199702262878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,2,balanced,0.06851199766000111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,2,balanced,0.09302399555842082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,2,balanced,0.14864533146222433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,4,power_law_1.2,0.08756480216979981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,2,balanced,0.24770132700602213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,2,balanced,0.44516801834106445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,4,power_law_1.2,0.09672319889068604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,2,balanced,0.6428800026575724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,2,balanced,0.6412320137023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,4,power_law_1.2,0.08866559863090515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,2,balanced,0.6451573371887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,4,power_law_1.2,0.10302720069885254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,2,balanced,0.6444533268610636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,4,power_law_1.2,0.1267840027809143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,2,balanced,0.6520266532897949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,4,power_law_1.2,0.14820480346679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,2,balanced,0.6582560141881307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,2,balanced,0.6611893177032471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,4,power_law_1.2,0.18517760038375855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,2,balanced,0.05003733436266581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,2,balanced,0.07051200171311696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,2,balanced,0.6671199798583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,4,power_law_1.2,0.1896064043045044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,2,balanced,0.6793013413747152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,2,balanced,0.053904001911481224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,4,power_law_1.2,0.18655359745025635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,2,balanced,0.07448533177375793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,2,balanced,0.6803466478983561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,2,balanced,0.07251733541488647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,4,power_law_1.2,0.20463359355926514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,2,balanced,0.10430399576822917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,2,balanced,0.0832426647345225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,2,balanced,0.7001492977142334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,2,balanced,0.1595146656036377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,4,power_law_1.2,0.21230719089508057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,2,balanced,0.11125333110491435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,2,balanced,0.2582079966862996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,4,power_law_1.2,0.22108159065246583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,2,balanced,0.1585653324921926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,2,balanced,0.7238933245340983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,2,balanced,0.363317330678304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,2,balanced,0.2395253380139669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,4,power_law_1.2,0.22616961002349853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,2,balanced,0.3622026840845744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,2,balanced,0.3223680059115092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,2,balanced,0.742965300877889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,4,power_law_1.2,0.23720960617065429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,2,balanced,0.3680533170700073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,2,balanced,0.32659733295440674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,4,power_law_1.2,0.24450559616088868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,2,balanced,0.371237317721049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,2,balanced,0.33107733726501465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,2,balanced,0.3720533450444539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,4,power_law_1.2,0.268556809425354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,2,balanced,0.3338079849878947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,2,balanced,0.7934026718139648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,2,balanced,0.388213316599528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,2,balanced,0.3290666739145915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,4,power_law_1.2,0.29697279930114745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,2,balanced,0.3851840098698934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,4,power_law_1.2,0.3240511894226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,2,balanced,0.3295946717262268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,2,balanced,0.3946880102157593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,2,balanced,0.8409120241800944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,4,power_law_1.2,0.3861504077911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,2,balanced,0.33164799213409424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,2,balanced,0.4018186728159587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,4,power_law_1.2,0.43526401519775393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,2,balanced,0.33816532293955487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,2,balanced,0.40698134899139404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,4,power_law_1.2,0.5531648159027099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,2,balanced,0.3466720183690389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,2,balanced,0.4222613175710042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,2,balanced,0.9375147024790446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,4,power_law_1.2,0.6268352031707763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,2,balanced,0.3495200077692668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,2,balanced,0.44646934668223065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,4,power_law_1.2,0.8267583847045898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,2,balanced,0.3688746690750122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,2,balanced,0.4721599817276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,4,power_law_1.2,1.0060288429260253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,2,balanced,0.3909279902776082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,2,balanced,1.0133013725280762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,4,power_law_1.2,1.4561920166015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,2,balanced,0.5260853370030721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,2,balanced,0.4147946834564209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,4,power_law_1.2,1.8017152786254882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,2,balanced,0.4644266764322917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,2,balanced,0.5738879839579264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,4,power_law_1.2,2.1763776779174804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,2,balanced,1.195530652999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,2,balanced,0.5046399831771851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,4,power_law_1.2,3.336159896850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,2,balanced,0.6895306905110677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,4,power_law_1.2,6.514028930664063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,2,balanced,0.6049386660257975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,2,balanced,0.819642702738444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,2,balanced,1.4748694101969402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,2,balanced,0.7178986867268881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,2,balanced,1.0281813144683838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,2,power_law_1.01,0.06672000288963317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,2,balanced,0.9237706661224365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,2,power_law_1.01,0.09159039855003356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,2,power_law_1.01,0.14315520524978637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,2,power_law_1.01,0.10871679782867431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,2,balanced,1.9020105997721355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,2,power_law_1.01,0.1882367968559265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,2,balanced,1.313472032546997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,2,power_law_1.01,0.2745471954345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,2,power_law_1.01,0.3285311937332153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,2,balanced,1.224837303161621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,2,power_law_1.01,0.42629117965698243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,2,power_law_1.01,0.4318079948425293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,2,power_law_1.01,0.43953280448913573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,2,power_law_1.01,0.47104640007019044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,2,power_law_1.01,0.49426560401916503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,2,power_law_1.01,0.5134335994720459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,2,power_law_1.01,0.5263679981231689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,2,balanced,1.7341119448343914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,2,power_law_1.01,0.553004789352417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,2,balanced,2.4353973070780435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,2,power_law_1.01,0.5534912109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,2,balanced,1.6352532704671223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,2,power_law_1.01,0.6044032096862793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,2,power_law_1.01,0.6381120204925537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,2,power_law_1.01,0.6893631935119628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,2,power_law_1.01,0.8090496063232422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,2,power_law_1.01,0.8734399795532226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,2,power_law_1.01,1.0926272392272949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,2,power_law_1.01,0.04928640127182007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,2,power_law_1.01,1.1509183883666991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,2,power_law_1.01,1.4133312225341796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,2,power_law_1.01,0.06921600103378296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,2,balanced,2.3265280723571777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,2,power_law_1.01,1.6102592468261718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,2,power_law_1.01,0.0974399983882904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,2,power_law_1.01,2.1449472427368166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,2,power_law_1.01,0.08942080140113831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,2,balanced,2.1875839233398438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,2,power_law_1.01,0.1238592028617859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,2,power_law_1.01,2.805446434020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,2,balanced,2.8909120559692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,2,power_law_1.01,0.1679487943649292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,2,power_law_1.01,3.265971374511719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,2,power_law_1.01,0.19981440305709838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,2,power_law_1.01,0.23771519660949708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,2,power_law_1.01,4.821113586425781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,2,power_law_1.01,0.06631039977073669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,2,power_law_1.01,0.25340158939361573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,2,power_law_1.01,0.266431999206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,2,power_law_1.01,0.07517439723014832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,2,power_law_1.2,0.0679744005203247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,2,power_law_1.01,0.2857215881347656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,2,power_law_1.01,0.10280959606170655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,2,power_law_1.01,8.827027130126954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,2,power_law_1.01,0.1020799994468689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,2,power_law_1.01,0.2973439931869507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,2,power_law_1.2,0.09358720183372497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,2,power_law_1.01,0.13219200372695922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,2,power_law_1.01,0.3106688022613525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,2,power_law_1.2,0.09966080188751221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,2,balanced,2.8292907079060874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,2,power_law_1.01,0.16556160449981688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,2,power_law_1.01,0.31378560066223143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,2,power_law_1.2,0.11704319715499878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,2,balanced,2.736015955607096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,2,power_law_1.01,0.18929920196533204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,2,power_law_1.01,0.3293888092041016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,2,power_law_1.2,0.17637120485305785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,2,power_law_1.01,0.3415744066238403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,2,power_law_1.01,0.23072640895843505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,2,power_law_1.2,0.23976318836212157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,2,power_law_1.01,0.2377919912338257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,2,power_law_1.01,0.36446080207824705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,2,power_law_1.2,0.2926079988479614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,2,power_law_1.01,0.2529727935791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,2,power_law_1.2,0.4008959770202637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,2,power_law_1.01,0.4018303871154785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,2,power_law_1.2,0.4166207790374756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,2,power_law_1.01,0.25436160564422605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,2,power_law_1.01,0.44551677703857423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,2,power_law_1.2,0.4465023994445801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,2,power_law_1.2,0.47267842292785645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,2,power_law_1.01,0.25922560691833496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,2,balanced,4.376570701599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,2,power_law_1.01,0.5126848220825195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,2,power_law_1.2,0.47190399169921876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,2,power_law_1.01,0.27082879543304444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,2,power_law_1.01,0.5972799777984619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,2,power_law_1.2,0.49313921928405763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,2,power_law_1.01,0.27844479084014895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,2,power_law_1.01,0.7310783863067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,2,power_law_1.01,0.2991296052932739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,2,power_law_1.2,0.5150400161743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,2,power_law_1.01,0.9251520156860351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,2,power_law_1.01,0.3052864074707031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,2,power_law_1.2,0.539404821395874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,2,power_law_1.01,1.2451135635375976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,2,power_law_1.01,0.31818881034851076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,2,power_law_1.2,0.5671167850494385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,2,power_law_1.01,1.4231167793273927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,2,power_law_1.01,0.35392639636993406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,2,power_law_1.2,0.5960256099700928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,2,power_law_1.01,0.37902719974517823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,2,power_law_1.2,0.6549568176269531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,2,power_law_1.01,1.8832639694213866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,2,power_law_1.01,0.4489727973937988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,2,power_law_1.2,0.7188479900360107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,2,balanced,4.2737226486206055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,2,power_law_1.01,2.5795007705688477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,2,power_law_1.01,0.5251520156860352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,2,power_law_1.2,0.8612735748291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,2,power_law_1.01,0.6499711990356445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,2,power_law_1.2,0.9488256454467774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,2,balanced,4.099664052327474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,2,power_law_1.01,3.1589311599731444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,2,power_law_1.01,0.775980806350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,2,power_law_1.2,1.1743743896484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,2,power_law_1.01,4.4299072265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,2,power_law_1.01,0.9952128410339356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,2,power_law_1.2,1.2223551750183106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,2,power_law_1.01,1.3553919792175293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,2,power_law_1.2,1.4968128204345703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,2,power_law_1.01,1.7954303741455078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,2,power_law_1.2,1.6905664443969726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,2,power_law_1.01,8.485228729248046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,2,power_law_1.01,2.223616027832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,2,power_law_1.2,2.3442304611206053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,2,power_law_1.2,2.8750207901000975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,2,power_law_1.01,2.8488704681396486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,2,power_law_1.2,3.612192153930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,2,power_law_1.01,4.281478500366211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,2,power_law_1.2,4.924217605590821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,2,power_law_1.01,8.075526428222656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,2,power_law_1.2,9.203199768066407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,2,balanced,8.257338841756185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,2,balanced,8.222245534261068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,2,balanced,7.829423904418945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,2,power_law_1.2,0.051667201519012454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,2,power_law_1.2,0.07232000231742859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,2,power_law_1.2,0.08062719702720642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,2,power_law_1.2,0.08823680281639099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,2,power_law_1.2,0.12197760343551636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,2,power_law_1.2,0.1583359956741333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,2,power_law_1.2,0.18167680501937866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,2,power_law_1.2,0.24495999813079833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,2,power_law_1.2,0.25022718906402586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,2,power_law_1.2,0.25705599784851074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,2,power_law_1.2,0.2689023971557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,2,power_law_1.2,0.2930943965911865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,2,power_law_1.2,0.31340799331665037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,2,power_law_1.2,0.3039680004119873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,2,power_law_1.2,0.32877440452575685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,2,power_law_1.2,0.3516416072845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,2,power_law_1.2,0.3704447984695435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,2,power_law_1.2,0.413804817199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,2,power_law_1.2,0.44458880424499514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,2,power_law_1.2,0.5255104064941406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,2,power_law_1.2,0.6088064193725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,2,power_law_1.2,0.7627776145935059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,2,power_law_1.2,0.9492416381835938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,2,power_law_1.2,1.2504511833190919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,2,power_law_1.2,1.4615103721618652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,2,power_law_1.2,1.9915647506713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,2,power_law_1.2,2.648908805847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,2,power_law_1.2,3.3106945037841795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,2,power_law_1.2,4.730195236206055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,2,power_law_1.2,0.06544640064239501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,2,power_law_1.2,8.961888122558594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,2,power_law_1.2,0.07661439776420594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,2,power_law_1.2,0.0874559998512268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,2,power_law_1.2,0.09560319781303406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,2,power_law_1.2,0.12448639869689941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,2,power_law_1.2,0.15599360466003417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,2,power_law_1.2,0.17811199426651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,2,power_law_1.2,0.22716159820556642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,2,power_law_1.2,0.23169920444488526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,2,power_law_1.2,0.2410111904144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,2,power_law_1.2,0.2542272090911865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,2,power_law_1.2,0.26143999099731446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,2,power_law_1.2,0.2678911924362183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,2,power_law_1.2,0.27956480979919435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,2,power_law_1.2,0.305132794380188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,2,power_law_1.2,0.2987456083297729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,2,power_law_1.2,0.3271424055099487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,2,power_law_1.2,0.3532288074493408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,2,power_law_1.2,0.3982079982757568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,2,power_law_1.2,0.46842241287231445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,1,balanced,0.06102933486302694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,2,power_law_1.2,0.5340991973876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,2,power_law_1.2,0.6950143814086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,1,balanced,0.07321066657702129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,1,balanced,0.09728533029556274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,2,power_law_1.2,0.7593088150024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,1,balanced,0.15550933281580606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,1,balanced,0.2544533411661784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,2,power_law_1.2,1.0223872184753418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,1,balanced,0.4517013231913249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,1,balanced,0.6510826746622721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,2,power_law_1.2,1.3453184127807618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,1,balanced,0.6661333243052164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,1,balanced,0.6579786539077759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,2,power_law_1.2,1.8648832321166993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,1,balanced,0.6768960158030192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,1,balanced,0.6817599932352701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,2,power_law_1.2,2.33123836517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,1,balanced,0.6902560393015543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,1,balanced,0.6936586697896322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,2,power_law_1.2,2.8554752349853514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,1,balanced,0.707914670308431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,1,balanced,0.7173279921213785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,2,power_law_1.2,4.36517105102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,1,balanced,0.056405335664749146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,1,balanced,0.7232000033060709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,1,balanced,0.740501324335734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,1,balanced,0.061066667238871254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,1,balanced,0.09386133154233296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,2,power_law_1.2,8.474034881591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,1,balanced,0.07930666704972585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,1,balanced,0.7860373655954996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,1,balanced,0.09924266735712688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,1,balanced,0.1170186698436737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,1,balanced,0.18074132998784384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,1,balanced,0.11342933773994446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,1,balanced,0.8232373396555582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,1,balanced,0.15054399768511453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,1,balanced,0.2905493378639221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,1,balanced,0.22254933913548788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,1,balanced,0.3994506597518921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,1,balanced,0.3343786795934041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,1,balanced,0.3994079828262329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,1,balanced,0.8948480288187662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,1,balanced,0.44859198729197186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,1,balanced,0.40223467350006104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,1,balanced,0.45044267177581787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,1,balanced,0.4026506741841634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,1,balanced,0.45232534408569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,1,balanced,0.9711093107859293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,1,balanced,0.40854934851328534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,1,balanced,0.454095999399821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,1,balanced,0.4607946475346883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,1,balanced,0.41654932498931885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,1,balanced,0.4647359848022461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,1,balanced,0.42026666800181073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,1,power_law_1.01,0.05375999808311462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,1,balanced,0.4657226800918579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,1,balanced,0.42863468329111737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,1,balanced,1.121519962946574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,1,power_law_1.01,0.07151359915733338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,1,balanced,0.47501333554585773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,1,balanced,0.43767468134562176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,1,power_law_1.01,0.09836800098419189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,1,balanced,0.480837345123291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,1,balanced,0.44830934206644696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,1,power_law_1.01,0.12357759475708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,1,balanced,0.49158398310343426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,1,balanced,0.4700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,1,balanced,1.327951987584432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,1,power_law_1.01,0.19714560508728027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,1,balanced,0.5100586811701456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,1,power_law_1.01,0.2749311923980713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,1,balanced,0.5272213220596313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,1,balanced,0.5507093270619711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,1,power_law_1.01,0.32792320251464846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,1,balanced,0.5813386837641398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,1,power_law_1.01,0.43289599418640134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,1,balanced,0.5876319805781046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,1,power_law_1.01,0.458892822265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,1,balanced,0.668725331624349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,1,balanced,1.675615946451823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,1,balanced,0.6704160372416178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,1,power_law_1.01,0.4662015914916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,1,power_law_1.01,0.49654397964477537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,1,balanced,0.7461653550465902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,1,power_law_1.01,0.5208896160125732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,1,balanced,0.747429370880127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,1,power_law_1.01,0.5364223957061768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,1,power_law_1.01,0.5557504177093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,1,balanced,0.9132853349049886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,1,balanced,0.9146827061971029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,1,power_law_1.01,0.577120018005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,1,power_law_1.01,0.6055295944213868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,1,balanced,2.1248265902201333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,1,power_law_1.01,0.6296703815460205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,1,power_law_1.01,0.05422080159187317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,1,balanced,1.131381352742513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,1,power_law_1.01,0.08952959775924682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,1,power_law_1.01,0.7006464004516602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,1,balanced,1.0869119962056477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,1,power_law_1.01,0.059552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,1,power_law_1.01,0.7582015991210938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,1,power_law_1.01,0.09271680116653443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,1,power_law_1.01,0.8849727630615234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,1,power_law_1.01,0.07754240036010743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,1,power_law_1.01,0.10785280466079712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,1,power_law_1.01,0.13426560163497925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,1,power_law_1.01,0.9588800430297851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,1,power_law_1.01,0.09797120094299316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,1,balanced,1.4644479751586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,1,balanced,1.4525173505147297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,1,power_law_1.01,0.18526079654693603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,1,power_law_1.01,1.1787455558776856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,1,power_law_1.01,0.14807679653167724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,1,power_law_1.01,0.23475840091705322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,1,balanced,2.791269302368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,1,power_law_1.01,0.19937280416488648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,1,power_law_1.01,1.463424015045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,1,power_law_1.01,0.26657919883728026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,1,power_law_1.01,0.22578558921813965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,1,power_law_1.01,1.9122880935668944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,1,power_law_1.01,0.3173376083374023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,1,power_law_1.01,0.2745215892791748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,1,power_law_1.01,0.33484160900115967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,1,power_law_1.01,2.176198387145996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,1,power_law_1.01,0.28815999031066897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,1,balanced,1.943285306294759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,1,balanced,1.9296053250630696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,1,power_law_1.01,0.34444799423217776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,1,power_law_1.01,0.2939647912979126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,1,power_law_1.01,2.900524711608887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,1,power_law_1.01,0.35571839809417727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,1,power_law_1.01,0.31102719306945803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,1,power_law_1.01,0.37706239223480226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,1,power_law_1.01,0.32999680042266843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,1,power_law_1.01,3.732223892211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,1,power_law_1.01,0.34364800453186034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,1,power_law_1.01,0.38220160007476806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,1,power_law_1.01,4.4993408203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,1,power_law_1.01,0.3650496006011963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,1,power_law_1.01,0.3941888093948364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,1,balanced,3.6761706670125327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,1,power_law_1.01,0.3930432081222534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,1,power_law_1.01,0.40993919372558596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,1,balanced,2.6311680475870767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,1,power_law_1.01,6.570751953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,1,power_law_1.01,0.4076096057891846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,1,power_law_1.01,0.421235179901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,1,balanced,2.6428319613138833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,1,power_law_1.01,0.44606719017028806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,1,power_law_1.01,0.44343042373657227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,1,power_law_1.01,0.5403135776519775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,1,power_law_1.01,0.4990399837493896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,1,power_law_1.01,12.689075469970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,1,power_law_1.01,0.5509119987487793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.01,0.5429696083068848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,1,power_law_1.01,0.6677887916564942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.01,0.640723180770874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,1,power_law_1.01,0.7517888069152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.01,0.7293312072753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,1,power_law_1.01,0.9608256340026855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.01,0.9293055534362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,1,power_law_1.01,1.1737279891967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.01,1.1221823692321777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,1,power_law_1.01,1.5695487976074218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.01,1.5191167831420898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,1,balanced,3.5187412897745767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,1,balanced,3.5754451751708984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,1,power_law_1.01,2.0198976516723635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,1,power_law_1.01,2.0304256439208985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,1,balanced,4.508437474568685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,1,power_law_1.01,2.760665512084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,1,power_law_1.01,2.7840639114379884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,1,power_law_1.01,3.557004928588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,1,power_law_1.01,3.4580543518066404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,1,power_law_1.01,4.208832168579102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,1,power_law_1.01,4.308832168579102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,1,power_law_1.01,6.616294097900391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,1,power_law_1.01,6.618125152587891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,1,power_law_1.01,12.803910827636718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,1,power_law_1.01,12.77117462158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,1,balanced,4.364432017008464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,1,balanced,4.467413266499837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,1,power_law_1.2,0.053472000360488894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,1,power_law_1.2,0.0717311978340149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,1,power_law_1.2,0.08473600149154663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,1,power_law_1.2,0.12373759746551513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,1,power_law_1.2,0.17768959999084472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,1,power_law_1.2,0.23885440826416016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,1,balanced,6.527546564737956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,1,power_law_1.2,0.291756796836853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,1,power_law_1.2,0.4133440017700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,1,power_law_1.2,0.43483519554138184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,1,power_law_1.2,0.46083841323852537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,1,power_law_1.2,0.479750394821167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,1,power_law_1.2,0.4915008068084717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,1,power_law_1.2,0.5171264171600342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,1,power_law_1.2,0.5487808227539063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,1,power_law_1.2,0.5788288116455078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,1,power_law_1.2,0.5834943771362304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,1,balanced,6.540042877197266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,1,power_law_1.2,0.6341119766235351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,1,power_law_1.2,0.7087615966796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,1,balanced,6.77186648050944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,1,power_law_1.2,0.7824512004852295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,1,power_law_1.2,0.9171135902404786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,1,power_law_1.2,1.0044095993041993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,1,power_law_1.2,1.211622428894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,1,power_law_1.2,1.5262847900390626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,1,power_law_1.2,1.9568960189819335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,1,power_law_1.2,2.2450368881225584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,1,power_law_1.2,2.9377599716186524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,1,power_law_1.2,3.806719970703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,1,power_law_1.2,4.5770751953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,1,power_law_1.2,6.70165786743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,1,power_law_1.2,12.812261962890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,1,power_law_1.2,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,1,power_law_1.2,0.0575872004032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,1,power_law_1.2,0.06759039759635925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,1,balanced,12.5206298828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,1,power_law_1.2,0.09600639939308167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,1,power_law_1.2,0.13123199939727784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,1,power_law_1.2,0.17879040241241456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,1,power_law_1.2,0.20327041149139405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,1,power_law_1.2,0.2602047920227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,1,power_law_1.2,0.2775615930557251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,1,power_law_1.2,0.28406400680541993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,1,balanced,12.645029703776041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,1,power_law_1.2,0.3010432004928589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,1,power_law_1.2,0.3248512029647827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,1,power_law_1.2,0.33851521015167235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,1,balanced,12.807963053385416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,1,power_law_1.2,0.36189439296722414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,1,power_law_1.2,0.3924736022949219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,1,power_law_1.2,0.41843199729919434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,1,power_law_1.2,0.4554880142211914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,1,power_law_1.2,0.5535039901733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,1,power_law_1.2,0.5603263854980469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,1,power_law_1.2,0.6860735893249512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,1,power_law_1.2,0.7679615974426269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,1,power_law_1.2,0.9772607803344726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,1,power_law_1.2,1.2105792045593262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,1,power_law_1.2,1.61712646484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,1,power_law_1.2,2.0665855407714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,1,power_law_1.2,2.799711990356445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,1,power_law_1.2,3.565760040283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,1,power_law_1.2,4.324870300292969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,1,power_law_1.2,6.683161926269531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,1,power_law_1.2,0.08570240139961242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,1,power_law_1.2,0.09200639724731445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,1,power_law_1.2,12.825248718261719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,1,power_law_1.2,0.10274560451507568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,1,power_law_1.2,0.1309823989868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,1,power_law_1.2,0.17139840126037598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,1,power_law_1.2,0.21566081047058105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,1,power_law_1.2,0.24040958881378174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,1,power_law_1.2,0.3080960035324097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,1,power_law_1.2,0.3210560083389282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,1,power_law_1.2,0.3353087902069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,1,power_law_1.2,0.3495232105255127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,1,power_law_1.2,0.36557440757751464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,1,power_law_1.2,0.37093119621276854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,1,power_law_1.2,0.3849087953567505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,1,power_law_1.2,0.40446081161499026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,1,power_law_1.2,0.418009614944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,1,power_law_1.2,0.43968639373779295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,1,power_law_1.2,0.49755520820617677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.2,0.5522175788879394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.2,0.6538879871368408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,16,balanced,0.033573334415753685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,16,balanced,0.0334346666932106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.2,0.7454783916473389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,16,balanced,0.033546666304270424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,16,balanced,0.03279466678698858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.2,0.945644760131836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,16,balanced,0.033957332372665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,16,balanced,0.056159997979799904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,16,balanced,0.033600000043710075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,16,balanced,0.07754666606585185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.2,1.1422143936157227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,16,balanced,0.03749866783618927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,16,balanced,0.07841066519419353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,16,balanced,0.057818666100502014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,16,balanced,0.07941866914431255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,16,balanced,0.057909334699312844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.2,1.5229503631591796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,16,balanced,0.0804746647675832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,16,balanced,0.059861332178115845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,16,balanced,0.07979199786980946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,16,balanced,0.05819199979305267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,1,power_law_1.2,2.0540224075317384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,16,balanced,0.07966400186220805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,16,balanced,0.05983466903368632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,16,balanced,0.059845333298047386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,16,balanced,0.08098666866620381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,1,power_law_1.2,2.8100671768188477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,16,balanced,0.06033066908518473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,16,balanced,0.08052266637484233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,16,balanced,0.0825386643409729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,16,balanced,0.06228800117969513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,1,power_law_1.2,3.4325439453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,16,balanced,0.0612960010766983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,16,balanced,0.08213866750399272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,16,balanced,0.06408533453941345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,16,balanced,0.08518399794896443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,1,power_law_1.2,4.186188888549805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,16,balanced,0.06955733398596446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,16,balanced,0.08616000413894653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,16,balanced,0.06876799960931142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,16,balanced,0.08751466870307922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,1,power_law_1.2,6.6112060546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,16,balanced,0.07122133175532024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,16,balanced,0.09005332986513774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,16,balanced,0.0762613316377004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,16,balanced,0.09268266956011455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,16,balanced,0.08147733410199483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,1,power_law_1.2,12.850271606445313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,16,balanced,0.10569066802660625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,16,balanced,0.08925333619117737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,16,balanced,0.10791466633478801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,16,balanced,0.10090133547782898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,16,balanced,0.12274666627248128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,16,balanced,0.1237546702226003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,16,balanced,0.13661332925160727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,16,balanced,0.14730133612950644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,16,balanced,0.17021334171295166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,16,balanced,0.18430399894714355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,16,balanced,0.20641599098841348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,16,balanced,0.2290133237838745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,16,balanced,0.03379733363787333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,16,balanced,0.033813332517941795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,16,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,16,balanced,0.0396373321612676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,16,balanced,0.023797333240509033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,16,balanced,0.045066664616266884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,16,balanced,0.04632533093293508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,16,balanced,0.0480373352766037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,16,balanced,0.02310933421055476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,16,balanced,0.047370667258898415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,16,balanced,0.023904000719388325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,16,balanced,0.04737600187460581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,16,balanced,0.25728533665339154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,16,balanced,0.02332266668478648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,16,balanced,0.30880532662073773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,16,balanced,0.029338667790095013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,16,balanced,0.0481333335240682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,16,balanced,0.033071999748547874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,16,balanced,0.04902400076389313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,16,balanced,0.04502933224042257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,16,balanced,0.0498933345079422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,16,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,16,balanced,0.0496373325586319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,16,balanced,0.06253866851329803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,16,power_law_1.01,0.07203199863433837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,16,balanced,0.050197333097457886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,16,balanced,0.06314133107662201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,16,power_law_1.01,0.04872959852218628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,16,balanced,0.31189332405726117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,16,balanced,0.38952000935872394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,16,balanced,0.05566399792830149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,16,power_law_1.01,0.05489280223846436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,16,power_law_1.01,0.04755840003490448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,16,balanced,0.06306666632493337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,16,balanced,0.05571199953556061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,16,power_law_1.01,0.03537279963493347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,16,balanced,0.06276800235112508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,16,power_law_1.01,0.058412802219390866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,16,balanced,0.061343997716903687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,16,power_law_1.01,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,16,balanced,0.05798399945100149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,16,power_law_1.01,0.05851519703865051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,16,balanced,0.05769066512584686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,16,power_law_1.01,0.07594879865646362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,16,power_law_1.01,0.04002560079097748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,16,balanced,0.07643199960390727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,16,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,16,power_law_1.01,0.07257599830627441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,16,power_law_1.01,0.04574080109596253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,16,balanced,0.07458666463692983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,16,power_law_1.01,0.07778559923171997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,16,balanced,0.08062933385372162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,16,power_law_1.01,0.052537602186203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,16,balanced,0.06962666908899943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,16,power_law_1.01,0.07580159902572632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,16,power_law_1.01,0.05597440004348755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,16,power_law_1.01,0.07514240145683289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,16,balanced,0.36531198024749756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,16,balanced,0.47092266877492267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,16,balanced,0.07950399816036224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,16,power_law_1.01,0.056992000341415404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,16,balanced,0.08261333405971527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,16,power_law_1.01,0.07738239765167236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,16,power_law_1.01,0.058457601070404056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,16,power_law_1.01,0.07935360074043274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,16,balanced,0.091839998960495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,16,power_law_1.01,0.055776000022888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,16,power_law_1.01,0.08075519800186157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,16,balanced,0.0867733359336853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,16,power_law_1.01,0.0593280017375946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,16,power_law_1.01,0.08284800052642823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,16,balanced,0.11083199580510457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,16,power_law_1.01,0.06104320287704468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,16,power_law_1.01,0.0886784017086029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,16,power_law_1.01,0.061510401964187625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,16,balanced,0.10309333602587382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,16,balanced,0.13212266564369202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,16,power_law_1.01,0.06444159746170045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,16,power_law_1.01,0.0944320023059845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,16,power_law_1.01,0.04824959933757782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,16,power_law_1.01,0.09181439876556396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,16,power_law_1.01,0.07041919827461243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,16,power_law_1.01,0.10250879526138305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,16,power_law_1.01,0.03715839982032776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,16,balanced,0.16211199760437012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,16,balanced,0.1144586702187856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,16,power_law_1.01,0.07491840124130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,16,power_law_1.01,0.11265280246734619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,16,power_law_1.01,0.035392001271247864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,16,balanced,0.5223573446273804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,16,balanced,0.7126506964365641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,16,power_law_1.01,0.0775168001651764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,16,power_law_1.01,0.14488320350646972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,16,balanced,0.19961067040761313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,16,power_law_1.01,0.03837440013885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,16,balanced,0.14446399609247842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,16,power_law_1.01,0.08575999736785889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,16,power_law_1.01,0.1503551959991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,16,power_law_1.01,0.040915200114250184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,16,balanced,0.29150400559107464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,16,power_law_1.01,0.09473919868469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,16,power_law_1.01,0.16242560148239135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,16,balanced,0.17321600516637167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,16,power_law_1.01,0.1143231987953186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,16,power_law_1.01,0.20670080184936523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,16,power_law_1.01,0.04524160027503967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,16,power_law_1.01,0.13007999658584596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,16,power_law_1.01,0.2738368034362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,16,power_law_1.01,0.045023998618125914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,16,power_law_1.01,0.15054719448089598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,16,power_law_1.01,0.3498239994049072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,16,power_law_1.01,0.20440320968627929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,16,power_law_1.01,0.4555840015411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,16,power_law_1.01,0.047577598690986635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,16,power_law_1.01,0.2849600076675415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,16,power_law_1.01,0.6656320095062256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,16,power_law_1.01,0.04673919975757599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,16,power_law_1.01,0.34221439361572265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,16,balanced,0.3643466631571452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,16,power_law_1.01,0.04787839949131012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,16,power_law_1.01,0.8743295669555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,16,power_law_1.01,0.047814399003982544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,16,power_law_1.01,0.5113152027130127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,16,power_law_1.01,1.289356803894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,16,power_law_1.01,0.04872959852218628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,16,balanced,0.23189866542816162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,16,power_law_1.01,0.613708782196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,16,power_law_1.01,0.04924159944057464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,16,power_law_1.01,0.05268480181694031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,16,power_law_1.01,2.0087615966796877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,16,power_law_1.01,0.7834559917449951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,16,power_law_1.01,0.05927039980888367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,16,power_law_1.01,0.06338559985160827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,16,power_law_1.01,1.1492863655090333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,16,power_law_1.01,0.06645119786262513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,16,power_law_1.01,0.026374399662017822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,16,power_law_1.01,0.07646080255508422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,16,power_law_1.01,2.364691162109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,16,balanced,0.5283519824345907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.01,0.0975488007068634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,16,power_law_1.01,0.02096640020608902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.01,0.10823040008544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,16,power_law_1.01,0.021196800470352172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.01,0.1054527997970581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,16,power_law_1.01,0.022067199647426605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.01,0.13434879779815673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,16,power_law_1.01,0.024774399399757386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,16,balanced,0.2941333254178365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.01,0.17192959785461426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,16,power_law_1.01,0.03312000036239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,16,power_law_1.01,0.04330880045890808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,16,power_law_1.01,0.04190720021724701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.01,0.22838399410247803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,16,power_law_1.01,0.06240000128746033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.01,0.2882944107055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,16,power_law_1.01,0.06231039762496948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,16,power_law_1.01,0.06264960169792175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.01,0.41443839073181155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,16,balanced,0.9552106857299805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,16,power_law_1.01,0.06907520294189454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.01,0.49458560943603513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,16,balanced,0.6954987049102783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,16,power_law_1.01,0.06212480068206787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,16,power_law_1.01,0.06480000019073487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,16,power_law_1.01,0.6672512054443359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,16,balanced,1.3522027333577473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,16,power_law_1.01,0.06330879926681518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,16,power_law_1.01,0.9364288330078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,16,power_law_1.01,0.06531839966773986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,16,power_law_1.01,0.06373760104179382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,16,power_law_1.01,1.8641792297363282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,16,power_law_1.01,0.06401919722557067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,16,balanced,0.3524266481399536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.01,0.0676800012588501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.01,0.07836800217628478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.01,0.09164159893989562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.01,0.11234560012817382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.01,0.14603519439697266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.01,0.20782079696655273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.01,0.25525119304656985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,16,balanced,0.8707946936289469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.01,0.3679935932159424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.01,0.4763328075408936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,16,power_law_1.01,0.6276544094085693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,16,power_law_1.01,0.9453632354736328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,16,power_law_1.01,1.7689216613769532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,16,balanced,0.5258506536483765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,16,balanced,1.375264008839925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,16,balanced,0.9896693229675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,16,balanced,2.7222185134887695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,16,power_law_1.2,0.07179520130157471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,16,power_law_1.2,0.05497599840164184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,16,power_law_1.2,0.037049600481987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,16,power_law_1.2,0.032915198802948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,16,power_law_1.2,0.040601599216461184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,16,power_law_1.2,0.03402239978313446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,16,power_law_1.2,0.05381119847297668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,16,power_law_1.2,0.03706879913806915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,16,power_law_1.2,0.053311997652053834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,16,power_law_1.2,0.04279040098190308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,16,power_law_1.2,0.07132800221443177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,16,power_law_1.2,0.054604798555374146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,16,power_law_1.2,0.07442560195922851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,16,power_law_1.2,0.05661439895629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,16,power_law_1.2,0.07768319845199585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,16,power_law_1.2,0.05702400207519531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,16,power_law_1.2,0.07594239711761475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,16,power_law_1.2,0.057036799192428586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,16,power_law_1.2,0.0754688024520874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,16,power_law_1.2,0.05103999972343445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,16,power_law_1.2,0.05783680081367493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,16,power_law_1.2,0.07933440208435058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,16,power_law_1.2,0.05985919833183288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,16,power_law_1.2,0.08127999901771546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,16,power_law_1.2,0.03466239869594574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,16,power_law_1.2,0.07980160117149353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,16,power_law_1.2,0.061440002918243405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,16,power_law_1.2,0.03580799996852875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,16,power_law_1.2,0.08508800268173218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,16,power_law_1.2,0.06438400149345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,16,power_law_1.2,0.03760640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,16,power_law_1.2,0.09477120041847228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,16,power_law_1.2,0.04182400107383728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,16,power_law_1.2,0.06462720036506653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,16,power_law_1.2,0.09742079973220825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,16,power_law_1.2,0.02592639923095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,16,power_law_1.2,0.04668160080909729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,16,power_law_1.2,0.09853439927101135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,8,balanced,0.031386665999889374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,16,power_law_1.2,0.07583360075950622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,16,power_law_1.2,0.11208319664001465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,16,power_law_1.2,0.04552319943904877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,16,power_law_1.2,0.0204352006316185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,16,power_law_1.2,0.07464320063591004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,16,power_law_1.2,0.021209600567817687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,8,balanced,0.032255999743938446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,16,power_law_1.2,0.12993919849395752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,16,power_law_1.2,0.04686720073223114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,16,power_law_1.2,0.08149120211601257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,16,power_law_1.2,0.0229312002658844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,16,power_law_1.2,0.16056959629058837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,16,power_law_1.2,0.04751999974250794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,16,power_law_1.2,0.02393600046634674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,8,balanced,0.03356266766786575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,16,power_law_1.2,0.0945855975151062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,16,power_law_1.2,0.16359039545059204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,16,power_law_1.2,0.03256320059299469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,16,power_law_1.2,0.04736000001430511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,16,power_law_1.2,0.10078719854354859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,8,balanced,0.055946667989095054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,16,power_law_1.2,0.04611839950084686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,16,power_law_1.2,0.22301440238952636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,16,power_law_1.2,0.04829440116882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,16,power_law_1.2,0.04127359986305237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,16,power_law_1.2,0.1192896008491516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,8,balanced,0.07845866680145264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,16,power_law_1.2,0.05134720206260681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,16,power_law_1.2,0.23911681175231933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,16,power_law_1.2,0.06339200139045716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,8,balanced,0.07896000146865845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,16,power_law_1.2,0.0625216007232666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,16,power_law_1.2,0.13556480407714844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,16,power_law_1.2,0.052883201837539674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,16,power_law_1.2,0.3247231960296631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,16,power_law_1.2,0.06355839967727661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,8,balanced,0.08051733175913493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,16,power_law_1.2,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,16,power_law_1.2,0.06547840237617493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,16,power_law_1.2,0.3983743906021118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,16,power_law_1.2,0.18647040128707887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,16,power_law_1.2,0.06407039761543273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,16,power_law_1.2,0.06197119951248169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,8,balanced,0.08069866895675659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,16,power_law_1.2,0.6705088138580322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,16,power_law_1.2,0.0587007999420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,16,power_law_1.2,0.2513279914855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,16,power_law_1.2,0.064300799369812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,8,balanced,0.08046933511892955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,16,power_law_1.2,0.06318719983100891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,16,power_law_1.2,0.8946111679077149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,16,power_law_1.2,0.35985279083251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,16,power_law_1.2,0.07146880030632019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,16,power_law_1.2,0.06116480231285095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,8,balanced,0.07923200229803722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,16,power_law_1.2,0.052339202165603636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,8,balanced,0.029792000850041706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,16,power_law_1.2,0.08186240196228027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,16,power_law_1.2,1.2420607566833497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,16,power_law_1.2,0.45290241241455076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,8,balanced,0.08044800162315369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,16,power_law_1.2,0.06295679807662964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,8,balanced,0.03140799949566523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.2,0.10103679895401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,16,power_law_1.2,1.7882240295410157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,8,balanced,0.03141333411137263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.2,0.0686079978942871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,8,balanced,0.08096000055472057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,16,power_law_1.2,0.6622335910797119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.2,0.11314560174942016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,8,balanced,0.03587199995915095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.2,0.08380799889564514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,8,balanced,0.057029331723848976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,8,balanced,0.08278400202592213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,16,power_law_1.2,4.10780143737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.2,0.10100480318069457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,16,power_law_1.2,0.922265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.2,0.11693439483642579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,8,balanced,0.058933332562446594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.2,0.1262719988822937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,8,balanced,0.08264000217119853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.2,0.17255680561065673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,8,balanced,0.061039999127388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,16,power_law_1.2,1.247116756439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.2,0.15171200037002563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,8,balanced,0.06018133461475372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.2,0.19719680547714233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,8,balanced,0.086325337489446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.2,0.214630389213562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,8,balanced,0.060191998879114784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.2,0.297273588180542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,8,balanced,0.06198399762312571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,16,power_law_1.2,1.6927423477172852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.2,0.2886847972869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,8,balanced,0.08898666501045227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,8,balanced,0.06137600044409434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.2,0.35320320129394533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.2,0.3963072061538696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,8,balanced,0.06418133278687795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,8,balanced,0.08993066350618999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.2,0.5835391998291015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.2,0.5084671974182129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,8,balanced,0.06402133405208588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,16,power_law_1.2,3.809983825683594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,8,balanced,0.09300266702969869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,8,balanced,0.06612266600131989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.2,0.6807168006896973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,16,power_law_1.2,0.6984447956085205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,8,balanced,0.07400000095367432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,16,power_law_1.2,0.7969535827636719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,8,balanced,0.09708799918492635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,16,power_law_1.2,1.1073151588439942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,8,balanced,0.07363200187683105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,8,balanced,0.07463466624418895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,16,power_law_1.2,1.4355648040771485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,16,power_law_1.2,2.1074304580688477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,8,balanced,0.11293866237004598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,8,balanced,0.08040533463160197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,8,balanced,0.09646933277448018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,16,power_law_1.2,3.877791976928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,8,balanced,0.12110400199890137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,8,balanced,0.10620266199111938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,8,balanced,0.13871999581654867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,8,balanced,0.11495467027028401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,8,balanced,0.1437279979387919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,8,balanced,0.1539253294467926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,8,balanced,0.17441066106160483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,8,balanced,0.19545066356658936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,8,balanced,0.035973332822322845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,8,balanced,0.0395413339138031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,8,balanced,0.23340266942977905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,8,balanced,0.04540266593297323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,8,balanced,0.0545066644748052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,8,balanced,0.07452266911665599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,8,balanced,0.0701279987891515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,8,balanced,0.07441066702206929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,8,balanced,0.24299200375874838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,8,balanced,0.07138133545716603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,8,balanced,0.06888533135255177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,8,balanced,0.07084799806276958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,8,balanced,0.2879413366317749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,8,balanced,0.07047466437021892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,8,balanced,0.07051733136177063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,8,balanced,0.07053333520889282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,8,balanced,0.021509334444999695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,8,balanced,0.0705866664648056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,8,balanced,0.08089600006739299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,8,balanced,0.023141334454218548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,8,balanced,0.07660800218582153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,8,balanced,0.02216000109910965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,8,balanced,0.023232000569502514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,8,balanced,0.31809600194295246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,8,balanced,0.07871999839941661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,8,balanced,0.02914133419593175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,8,balanced,0.3984106779098511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,8,balanced,0.0842026670773824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,8,balanced,0.03347733368476232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,8,balanced,0.03603733330965042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,8,balanced,0.08938666184743245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,8,balanced,0.03625066578388214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,8,balanced,0.037248000502586365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,8,balanced,0.038378665844599404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,8,balanced,0.09300266702969869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,8,balanced,0.03977066775163015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,8,balanced,0.042208001017570496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,8,balanced,0.04419200122356415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,8,balanced,0.3877493143081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,8,balanced,0.1046346624692281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,8,balanced,0.05179200073083242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,8,balanced,0.054485330979029335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,8,balanced,0.14274666706720987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,8,balanced,0.05783999959627787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,8,balanced,0.5019520123799642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,8,balanced,0.08729066451390584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,8,balanced,0.0936853289604187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,8,balanced,0.15963199734687805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,8,balanced,0.1151039997736613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,8,balanced,0.1556053360303243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,8,balanced,0.208186666170756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,8,balanced,0.11473600069681804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,8,balanced,0.4606026808420817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,8,balanced,0.13673599561055502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,8,balanced,0.25838400920232135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,8,balanced,0.6116746664047241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,8,balanced,0.15449600418408713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,8,balanced,0.22857065995534262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,8,balanced,0.36080535252888996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,8,balanced,0.26968000332514447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,8,balanced,0.69268798828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,8,balanced,0.4617120027542114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,8,balanced,0.3880159854888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,8,balanced,0.9365759690602621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,8,balanced,0.5053706566492716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,8,balanced,0.563647985458374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,8,balanced,0.6274186770121256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,8,power_law_1.01,0.05315200090408325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,8,power_law_1.01,0.03804160058498383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,8,power_law_1.01,0.036550399661064145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,8,power_law_1.01,0.06111999750137329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,8,power_law_1.01,0.0316864013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,8,power_law_1.01,0.039155200123786926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,8,power_law_1.01,0.03330560028553009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,8,power_law_1.01,0.0537663996219635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,8,balanced,0.8633759816487631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,8,power_law_1.01,0.04079999923706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,8,power_law_1.01,0.039084801077842714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,8,power_law_1.01,0.059468799829483034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,8,power_law_1.01,0.0463808000087738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,8,power_law_1.01,0.04350079894065857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,8,power_law_1.01,0.07319039702415467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,8,power_law_1.01,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,8,power_law_1.01,0.05432959794998169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,8,power_law_1.01,0.07484800219535828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,8,power_law_1.01,0.05620480179786682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,8,power_law_1.01,0.06033920049667359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,8,power_law_1.01,0.07802240252494812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,8,power_law_1.01,0.058169597387313844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,8,power_law_1.01,0.07022719979286193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,8,balanced,1.2951467037200928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,8,power_law_1.01,0.0576960027217865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,8,power_law_1.01,0.0753600001335144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,8,power_law_1.01,0.07354879975318909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,8,power_law_1.01,0.058924800157546996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,8,power_law_1.01,0.07895680069923401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,8,power_law_1.01,0.05957760214805603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,8,power_law_1.01,0.06959999799728393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,8,power_law_1.01,0.07751039862632751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,8,power_law_1.01,0.061868798732757566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,8,power_law_1.01,0.07753599882125854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,8,power_law_1.01,0.0642624020576477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,8,power_law_1.01,0.06949120163917541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,8,power_law_1.01,0.021670399606227873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,8,power_law_1.01,0.06695680022239685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,8,power_law_1.01,0.07828480005264282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,8,power_law_1.01,0.07090560197830201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,8,power_law_1.01,0.07460479736328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,8,power_law_1.01,0.0812608003616333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,8,balanced,1.791594664255778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,8,power_law_1.01,0.07647359967231751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,8,power_law_1.01,0.0722815990447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,8,power_law_1.01,0.02396160066127777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,8,power_law_1.01,0.08901119828224183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,8,power_law_1.01,0.0806335985660553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,8,power_law_1.01,0.07306879758834839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,8,power_law_1.01,0.09232640266418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,8,power_law_1.01,0.08795520067214965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,8,power_law_1.01,0.020742399990558623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,8,balanced,1.000869353612264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,8,power_law_1.01,0.0735040009021759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,8,power_law_1.01,0.0936191976070404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,8,power_law_1.01,0.10728319883346557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,8,power_law_1.01,0.023545600473880768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,8,power_law_1.01,0.10279680490493774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,8,power_law_1.01,0.07529600262641907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,8,power_law_1.01,0.13011840581893921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,8,power_law_1.01,0.02391040027141571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,8,power_law_1.01,0.13657599687576294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,8,power_law_1.01,0.11508480310440064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,8,power_law_1.01,0.07975680232048035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,8,power_law_1.01,0.027641600370407103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,8,power_law_1.01,0.18119679689407348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,8,power_law_1.01,0.1378175973892212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,8,power_law_1.01,0.029228800535202028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,8,power_law_1.01,0.07989119887351989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,8,power_law_1.01,0.22298879623413087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,8,power_law_1.01,0.02842240035533905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,8,power_law_1.01,0.13408000469207765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,8,power_law_1.01,0.08850560188293458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,8,power_law_1.01,0.31529600620269777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,8,power_law_1.01,0.030611199140548707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,8,power_law_1.01,0.0972544014453888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,8,power_law_1.01,0.18142720460891723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,8,power_law_1.01,0.3954751968383789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,8,power_law_1.01,0.03251200020313263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.01,0.10358400344848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,8,power_law_1.01,0.5574656009674073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,8,power_law_1.01,0.22126080989837646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,8,power_law_1.01,0.7460288047790528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,8,power_law_1.01,0.03372800052165985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.01,0.11985280513763427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,8,power_law_1.01,0.2953088045120239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,8,power_law_1.01,0.8914560317993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,8,power_law_1.01,0.03449600040912628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.01,0.14360320568084717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,8,power_law_1.01,0.348140811920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,8,power_law_1.01,1.2926143646240233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,8,power_law_1.01,0.03947519958019256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.01,0.1993600010871887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,8,power_law_1.01,0.48798718452453616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,8,power_law_1.01,0.04716159999370575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.01,0.2182528018951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,8,power_law_1.01,0.5927423954010009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,8,power_law_1.01,2.8964864730834963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,8,power_law_1.01,0.04935039877891541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.01,0.3039743900299072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,8,power_law_1.01,0.05088000297546387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,8,power_law_1.01,0.7251200199127197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.01,0.37737600803375243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,8,power_law_1.01,0.06895359754562377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,8,power_law_1.01,1.1262271881103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,8,power_law_1.01,0.08376320004463196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.01,0.5157631874084473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.01,0.09568639993667602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.01,0.7144192218780517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.01,0.1164736032485962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,8,power_law_1.01,2.0388288497924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,8,power_law_1.01,0.7922111988067627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.01,0.08494719862937927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.01,0.10787839889526367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,8,power_law_1.01,1.279423999786377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.01,0.12472319602966309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.01,0.16989439725875854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,8,power_law_1.01,2.7164159774780274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.01,0.21276800632476806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.01,0.31740798950195315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.01,0.4027584075927734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,8,balanced,1.6566133499145508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,8,power_law_1.01,0.4737664222717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,8,power_law_1.01,0.746553611755371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,8,power_law_1.01,1.4015104293823242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,8,balanced,1.973514715830485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,8,power_law_1.2,0.05314559936523437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,8,power_law_1.2,0.036051198840141296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,8,power_law_1.2,0.036908799409866334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,8,power_law_1.2,0.037567999958992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,8,power_law_1.2,0.051551997661590576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,8,power_law_1.2,0.052825599908828735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,8,power_law_1.2,0.030687999725341798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,8,power_law_1.2,0.06849279999732971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,8,power_law_1.2,0.03377920091152191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,8,power_law_1.2,0.07660800218582153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,8,power_law_1.2,0.038796800374984744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,8,power_law_1.2,0.060755199193954466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,8,power_law_1.2,0.07534080147743225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,8,power_law_1.2,0.04073599874973297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,8,power_law_1.2,0.04268159866333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,8,power_law_1.2,0.07946239709854126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,8,power_law_1.2,0.053939199447631835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,8,power_law_1.2,0.047014400362968445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,8,power_law_1.2,0.07756800055503846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,8,power_law_1.2,0.05683839917182922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,8,power_law_1.2,0.05273600220680237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,8,power_law_1.2,0.07960960268974304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,8,power_law_1.2,0.05753600001335144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,8,power_law_1.2,0.05767040252685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,8,power_law_1.2,0.08220800161361694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,8,power_law_1.2,0.06892160177230836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,8,power_law_1.2,0.05808640122413635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,8,power_law_1.2,0.07021440267562866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,8,power_law_1.2,0.08313599824905396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,8,power_law_1.2,0.05761280059814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,8,power_law_1.2,0.07008640170097351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,8,power_law_1.2,0.08423039913177491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,8,power_law_1.2,0.06079999804496765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,8,power_law_1.2,0.07184640169143677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,8,power_law_1.2,0.06264960169792175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,8,power_law_1.2,0.09292799830436707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,8,power_law_1.2,0.02059520035982132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,8,power_law_1.2,0.07411839962005615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,8,power_law_1.2,0.0646016001701355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,8,power_law_1.2,0.09500160217285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,8,power_law_1.2,0.07346559762954712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,8,power_law_1.2,0.06946560144424438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,8,power_law_1.2,0.07575680017471313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,8,power_law_1.2,0.09961599707603455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,8,power_law_1.2,0.01990399956703186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,8,power_law_1.2,0.07465599775314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,8,power_law_1.2,0.07187839746475219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,8,power_law_1.2,0.10894720554351807
TRTLLM,1.2.0rc5,NVIDIA B200,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 B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,8,power_law_1.2,0.07694720029830933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,8,power_law_1.2,0.07608960270881653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,8,power_law_1.2,0.12436480522155761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,8,power_law_1.2,0.08196480274200439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,8,power_law_1.2,0.021139200031757354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,8,power_law_1.2,0.08110079765319825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,8,power_law_1.2,0.08562560081481933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,8,power_law_1.2,0.16266239881515504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,8,power_law_1.2,0.023366400599479677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,8,power_law_1.2,0.09125760197639465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,8,power_law_1.2,0.0898751974105835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,8,power_law_1.2,0.027116799354553224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,8,power_law_1.2,0.16108160018920897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,8,power_law_1.2,0.10056960582733154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,8,power_law_1.2,0.11292159557342529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,8,power_law_1.2,0.030425599217414855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.2,0.11259520053863525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,8,power_law_1.2,0.22136321067810058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,8,power_law_1.2,0.13180160522460938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,8,power_law_1.2,0.029817599058151244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.2,0.12869759798049926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,8,power_law_1.2,0.030656000971794127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,8,power_law_1.2,0.14775680303573607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,8,power_law_1.2,0.23383040428161622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.2,0.1494976043701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,8,power_law_1.2,0.032652801275253295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,4,balanced,0.031744000812371574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.2,0.196563196182251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,8,power_law_1.2,0.3731584072113037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,8,power_law_1.2,0.2028287887573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,4,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.2,0.25408000946044923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,8,power_law_1.2,0.2346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,8,power_law_1.2,0.3999295949935913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,8,power_law_1.2,0.03410559892654419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,4,balanced,0.029680001238981884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,4,balanced,0.03362133353948593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.2,0.33882238864898684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,8,power_law_1.2,0.3220223903656006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,4,balanced,0.04387733340263367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,8,power_law_1.2,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,8,power_law_1.2,0.6388864040374755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,4,balanced,0.035162667433420815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.2,0.4800576210021973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,4,balanced,0.01950399950146675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,4,balanced,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,4,balanced,0.04957333207130432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,8,power_law_1.2,0.425216007232666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,4,balanced,0.03339199970165888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,4,balanced,0.05798399945100149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,4,balanced,0.06382399797439575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.2,0.6318528175354003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,8,power_law_1.2,0.039001598954200745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,8,power_law_1.2,0.6624063968658447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,4,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,8,power_law_1.2,0.7750912189483643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,4,balanced,0.021018666525681812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,4,balanced,0.07972800234953563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,4,balanced,0.08488532900810242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,4,balanced,0.058592001597086586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,8,power_law_1.2,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.2,0.8566911697387696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,4,balanced,0.08137066662311554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,8,power_law_1.2,0.9223615646362304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,4,balanced,0.08656000097592671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,4,balanced,0.02165866643190384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,4,balanced,0.06202666461467743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,8,power_law_1.2,0.8953087806701661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,4,balanced,0.08292800188064575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,8,power_law_1.2,0.05002880096435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,8,power_law_1.2,0.9119423866271973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,8,power_law_1.2,1.1263872146606446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,4,balanced,0.08349333206812541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,4,balanced,0.06412800153096516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,4,balanced,0.023258666197458904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,8,power_law_1.2,1.347475242614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,4,balanced,0.0846560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,8,power_law_1.2,0.05203199982643127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,8,power_law_1.2,1.9231359481811523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,4,balanced,0.02932800104220708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,4,balanced,0.06216000020503998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,4,balanced,0.08268799881140391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,8,power_law_1.2,0.07064319849014282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,4,balanced,0.0832586685816447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,8,power_law_1.2,1.8227840423583985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,4,balanced,0.03133333226044973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,4,balanced,0.063701331615448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,8,power_law_1.2,3.609164810180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,8,power_law_1.2,3.555545425415039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,4,balanced,0.08457066615422566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,8,power_law_1.2,0.08728960156440735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,4,balanced,0.033941333492596946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,4,balanced,0.08375466863314311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,8,power_law_1.2,3.956697463989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,4,balanced,0.06299733122189839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,4,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,4,balanced,0.08343999584515889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,4,balanced,0.082096000512441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.2,0.09512320160865784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,4,balanced,0.06492800017197926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,4,balanced,0.03762666632731756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,4,balanced,0.08425600330034892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,4,balanced,0.06842133402824402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,4,balanced,0.08687999844551086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,4,balanced,0.08181333541870117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.2,0.12394239902496337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,4,balanced,0.04135466615358988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,4,balanced,0.08828799923261006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,4,balanced,0.06842133402824402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,4,balanced,0.08688533306121826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.2,0.08929920196533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,4,balanced,0.04008533308903376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,4,balanced,0.07187733550866444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,4,balanced,0.09333333373069763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,4,balanced,0.07941866914431255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,4,balanced,0.04225599765777588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,4,balanced,0.09485866626103719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.2,0.10831999778747559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,4,balanced,0.08684800068537395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,4,balanced,0.046154667933781944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,4,balanced,0.08002133170763652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,4,balanced,0.09915199875831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.2,0.13393919467926024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,4,balanced,0.08276266853014629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,4,balanced,0.04614933331807455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,4,balanced,0.08321600159009297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,4,balanced,0.10496000448862712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,4,balanced,0.09471467137336731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.2,0.18927359580993652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,4,balanced,0.04816000163555145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,4,balanced,0.09237866600354512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,4,balanced,0.11508799592653911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,4,balanced,0.09099200367927551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.2,0.23120639324188233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,4,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,4,balanced,0.10191999872525533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,4,balanced,0.1271573305130005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,4,balanced,0.05966933568318685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,4,balanced,0.12243733803431193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,4,balanced,0.0950879951318105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.2,0.3281280040740967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,4,balanced,0.10233066479365031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,4,balanced,0.06366399923960368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,4,balanced,0.14334932963053384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,4,balanced,0.13819733262062073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,4,balanced,0.11085333426793416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.2,0.4229248046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,8,power_law_1.2,0.528492784500122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,4,balanced,0.10017599662144978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,4,balanced,0.123690664768219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,8,power_law_1.2,0.8242752075195312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,4,balanced,0.1762346625328064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,4,balanced,0.17940266927083334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,4,balanced,0.1262986660003662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,4,balanced,0.14129066467285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,8,power_law_1.2,1.6152576446533202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,4,balanced,0.21253865957260132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,4,balanced,0.20759467283884683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,4,balanced,0.11382399996121724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,4,balanced,0.19867199659347534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,4,balanced,0.13025066256523132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,4,balanced,0.22817599773406982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,4,balanced,0.27265600363413495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,4,balanced,0.29708800713221234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,4,balanced,0.14096533258756003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,4,balanced,0.3105439941088359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,4,balanced,0.3377546469370524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,4,balanced,0.3868853251139323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,4,balanced,0.2175146738688151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,4,balanced,0.3923413356145223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,4,balanced,0.24841066201527914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,4,balanced,0.4617120027542114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,4,balanced,0.5532480080922445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,4,balanced,0.5569279988606771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,4,balanced,0.3540960152943929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,4,balanced,0.5815146764119467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,4,balanced,0.7133973439534506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,4,balanced,0.7276533444722494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,4,balanced,0.4625653425852458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,4,balanced,0.6985387007395426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,4,balanced,0.8785280386606852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,4,balanced,0.8877013524373373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,4,balanced,0.5724320014317831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,4,balanced,1.0604480107625325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,4,balanced,1.3647039731343586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,4,balanced,1.3838079770406086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,4,balanced,0.9099520047505697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,4,power_law_1.01,0.036671999096870425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,4,power_law_1.01,0.03232640027999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,4,power_law_1.01,0.05431680083274841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,4,power_law_1.01,0.035308799147605895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,4,power_law_1.01,0.03189760148525238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,4,power_law_1.01,0.04936319887638092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,4,power_law_1.01,0.05746560096740723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,4,power_law_1.01,0.030406400561332703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,4,power_law_1.01,0.04650880098342895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,4,power_law_1.01,0.03315840065479279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,4,power_law_1.01,0.07279999852180481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,4,power_law_1.01,0.05281919836997986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,4,power_law_1.01,0.03934719860553741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,4,power_law_1.01,0.07410560250282287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,4,power_law_1.01,0.061222398281097413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,4,power_law_1.01,0.04255360066890716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,4,power_law_1.01,0.07731199860572815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,4,power_law_1.01,0.057740801572799684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,4,power_law_1.01,0.0689087986946106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,4,power_law_1.01,0.07980160117149353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,4,power_law_1.01,0.05817599892616272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,4,power_law_1.01,0.08259199857711792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,4,power_law_1.01,0.08254719972610473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,4,power_law_1.01,0.06042240262031555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,4,power_law_1.01,0.08043519854545593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,4,power_law_1.01,0.08168320059776306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,4,power_law_1.01,0.06020479798316956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,4,power_law_1.01,0.08291199803352356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,4,power_law_1.01,0.08364800214767457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,4,power_law_1.01,0.06063359975814819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,4,power_law_1.01,0.08447999954223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,4,power_law_1.01,0.08453760147094727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,4,power_law_1.01,0.0651968002319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,4,power_law_1.01,0.09008640050888062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,4,power_law_1.01,0.06666240096092224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,4,power_law_1.01,0.08329600095748901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,4,power_law_1.01,0.09747200012207032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,4,power_law_1.01,0.06888960003852844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,4,power_law_1.01,0.0843775987625122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,4,power_law_1.01,0.07528319954872131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,4,power_law_1.01,0.09861760139465332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,4,power_law_1.01,0.08444160223007202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,4,power_law_1.01,0.08055679798126221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,4,power_law_1.01,0.10268800258636475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,4,power_law_1.01,0.08867200016975403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,4,power_law_1.01,0.08727040290832519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,4,power_law_1.01,0.1128383994102478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,4,power_law_1.01,0.08972160220146179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,4,power_law_1.01,0.08607360124588012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,4,power_law_1.01,0.12798080444335938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,4,power_law_1.01,0.10129280090332031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,4,power_law_1.01,0.09356799721717834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,4,power_law_1.01,0.15205760002136232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,4,power_law_1.01,0.1118656039237976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,4,power_law_1.01,0.09639679789543151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,4,power_law_1.01,0.16577279567718506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,4,power_law_1.01,0.1355712056159973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,4,power_law_1.01,0.10014079809188843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,4,power_law_1.01,0.19982080459594725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,4,power_law_1.01,0.15907200574874877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,4,power_law_1.01,0.11434240341186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,4,power_law_1.01,0.2055488109588623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,4,power_law_1.01,0.24311039447784424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.01,0.1289728045463562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,4,power_law_1.01,0.25471360683441163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,4,power_law_1.01,0.3329087972640991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.01,0.14966399669647218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,4,power_law_1.01,0.3554111957550049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,4,power_law_1.01,0.38161280155181887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.01,0.1695680022239685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,4,power_law_1.01,0.47162880897521975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,4,power_law_1.01,0.659116792678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.01,0.2428607940673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,4,power_law_1.01,0.7091392040252685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,4,power_law_1.01,0.7501503944396972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.01,0.2802432060241699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,4,power_law_1.01,0.9047679901123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,4,power_law_1.01,0.8654272079467773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.01,0.3802495956420898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,4,power_law_1.01,1.0685376167297362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.01,0.49854722023010256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,4,power_law_1.01,1.3542271614074708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,4,power_law_1.01,1.7652223587036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.01,0.7182655811309815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.01,0.8295424461364747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,4,power_law_1.01,3.129100799560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,4,power_law_1.01,2.834835243225098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,4,power_law_1.01,1.1332032203674316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,4,power_law_1.01,1.828268814086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,4,balanced,2.0356106758117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,4,balanced,2.585514704386393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,4,power_law_1.01,3.4551998138427735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,4,balanced,2.6804161071777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,4,power_law_1.01,0.01860480010509491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,4,balanced,1.789445400238037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,4,power_law_1.01,0.01934719979763031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,4,power_law_1.01,0.02003840059041977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,4,power_law_1.01,0.02099840044975281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,4,power_law_1.01,0.02389120012521744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,4,power_law_1.01,0.027456000447273254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,4,power_law_1.01,0.029465600848197937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,4,power_law_1.01,0.03218559920787811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,4,power_law_1.01,0.03487359881401062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,4,power_law_1.01,0.03726080060005188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,4,power_law_1.01,0.034822401404380796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,4,power_law_1.01,0.03751679956912994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,4,power_law_1.01,0.039776000380516055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,4,power_law_1.01,0.03823359906673431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,4,power_law_1.01,0.04200960099697113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,4,power_law_1.01,0.04450559914112091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,4,power_law_1.01,0.046265599131584165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,4,power_law_1.01,0.054688000679016115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.01,0.093612802028656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.01,0.10574719905853272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.01,0.09286400079727172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.01,0.11146240234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.01,0.13246079683303832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.01,0.17032320499420167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.01,0.21208319664001465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.01,0.30410239696502683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.01,0.3806335926055908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,4,power_law_1.01,0.45058560371398926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,4,power_law_1.01,0.6866879940032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,4,power_law_1.01,1.432960033416748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,4,power_law_1.2,0.037913599610328676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,4,power_law_1.2,0.032102400064468385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,4,power_law_1.2,0.03713279962539673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,4,power_law_1.2,0.043808001279830935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,4,power_law_1.2,0.05555840134620667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,4,power_law_1.2,0.07376639842987061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,4,power_law_1.2,0.07631360292434693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,4,power_law_1.2,0.07908480167388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,4,power_law_1.2,0.08109440207481385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,4,power_law_1.2,0.08424320220947265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,4,power_law_1.2,0.08147839903831482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,4,power_law_1.2,0.08327040076255798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,2,balanced,0.031712000568707786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,4,power_law_1.2,0.052767997980117796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,4,power_law_1.2,0.018918399512767792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,4,power_law_1.2,0.08599680066108703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,4,power_law_1.2,0.032416000962257385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,2,balanced,0.03176533430814743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,4,power_law_1.2,0.09245439767837524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,4,power_law_1.2,0.018931199610233308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,2,balanced,0.0335359995563825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,4,power_law_1.2,0.04323840141296387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,4,power_law_1.2,0.09822720289230347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,4,power_law_1.2,0.020121599733829498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,4,power_law_1.2,0.03115519881248474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,2,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,2,balanced,0.03267733256022135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,4,power_law_1.2,0.1041152000427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,4,power_law_1.2,0.019923199713230134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,2,power_law_1.01,0.03683840036392212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,4,power_law_1.2,0.045817598700523376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,4,power_law_1.2,0.03418239951133728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,2,balanced,0.045797333121299744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,2,balanced,0.0378506655494372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,4,power_law_1.2,0.10737279653549195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,4,power_law_1.2,0.023219199478626253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,4,power_law_1.2,0.05431680083274841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,2,balanced,0.06870933373769124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,4,power_law_1.2,0.027654400467872618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,2,balanced,0.05852800110975901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,4,power_law_1.2,0.11585279703140258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,4,power_law_1.2,0.04005120098590851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,2,power_law_1.01,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,4,power_law_1.2,0.02968960106372833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,2,balanced,0.0697920024394989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,4,power_law_1.2,0.06261119842529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,2,balanced,0.08499200145403545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,4,power_law_1.2,0.13082879781723022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,4,power_law_1.2,0.041177600622177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,4,power_law_1.2,0.03200640082359314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,2,power_law_1.01,0.03616639971733093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,2,balanced,0.07032000025113423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,4,power_law_1.2,0.08078719973564148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,2,balanced,0.08724266290664673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,4,power_law_1.2,0.1593791961669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,2,power_law_1.01,0.04912639856338501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,4,power_law_1.2,0.05724160075187683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,4,power_law_1.2,0.03528960049152374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,2,balanced,0.07449066638946533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,4,power_law_1.2,0.08579840064048767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,2,power_law_1.01,0.05690240263938904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,2,balanced,0.08761066198348999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,4,power_law_1.2,0.18709759712219237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,4,power_law_1.2,0.05958399772644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,4,power_law_1.2,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,4,power_law_1.2,0.08280959725379944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,2,balanced,0.07528533538182576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,2,balanced,0.08967999617258708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,2,power_law_1.01,0.07719680070877075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,4,power_law_1.2,0.06164479851722717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,4,power_law_1.2,0.21345279216766358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,2,balanced,0.08990933497746785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,2,power_law_1.01,0.07965440154075623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,4,power_law_1.2,0.06106240153312683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,4,power_law_1.2,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,2,balanced,0.07625066737333934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,4,power_law_1.2,0.08686720132827759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,2,balanced,0.09040533502896626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,2,power_law_1.01,0.0844223976135254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,2,balanced,0.07785599927107494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,4,power_law_1.2,0.06227200031280518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,4,power_law_1.2,0.0847104012966156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,4,power_law_1.2,0.03800959885120392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,4,power_law_1.2,0.2747136116027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,2,balanced,0.09238933523495992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,2,power_law_1.01,0.08405759930610657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,2,balanced,0.08269866804281871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,4,power_law_1.2,0.06455680131912231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,4,power_law_1.2,0.040300801396369934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,4,power_law_1.2,0.08501120209693909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,4,power_law_1.2,0.3911679983139038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,2,power_law_1.01,0.08698880076408386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,2,balanced,0.09338666995366414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,2,balanced,0.0806879997253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,4,power_law_1.2,0.03895680010318756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,4,power_law_1.2,0.08565760254859925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,4,power_law_1.2,0.0673792004585266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,2,power_law_1.01,0.0905023992061615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,2,balanced,0.09518399834632874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,2,power_law_1.01,0.030713599920272828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,4,power_law_1.2,0.4258944034576416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,2,balanced,0.08443733056386311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,4,power_law_1.2,0.04223999977111816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,4,power_law_1.2,0.08958719968795777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,2,power_law_1.01,0.09462400078773499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,4,power_law_1.2,0.07166720032691956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,2,balanced,0.0976106623808543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,4,power_law_1.2,0.6544320106506347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,4,power_law_1.2,0.044537600874900815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,2,balanced,0.093231995900472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,2,power_law_1.01,0.09617919921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,4,power_law_1.2,0.09065600037574768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,4,power_law_1.2,0.07612159848213196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,2,power_law_1.01,0.032332798838615416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,2,balanced,0.10313066840171814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,4,power_law_1.2,0.04748159945011139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,2,power_law_1.01,0.09720960259437561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,4,power_law_1.2,0.09530240297317505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,4,power_law_1.2,0.08213760256767273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,2,balanced,0.09506666660308838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,4,power_law_1.2,0.866431999206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,2,balanced,0.1065013309319814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,2,power_law_1.01,0.034969601035118106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,4,power_law_1.2,0.05567359924316406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,2,power_law_1.01,0.10678399801254272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,4,power_law_1.2,0.09918720126152039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,4,power_law_1.2,0.08909440040588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.2,0.0942080020904541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,2,balanced,0.11318399508794148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,2,balanced,0.09849066535631816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,2,power_law_1.01,0.04276480078697205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,2,power_law_1.01,0.10743680000305175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,4,power_law_1.2,1.0626751899719238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,4,power_law_1.2,0.08979200124740601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,4,power_law_1.2,0.11073919534683227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.2,0.11083519458770752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,2,balanced,0.13240533073743185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,2,power_law_1.01,0.1331455945968628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,2,power_law_1.01,0.04835200011730194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,2,balanced,0.11471466223398845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,4,power_law_1.2,1.7682559967041016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,2,power_law_1.01,0.15196160078048707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,2,power_law_1.01,0.06594560146331788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.2,0.09440640211105347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,4,power_law_1.2,0.11646080017089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,4,power_law_1.2,0.10421119928359986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,2,balanced,0.13222400347391763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,2,balanced,0.13357333342234293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,2,power_law_1.01,0.14362879991531372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,2,power_law_1.01,0.06445440053939819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,4,power_law_1.2,3.8134910583496096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.2,0.1178879976272583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,2,power_law_1.01,0.17426559925079346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.2,0.1296064019203186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,4,power_law_1.2,0.12024960517883301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,2,power_law_1.01,0.0691648006439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.2,0.1344831943511963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,2,power_law_1.01,0.20741119384765624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,2,power_law_1.01,0.07070080041885377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,2,balanced,0.16159466902414957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.2,0.15402239561080933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,2,balanced,0.15717867016792297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,4,power_law_1.2,0.14405759572982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,2,power_law_1.01,0.2688256025314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.2,0.18280320167541503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,2,power_law_1.01,0.07440000176429748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.2,0.18418560028076172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,4,power_law_1.2,0.16208640336990357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,2,power_law_1.01,0.3377279996871948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.2,0.2214143991470337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,2,power_law_1.01,0.07422080039978027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.2,0.25095040798187257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,2,balanced,0.17679466803868613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,2,power_law_1.01,0.45951361656188966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,2,balanced,0.18159999450047812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.2,0.30662400722503663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,2,power_law_1.01,0.07747840285301208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,4,power_law_1.2,0.22442240715026857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.2,0.3171456098556519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,2,power_law_1.01,0.5823872089385986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,2,power_law_1.01,0.07971199750900268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.2,0.3979648113250732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,2,power_law_1.01,0.8322943687438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,2,power_law_1.01,0.08539519906044006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,4,power_law_1.2,0.3047231912612915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.2,0.41139841079711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,4,power_law_1.2,0.49161601066589355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,2,balanced,0.24864532550175986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,2,power_law_1.01,0.09221119880676269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,2,power_law_1.01,1.0469504356384278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,2,balanced,0.26336532831192017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.2,0.5370751857757569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,4,power_law_1.2,0.40073599815368655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,4,power_law_1.2,0.777510404586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,2,power_law_1.01,0.0964352011680603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,2,power_law_1.01,1.3006272315979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.2,0.7976511955261231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,4,power_law_1.2,0.5038911819458007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,2,power_law_1.01,0.10524159669876099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,4,power_law_1.2,1.5316160202026368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,2,power_law_1.01,2.036390495300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,2,power_law_1.01,0.12254719734191895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,2,balanced,0.2834826707839966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,4,power_law_1.2,0.9118720054626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.2,1.1068927764892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,2,balanced,0.31455467144648236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,2,power_law_1.01,0.1420799970626831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,2,power_law_1.01,3.873971176147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,4,power_law_1.2,1.0605695724487305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,2,power_law_1.01,0.1820736050605774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,4,power_law_1.2,1.313804817199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,2,power_law_1.01,0.21240320205688476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,4,power_law_1.2,1.3421312332153321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,2,power_law_1.01,0.30434560775756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,4,power_law_1.2,2.127289581298828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,2,balanced,0.40465601285298664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,2,power_law_1.01,0.35759360790252687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,4,power_law_1.2,2.169657516479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,2,balanced,0.45742400487263996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,2,power_law_1.01,0.5249023914337159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,4,power_law_1.2,4.299193572998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,2,power_law_1.01,0.6496384143829346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,4,power_law_1.2,3.953753662109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,2,power_law_1.01,0.9953408241271973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,2,power_law_1.01,1.3452608108520507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,2,balanced,0.5053919951121012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,2,power_law_1.01,1.5706368446350099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,2,balanced,0.5849386850992838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,2,power_law_1.01,2.40832633972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,2,power_law_1.01,4.795065689086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,2,balanced,0.7197973728179932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,2,balanced,0.8519466718037924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,2,balanced,0.9137653509775797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,2,balanced,1.113637367884318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,2,balanced,1.1256799697875977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,2,balanced,1.384895960489909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,2,balanced,1.7489919662475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,2,balanced,2.147973378499349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,2,power_law_1.2,0.0366784006357193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,2,balanced,3.3543148040771484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,2,power_law_1.2,0.033369600772857666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,2,power_law_1.2,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,2,balanced,4.1758880615234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,2,power_law_1.2,0.04496639966964722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,2,power_law_1.2,0.05752320289611816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,2,power_law_1.2,0.07879040241241456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,2,power_law_1.2,0.08064640164375306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,2,power_law_1.2,0.08669440150260925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,2,power_law_1.2,0.08538240194320679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,2,power_law_1.2,0.08734080195426941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,1,balanced,0.035690667728583016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,2,power_law_1.2,0.09397119879722596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,2,power_law_1.2,0.03078399896621704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,2,power_law_1.2,0.09618560075759888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,1,balanced,0.03916800022125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,2,power_law_1.2,0.09992960095405579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,2,power_law_1.2,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,1,balanced,0.045466666420300804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,2,power_law_1.2,0.10145920515060425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,2,power_law_1.2,0.035071998834609985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,2,power_law_1.2,0.1076159954071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,2,power_law_1.2,0.04173440039157868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,1,balanced,0.06854400038719177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,2,power_law_1.2,0.11156480312347412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,2,power_law_1.2,0.04625279903411865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,1,balanced,0.10154666503270467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,2,power_law_1.2,0.1335935950279236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,2,power_law_1.2,0.0648959994316101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,1,balanced,0.10358400146166484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,2,power_law_1.2,0.15569280385971068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,2,power_law_1.2,0.06709759831428527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,1,balanced,0.1056106686592102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,2,power_law_1.2,0.1497599959373474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,1,balanced,0.10597866773605347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,2,power_law_1.2,0.06912000179290771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,2,power_law_1.2,0.18969600200653075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,1,balanced,0.10565333565076192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,2,power_law_1.2,0.0725055992603302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,2,power_law_1.2,0.22225279808044435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,1,balanced,0.10716799894968669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,2,power_law_1.2,0.07393919825553893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,2,power_law_1.2,0.2764415979385376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,1,balanced,0.11147200067838033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,2,power_law_1.2,0.34415359497070314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,2,power_law_1.2,0.07542399764060974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,1,balanced,0.11326400438944499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,1,balanced,0.035573333501815796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,2,power_law_1.2,0.4799232006072998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,2,power_law_1.2,0.07994239926338195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,1,balanced,0.11661332845687866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,2,power_law_1.2,0.590444803237915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,1,balanced,0.12078400452931722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,1,power_law_1.01,0.03162240087985992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,1,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,2,power_law_1.2,0.833625602722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,1,power_law_1.01,0.03208959996700287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,2,power_law_1.2,0.08133760094642639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,1,balanced,0.12821333607037863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,1,balanced,0.042026668787002563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,2,power_law_1.2,1.1843584060668946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,1,balanced,0.1325973371664683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,2,power_law_1.2,0.08662400245666504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,1,power_law_1.01,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,1,power_law_1.01,0.03468160033226013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,1,balanced,0.055786664287249245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,2,power_law_1.2,1.278048038482666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,1,power_law_1.01,0.03901439905166626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,2,power_law_1.2,0.09312639832496643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,1,balanced,0.14218133687973022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,1,balanced,0.08762666583061218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,2,power_law_1.2,0.09956480264663696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,1,power_law_1.01,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,1,power_law_1.01,0.041606399416923526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,2,power_law_1.2,2.112351989746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,1,balanced,0.16214932998021445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,1,balanced,0.08457600076993306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,1,power_law_1.01,0.056979197263717654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,2,power_law_1.2,0.10531840324401856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,1,power_law_1.01,0.053446400165557864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,1,balanced,0.0886400043964386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,1,balanced,0.1753013332684835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,2,power_law_1.2,4.232908630371094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,1,power_law_1.01,0.07642880082130432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,2,power_law_1.2,0.12914559841156006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,1,power_law_1.01,0.06444159746170045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,1,balanced,0.08738133311271667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,1,power_law_1.01,0.07976319789886474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,2,power_law_1.2,0.14479360580444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,1,balanced,0.25250667333602905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,1,power_law_1.01,0.09308159947395325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,1,balanced,0.08906132976214091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,1,power_law_1.01,0.08174719810485839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,2,power_law_1.2,0.1889024019241333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,1,power_law_1.01,0.09474560022354125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,1,balanced,0.27373866240183514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,2,power_law_1.2,0.2256319999694824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,1,power_law_1.01,0.08632959723472595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,1,balanced,0.09098666906356812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,1,power_law_1.01,0.09733759760856628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,1,power_law_1.01,0.08858879804611205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,1,balanced,0.09399466713269551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,1,power_law_1.01,0.09897599816322326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,1,balanced,0.4004960060119629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,2,power_law_1.2,0.3091520071029663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,1,balanced,0.09805333614349365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,1,power_law_1.01,0.0955136001110077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,1,power_law_1.01,0.10202239751815796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,2,power_law_1.2,0.38657279014587403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,1,balanced,0.4766720136006673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,1,power_law_1.01,0.10165760517120362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,1,balanced,0.09969600041707356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,1,power_law_1.01,0.10459519624710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,2,power_law_1.2,0.5436543941497802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,1,power_law_1.01,0.11025279760360718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,1,power_law_1.01,0.10680320262908935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,1,balanced,0.10731200377146403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,1,power_law_1.01,0.10982400178909302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,2,power_law_1.2,0.7189631938934327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,1,power_law_1.01,0.10684800148010254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,1,balanced,0.11834133664766948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,1,power_law_1.01,0.13767679929733276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,1,power_law_1.01,0.11632000207901001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,1,power_law_1.01,0.14298880100250244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,1,balanced,0.12196800112724304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,2,power_law_1.2,1.0110719680786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,1,power_law_1.01,0.12455040216445923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,1,power_law_1.01,0.14755200147628783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,1,balanced,0.6831733385721842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,1,balanced,0.13026133179664612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,1,power_law_1.01,0.1628991961479187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,1,power_law_1.01,0.1361024022102356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,1,power_law_1.01,0.1843392014503479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,1,balanced,0.15270400047302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,2,power_law_1.2,1.267289638519287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,1,power_law_1.01,0.16551680564880372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,1,power_law_1.01,0.1964095950126648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,1,power_law_1.01,0.19808000326156616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,1,balanced,0.18745599190394083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,2,power_law_1.2,1.5103039741516113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,1,power_law_1.01,0.2525376081466675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,1,balanced,0.8594240347544352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,1,power_law_1.01,0.2562560081481934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,1,balanced,0.2330026626586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,1,power_law_1.01,0.3032576084136963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,2,power_law_1.2,2.4986431121826174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,1,power_law_1.01,0.31256320476531985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,1,power_law_1.01,0.4050943851470947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,1,power_law_1.01,0.4280128002166748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,1,power_law_1.01,0.5059199810028077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,1,balanced,0.279423991839091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,2,power_law_1.2,5.824307250976562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,1,power_law_1.01,0.6890687942504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,1,power_law_1.01,0.5405312061309815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,1,power_law_1.01,0.8800448417663574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,1,power_law_1.01,0.7614719867706299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,1,balanced,1.2515679995218914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,1,power_law_1.01,1.2418368339538575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,1,balanced,0.4137386480967204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,1,power_law_1.01,0.9892736434936523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,1,power_law_1.01,1.5996543884277343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,1,power_law_1.01,1.962758445739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,1,power_law_1.01,1.4355327606201171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,1,balanced,0.5124906698862711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,1,power_law_1.01,3.0554176330566407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,1,power_law_1.01,1.871366310119629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,1,power_law_1.01,5.946124649047851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,1,power_law_1.01,2.3243839263916017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,1,balanced,0.748522679011027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,1,power_law_1.01,3.6258880615234377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,1,balanced,1.6090985933939617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,1,power_law_1.01,7.148057556152343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,1,balanced,0.9758666356404623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,1,balanced,1.4455199241638184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,1,balanced,2.005669275919596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,1,power_law_1.2,0.0318336009979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,1,power_law_1.2,0.036160001158714296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,1,power_law_1.2,0.039417600631713866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,1,power_law_1.2,0.03089280128479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,1,power_law_1.2,0.0490880012512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,1,power_law_1.2,0.06370559930801392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,1,power_law_1.2,0.03436160087585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,1,power_law_1.2,0.03814400136470795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,1,power_law_1.2,0.09427199959754944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,1,power_law_1.2,0.045798400044441225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,1,power_law_1.2,0.09548159837722778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,1,balanced,1.8710133234659831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,1,power_law_1.2,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,1,power_law_1.2,0.09958400130271912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,1,power_law_1.2,0.07621759772300721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,1,power_law_1.2,0.10176639556884766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,1,power_law_1.2,0.0797439992427826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,1,power_law_1.2,0.10215679407119752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,1,power_law_1.2,0.08389760255813598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,1,power_law_1.2,0.10648319721221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,1,power_law_1.2,0.08703359961509705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,1,power_law_1.2,0.11006720066070556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,1,power_law_1.2,0.0899392008781433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,1,power_law_1.2,0.11299200057983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,1,power_law_1.2,0.09596800208091735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,1,power_law_1.2,0.1373695969581604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,1,power_law_1.2,0.10206079483032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,1,power_law_1.2,0.10759680271148682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,1,power_law_1.2,0.14448000192642213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,1,power_law_1.2,0.14828799962997435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,1,power_law_1.2,0.10729600191116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,1,power_law_1.2,0.16479359865188598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,1,power_law_1.2,0.11792639493942261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,1,power_law_1.2,0.18609919548034667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,1,power_law_1.2,0.12494080066680908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,1,power_law_1.2,0.19711999893188475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,1,power_law_1.2,0.1354367971420288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,1,power_law_1.2,0.25089919567108154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,1,power_law_1.2,0.16573439836502074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,1,balanced,2.333397388458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,1,balanced,3.0930185317993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,1,power_law_1.2,0.3067199945449829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,1,power_law_1.2,0.20074880123138428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,1,power_law_1.2,0.4046016216278076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,1,power_law_1.2,0.2569216012954712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,1,power_law_1.2,0.508838415145874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,1,power_law_1.2,0.3135296106338501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,1,power_law_1.2,0.6852863788604736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,1,power_law_1.2,0.4254784107208252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,1,power_law_1.2,0.8805503845214844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,1,power_law_1.2,0.5379327774047852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,1,power_law_1.2,0.7546688079833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,1,power_law_1.2,1.2406720161437987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,1,power_law_1.2,0.9874624252319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,1,power_law_1.2,1.593734359741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,1,power_law_1.2,1.430675220489502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,1,power_law_1.2,1.9540672302246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,1,power_law_1.2,1.8633920669555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,1,power_law_1.2,3.047871971130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,1,power_law_1.2,2.3144704818725588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,1,power_law_1.2,5.9338623046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,1,power_law_1.2,3.6242176055908204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,1,power_law_1.2,7.140275573730468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,1,balanced,3.6505438486735025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,16,balanced,0.05197333296140035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,16,balanced,0.05487466851870219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,16,balanced,0.07593066493670146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,16,balanced,0.10372266173362732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,16,balanced,0.1628320018450419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,16,balanced,0.18753600120544434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,16,balanced,0.18491200606028238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,16,balanced,0.1876266598701477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,16,balanced,0.1880106727282206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,16,balanced,0.18690667549769083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,16,balanced,0.18817599614461264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,16,balanced,0.18965866168340048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,16,balanced,0.19103467464447021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,16,balanced,0.1884160041809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,16,balanced,0.19477866093317667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,16,balanced,0.19606399536132812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,16,balanced,0.19876799980799356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,16,balanced,0.21794132391611734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,16,balanced,0.20915732781092325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,16,balanced,0.24499199787775675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,16,balanced,0.23478400707244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,1,balanced,5.977039972941081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,16,balanced,0.3326080044110616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,16,balanced,0.3006773392359416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,16,balanced,0.04569066564242045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,16,balanced,0.45044267177581787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,16,balanced,0.04584000011285146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,16,balanced,0.050053333242734276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,16,balanced,0.07656533519426982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,16,balanced,0.10521599650382996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,16,balanced,0.11753066380818684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,16,balanced,0.11915733416875203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,16,balanced,0.11916800340016682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,16,balanced,0.45770132541656494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,16,balanced,0.12103999654452006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,16,balanced,0.11957333485285442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,16,balanced,0.12230400244394939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,16,balanced,0.12302399675051372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,1,balanced,7.149381637573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,16,balanced,0.12276267011960347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,16,balanced,0.12719466288884482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,16,balanced,0.13088533282279968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,16,balanced,0.05118933320045471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,16,balanced,0.13332800070444742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,16,balanced,0.04966933528582255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,16,balanced,0.13742933670679727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,16,balanced,0.050000001986821495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,16,balanced,0.1537866691748301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,16,balanced,0.7449173132578532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,16,balanced,0.06214933097362518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,16,balanced,0.08868267138799031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,16,balanced,0.15402133266131082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,16,balanced,0.09477866689364116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,16,balanced,0.09585600097974141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,16,balanced,0.09841066598892212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,16,balanced,0.1968160072962443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,16,balanced,0.09826133648554485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,16,balanced,0.09802666306495667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,16,balanced,0.1889866590499878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,16,balanced,0.09938133756319682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,16,balanced,0.09963732957839966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,16,balanced,0.09896533687909444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,16,balanced,0.04865066707134247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,16,balanced,0.10294933120409648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,16,balanced,0.2462773323059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,16,balanced,0.1053600013256073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,16,balanced,0.04128533353408178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,16,balanced,0.10807466506958008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,16,balanced,0.7741013367970785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,16,balanced,0.043978666265805565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,16,balanced,0.1107413371404012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,16,balanced,0.05014933149019877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,16,balanced,0.25974400838216144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,16,balanced,0.05834666887919108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,16,balanced,0.06623466809590657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,16,balanced,0.11754666765530904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,16,balanced,0.08628799517949422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,16,balanced,0.12366933623949687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,16,balanced,0.08296533425649007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,16,balanced,0.10443733135859172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,16,balanced,0.1507200002670288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,16,balanced,0.10982400178909302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,16,balanced,0.3991573254267375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,16,balanced,0.11635733644167583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,16,balanced,0.14441600441932678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,16,balanced,0.15227199594179788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,16,balanced,0.1241973340511322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,16,balanced,0.1302773356437683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,16,balanced,0.15083733201026917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,16,balanced,0.18751466274261475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,16,balanced,0.1511893371740977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,16,balanced,0.42819201946258545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,16,balanced,0.1368106702963511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,16,balanced,0.15228266517321268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,16,balanced,0.21094934145609537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,16,balanced,0.8744586308797201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,16,balanced,0.179802676041921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,16,balanced,0.22739199797312418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,16,balanced,0.2639146645863851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,16,balanced,0.2934933304786682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,16,balanced,0.6315786838531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,16,balanced,0.33953599135080975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,16,balanced,0.3437439997990926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,16,balanced,0.4128799835840861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,16,balanced,0.616373340288798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,16,balanced,0.7522986729939779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,16,balanced,0.45918933550516766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,16,balanced,1.3552427291870117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,16,balanced,0.7812053362528483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,16,power_law_1.01,0.15656319856643677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,16,power_law_1.01,0.16826879978179932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,16,balanced,0.5796746810277303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,16,power_law_1.01,0.08595839738845826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,16,power_law_1.01,0.10486400127410889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,16,balanced,0.8666133085886637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,16,power_law_1.01,0.12558720111846924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,16,balanced,1.1534453233083088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,16,power_law_1.01,0.1625663995742798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,16,power_law_1.01,0.16509439945220947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,16,power_law_1.01,0.17500799894332886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,16,power_law_1.01,0.164409601688385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,16,power_law_1.01,0.171942400932312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,16,power_law_1.01,0.17770880460739136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,16,power_law_1.01,0.09792640209197997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,16,power_law_1.01,0.18109439611434935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,16,power_law_1.01,0.1823232054710388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,16,power_law_1.01,0.10533119440078735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,16,power_law_1.01,0.18647040128707887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,16,power_law_1.01,0.061561602354049685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,16,power_law_1.01,0.20581119060516356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,16,power_law_1.01,0.07745280265808105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,16,power_law_1.01,0.08494719862937927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,16,power_law_1.01,0.19992320537567138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,16,power_law_1.01,0.10704640150070191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,16,power_law_1.01,0.22746880054473878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,16,balanced,0.6837333043416342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,16,power_law_1.01,0.11274240016937256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,16,power_law_1.01,0.26177918910980225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,16,balanced,1.5131413141886394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,16,power_law_1.01,0.10398720502853394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,16,power_law_1.01,0.2610687971115112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,16,power_law_1.01,0.11160320043563843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,16,power_law_1.01,0.3398144006729126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,16,power_law_1.01,0.11493120193481446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,16,power_law_1.01,0.3439167976379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,16,power_law_1.01,0.11809920072555542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,16,power_law_1.01,0.4273087978363037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,16,power_law_1.01,0.12005759477615356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,16,power_law_1.01,0.46149120330810545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,16,power_law_1.01,0.1267135977745056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,16,power_law_1.01,0.12220799922943115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,16,power_law_1.01,0.5805247783660888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,16,power_law_1.01,0.136735999584198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,16,power_law_1.01,0.8596159934997558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,16,power_law_1.01,0.14217599630355834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,16,power_law_1.01,1.0846400260925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,16,power_law_1.01,0.14707839488983154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,16,balanced,1.3703306516011555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,16,power_law_1.01,1.3402815818786622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,16,power_law_1.01,0.17909760475158693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,16,power_law_1.01,0.19667199850082398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,16,power_law_1.01,1.7449535369873046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,16,power_law_1.01,0.2594304084777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,16,power_law_1.01,3.013216018676758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,16,power_law_1.01,0.23761920928955077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,16,power_law_1.01,0.30443520545959474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,16,balanced,1.8290026982625325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,16,power_law_1.01,0.3991039991378784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,16,power_law_1.01,5.404415893554687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,16,power_law_1.01,0.5002304077148437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,16,power_law_1.01,0.6157824039459229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,16,power_law_1.01,0.09241600036621093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,16,power_law_1.01,0.9078528404235839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,16,power_law_1.01,0.08872960209846496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,16,power_law_1.01,1.06876802444458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,16,balanced,2.5134506225585938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,16,balanced,1.0521706740061443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,16,power_law_1.01,0.060134398937225345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,16,power_law_1.01,1.4873663902282714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,16,power_law_1.01,0.0692031979560852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,16,power_law_1.01,2.1722944259643553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,16,power_law_1.01,0.07623040080070495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,16,power_law_1.01,0.09096320271492005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,16,power_law_1.01,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,16,power_law_1.01,4.991635131835937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,16,power_law_1.01,0.09532160162925721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,16,power_law_1.01,0.09515519738197327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,16,power_law_1.01,0.09498239755630493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,16,power_law_1.01,0.09568639993667602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,16,power_law_1.01,0.09841279983520508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,16,power_law_1.01,0.09913600087165833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,16,power_law_1.01,0.10241279602050782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,16,power_law_1.01,0.10812159776687622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,16,power_law_1.01,0.10929919481277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,16,power_law_1.01,0.11507840156555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,16,balanced,2.9590559005737305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,16,power_law_1.01,0.1339840054512024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.01,0.14782079458236694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,16,power_law_1.01,0.05058559775352478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.01,0.1728384017944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.01,0.18408960103988647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,16,power_law_1.01,0.053958398103713986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.01,0.24493439197540284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,16,power_law_1.01,0.039129599928855896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.01,0.3059776067733765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,16,power_law_1.01,0.04595839977264404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.01,0.40353918075561523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,16,power_law_1.01,0.050316798686981204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.01,0.481279993057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,16,power_law_1.01,0.06672000288963317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,16,power_law_1.01,0.08673279881477355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.01,0.6688447952270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,16,power_law_1.01,0.08098559975624084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.01,0.8618623733520507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,16,power_law_1.01,0.07697280049324036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,16,power_law_1.01,1.1170751571655273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,16,power_law_1.01,0.11143679618835449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,16,balanced,2.6117226282755532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,16,power_law_1.01,0.11335680484771729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,16,power_law_1.01,1.7182016372680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,16,power_law_1.01,0.11451519727706909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,16,power_law_1.01,0.14084479808807374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,16,power_law_1.01,3.193471908569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,16,power_law_1.01,0.1310528039932251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,16,power_law_1.01,0.13264000415802002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,16,power_law_1.01,0.1283519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,16,power_law_1.01,0.13271039724349976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,16,power_law_1.01,0.11349120140075683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,16,balanced,2.0038933753967285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.01,0.11907199621200562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.01,0.13965439796447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.01,0.16856319904327394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.01,0.21854720115661622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.01,0.2617919921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.01,0.3727360010147095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.01,0.44659199714660647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.01,0.6462463855743408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.01,0.9013504028320313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,16,power_law_1.01,1.1366975784301758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,16,power_law_1.01,1.7262527465820312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,16,power_law_1.01,3.701222229003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,16,balanced,5.790597279866536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,16,power_law_1.2,0.15614720582962036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,16,power_law_1.2,0.10959999561309815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,16,power_law_1.2,0.09025279879570007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,16,power_law_1.2,0.10553599596023559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,16,power_law_1.2,0.1168895959854126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,16,power_law_1.2,0.15345920324325563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,16,power_law_1.2,0.09969919919967651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,16,power_law_1.2,0.17039999961853028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,16,power_law_1.2,0.16981760263442994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,16,power_law_1.2,0.06316800117492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,16,power_law_1.2,0.17292159795761108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,16,power_law_1.2,0.06419839859008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,16,power_law_1.2,0.17297279834747314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,16,power_law_1.2,0.08110719919204712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,16,power_law_1.2,0.18293119668960572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,16,power_law_1.2,0.08349440097808838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,16,power_law_1.2,0.1889855980873108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,16,power_law_1.2,0.10497280359268188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,16,power_law_1.2,0.1929471969604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,16,power_law_1.2,0.11115520000457764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,16,power_law_1.2,0.185971200466156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,16,power_law_1.2,0.10926079750061035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,16,power_law_1.2,0.21159679889678956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,16,power_law_1.2,0.11084799766540528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,16,power_law_1.2,0.21611518859863282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,16,power_law_1.2,0.11654399633407593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,16,power_law_1.2,0.09011200070381165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,16,power_law_1.2,0.24285440444946288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,16,power_law_1.2,0.12554880380630493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,16,power_law_1.2,0.2864192008972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,16,power_law_1.2,0.12674560546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,16,power_law_1.2,0.06396160125732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,16,power_law_1.2,0.2955647945404053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,16,power_law_1.2,0.12711679935455322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,16,power_law_1.2,0.056620800495147706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,16,power_law_1.2,0.12992000579833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,16,power_law_1.2,0.3598975896835327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,16,power_law_1.2,0.07189120054244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,16,power_law_1.2,0.1357375979423523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,16,power_law_1.2,0.3814527988433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,16,power_law_1.2,0.14856959581375123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,16,power_law_1.2,0.0722815990447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,16,power_law_1.2,0.47404799461364744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,16,power_law_1.2,0.1573184013366699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,16,power_law_1.2,0.08735359907150268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,16,power_law_1.2,0.1903040051460266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,16,power_law_1.2,0.5472447872161865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,16,power_law_1.2,0.0883840024471283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,16,power_law_1.2,0.2178431987762451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,16,power_law_1.2,0.8474687576293946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,16,power_law_1.2,0.09148160219192505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,16,power_law_1.2,0.26679680347442625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,16,power_law_1.2,0.9350144386291503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,16,power_law_1.2,0.09444479942321778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,16,power_law_1.2,0.2479487895965576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,16,power_law_1.2,0.09390079975128174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,16,power_law_1.2,1.3997440338134766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,16,power_law_1.2,0.34647040367126464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,16,power_law_1.2,0.09637119770050048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,16,power_law_1.2,0.43386240005493165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,16,power_law_1.2,2.0015615463256835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,16,power_law_1.2,0.09631360173225403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,16,power_law_1.2,0.5604095935821534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,16,power_law_1.2,0.09711999893188476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,16,power_law_1.2,2.4257152557373045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,16,power_law_1.2,0.6715199947357178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,16,power_law_1.2,0.1011199951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,16,power_law_1.2,0.9782464027404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,16,power_law_1.2,2.9339712142944334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,16,power_law_1.2,0.10915839672088623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,16,power_law_1.2,1.272812843322754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,16,power_law_1.2,0.11425280570983887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,16,power_law_1.2,1.8565696716308593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,16,power_law_1.2,0.12012159824371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,16,power_law_1.2,7.561023712158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,16,power_law_1.2,0.13603199720382692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,16,power_law_1.2,2.9290176391601563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.2,0.15486079454421997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.2,0.18616960048675538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,16,power_law_1.2,6.605830383300781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.2,0.22342400550842284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.2,0.2609663963317871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,16,power_law_1.2,0.051526397466659546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,8,balanced,0.044079999128977455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.2,0.3529599905014038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,16,power_law_1.2,0.03733119964599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.2,0.4494463920593262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,8,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,16,power_law_1.2,0.04167680144309997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.2,0.6373631954193115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,8,balanced,0.06817600131034851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,16,power_law_1.2,0.04199680089950562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,8,balanced,0.1002506713072459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,16,power_law_1.2,0.047660800814628604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.2,0.8413887977600097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,8,balanced,0.17537599802017212
TRTLLM,1.2.0rc5,NVIDIA B200,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 B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,8,balanced,0.18120533227920532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,16,power_law_1.2,0.08454399704933166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.2,1.2804160118103027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,8,balanced,0.18222399552663168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,16,power_law_1.2,0.08076800107955932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,8,balanced,0.18404799699783325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,8,balanced,0.04370133578777313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,16,power_law_1.2,1.3099455833435059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,8,balanced,0.18492267529169717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,16,power_law_1.2,0.07388160228729249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,8,balanced,0.18362667163213095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,16,power_law_1.2,0.11527680158615113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,8,balanced,0.04574933151404063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,16,power_law_1.2,2.371232032775879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,8,balanced,0.18640534083048502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,16,power_law_1.2,0.11291520595550537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,8,balanced,0.04975466430187225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,8,balanced,0.18860799074172974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,8,balanced,0.07462933162848155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,8,balanced,0.1890986760457357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,16,power_law_1.2,0.11376639604568481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,16,power_law_1.2,5.297433471679687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,8,balanced,0.10686933000882466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,8,balanced,0.18920000394185385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,16,power_law_1.2,0.12754559516906738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,8,balanced,0.11739733815193176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,16,power_law_1.2,0.12298879623413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,8,balanced,0.19447465737660727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,8,balanced,0.12006933490435283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,16,power_law_1.2,0.11907199621200562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,8,balanced,0.1949333349863688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,8,balanced,0.11940800150235494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,16,power_law_1.2,0.11726080179214478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,16,power_law_1.2,0.10870399475097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,8,balanced,0.12141866485277812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,8,balanced,0.19954667488733926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,16,power_law_1.2,0.1097983956336975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,8,balanced,0.12123200297355652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.2,0.13400319814682007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,8,balanced,0.12418133020401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,8,balanced,0.2180160085360209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.2,0.14513920545578002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,8,balanced,0.12416533629099528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.2,0.17270400524139404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,8,balanced,0.12357333302497864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,8,balanced,0.2156426707903544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.2,0.2184959888458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,8,balanced,0.1281599998474121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.2,0.31588480472564695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.2,0.3825664043426514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,8,balanced,0.25382934014002484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,8,balanced,0.05596266686916351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,8,balanced,0.1337546706199646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.2,0.5337024211883545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,8,balanced,0.13750400145848593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,8,balanced,0.24231467644373575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,8,balanced,0.06243733565012614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.2,0.7681920051574707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,8,balanced,0.025311999022960663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,8,balanced,0.06924800078074138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,8,balanced,0.14126400152842203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,8,balanced,0.0937439997990926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.2,1.1067263603210449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,8,balanced,0.026816000541051228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,8,balanced,0.13577066858609518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,8,balanced,0.15827733278274536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,8,balanced,0.03507733345031738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,8,balanced,0.151418666044871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,8,balanced,0.33032000064849854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,16,power_law_1.2,1.3236479759216309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,8,balanced,0.16107199589411417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,8,balanced,0.1602186659971873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,8,balanced,0.043893332282702126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,8,balanced,0.15064533551534018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,8,balanced,0.05590933561325073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,16,power_law_1.2,2.058131217956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,8,balanced,0.15075199802716574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,8,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,8,balanced,0.1954773267110189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,8,balanced,0.13984533150990805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,16,power_law_1.2,3.9743167877197267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,8,balanced,0.3176106611887614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,8,balanced,0.14006933569908142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,8,balanced,0.06750933329264323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,8,balanced,0.20870399475097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,8,balanced,0.14812800288200378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,8,balanced,0.07216000060240428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,8,balanced,0.1467359960079193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,8,balanced,0.26764800151189166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,8,balanced,0.07783466577529907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,8,balanced,0.14891200264294943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,8,balanced,0.4689226547876994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,8,balanced,0.08334400256474812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,8,balanced,0.1569439967473348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,8,balanced,0.07997333506743114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,8,balanced,0.08345599969228108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,8,balanced,0.15592533349990845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,8,balanced,0.30270934104919434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,8,balanced,0.08688533306121826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,8,balanced,0.15955199797948202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,8,balanced,0.094842662413915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,8,balanced,0.174234668413798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,8,balanced,0.10271466771761577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,8,balanced,0.16595733165740967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,8,balanced,0.481056014696757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,8,balanced,0.11315733194351196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,8,balanced,0.42353065808614093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,8,balanced,0.17226133743921915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,8,balanced,0.1750239928563436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,8,balanced,0.1855199933052063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,8,balanced,0.21005332469940186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,8,balanced,0.2434879938761393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,8,balanced,0.280239999294281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,8,balanced,0.32683199644088745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,8,balanced,0.5002880096435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,8,balanced,0.7383093039194742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,8,balanced,0.29416000843048096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,8,balanced,0.21358933051427206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,8,balanced,0.25891733169555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,8,balanced,0.4129759867986043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,8,balanced,0.29362666606903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,8,balanced,0.6924053033192953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,8,balanced,0.5428800185521444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,8,balanced,0.43931734561920166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,8,balanced,0.8277920087178549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,8,balanced,0.5292213360468546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,8,balanced,0.8874719937642416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,8,balanced,0.6586346626281738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,8,balanced,0.7778986295064291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,8,balanced,0.9080800215403239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,8,balanced,0.8946293195088705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,8,balanced,1.0326080322265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,8,balanced,1.0157333215077717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,8,balanced,1.0148746967315674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,8,balanced,1.4616319338480632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,8,balanced,1.1934026877085369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,8,balanced,1.6272640228271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,8,power_law_1.01,0.07048320174217224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,8,power_law_1.01,0.1003648042678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,8,power_law_1.01,0.07568640112876893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,8,balanced,1.5826880137125652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,8,power_law_1.01,0.12223360538482667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,8,power_law_1.01,0.061913597583770755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,8,power_law_1.01,0.07571840286254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,8,power_law_1.01,0.07899519801139832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,8,power_law_1.01,0.08333439826965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,8,power_law_1.01,0.10450559854507446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,8,power_law_1.01,0.10534399747848511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,8,power_law_1.01,0.11185920238494873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,8,power_law_1.01,0.1121216058731079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,8,power_law_1.01,0.14766080379486085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,8,power_law_1.01,0.11195520162582398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,8,power_law_1.01,0.1670912027359009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,8,power_law_1.01,0.11112320423126221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,8,power_law_1.01,0.17097599506378175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,8,power_law_1.01,0.10060160160064698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,8,power_law_1.01,0.1153216004371643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,8,power_law_1.01,0.17260799407958985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,8,power_law_1.01,0.11559679508209228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,8,balanced,1.9453493754069011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,8,power_law_1.01,0.1736448049545288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,8,power_law_1.01,0.11870720386505126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,8,power_law_1.01,0.1189120054244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,8,power_law_1.01,0.1729920029640198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,8,power_law_1.01,0.071424001455307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,8,power_law_1.01,0.12298239469528198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,8,power_law_1.01,0.1750272035598755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,8,power_law_1.01,0.127948796749115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,8,power_law_1.01,0.09899520277976989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,8,power_law_1.01,0.17985919713974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,8,power_law_1.01,0.13653119802474975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,8,power_law_1.01,0.10660480260848999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,8,power_law_1.01,0.186572802066803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,8,power_law_1.01,0.145632004737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,8,power_law_1.01,0.13185919523239137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,8,power_law_1.01,0.1970687985420227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,8,power_law_1.01,0.14716800451278686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,8,power_law_1.01,0.13610880374908446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,8,power_law_1.01,0.2037503957748413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,8,power_law_1.01,0.17470719814300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,8,power_law_1.01,0.14382719993591309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,8,power_law_1.01,0.17985279560089112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,8,power_law_1.01,0.21198720932006837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,8,power_law_1.01,0.1435647964477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,8,power_law_1.01,0.21699841022491456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,8,power_law_1.01,0.2352447986602783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,8,power_law_1.01,0.23915519714355468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,8,power_law_1.01,0.14288640022277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,8,power_law_1.01,0.23392000198364257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,8,power_law_1.01,0.3076159954071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,8,power_law_1.01,0.14521600008010865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,8,power_law_1.01,0.28399360179901123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,8,power_law_1.01,0.37564160823822024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,8,power_law_1.01,0.14926719665527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,8,power_law_1.01,0.3077183961868286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,8,power_law_1.01,0.5001279830932617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,8,power_law_1.01,0.14402559995651246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,8,power_law_1.01,0.38302719593048096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,8,power_law_1.01,0.636191987991333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,8,power_law_1.01,0.1441472053527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,8,power_law_1.01,0.43091840744018556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,8,power_law_1.01,0.92806396484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,8,power_law_1.01,0.14802559614181518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,8,power_law_1.01,0.5559167861938477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,8,power_law_1.01,1.1700287818908692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,8,power_law_1.01,0.15436160564422607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,8,power_law_1.01,0.6441792011260986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,8,power_law_1.01,0.16287360191345215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,8,power_law_1.01,1.5688128471374512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,8,power_law_1.01,0.1828927993774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,8,power_law_1.01,0.9654080390930175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,8,power_law_1.01,2.146873664855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.01,0.19447040557861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,8,power_law_1.01,1.155795192718506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.01,0.23067519664764405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,8,power_law_1.01,4.485081481933594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,8,power_law_1.01,1.4154111862182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.01,0.25418241024017335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,8,power_law_1.01,0.03550080060958862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,8,balanced,2.7319412231445312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.01,0.32617599964141847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,8,power_law_1.01,1.9565887451171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.01,0.40020480155944826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,8,power_law_1.01,0.04625279903411865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.01,0.5457151889801025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,8,power_law_1.01,0.02898559868335724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,8,power_law_1.01,4.443891143798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.01,0.7003967761993408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,8,power_law_1.01,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.01,0.9428031921386719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,8,power_law_1.01,0.04622719883918762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.01,1.1827391624450683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,8,power_law_1.01,0.05716480016708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,8,balanced,3.133397420247396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,8,power_law_1.01,1.4866687774658203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,8,power_law_1.01,0.057120001316070555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,8,power_law_1.01,2.0778560638427734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,8,power_law_1.01,0.06109439730644226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,8,power_law_1.01,0.06216319799423218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,8,power_law_1.01,4.254912185668945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,8,power_law_1.01,0.06627839803695679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,8,power_law_1.01,0.06260480284690857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,8,balanced,3.10424009958903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,8,power_law_1.01,0.06546559929847717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,8,power_law_1.01,0.07025920152664185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,8,power_law_1.01,0.07933440208435058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,8,power_law_1.01,0.08287360072135926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,8,power_law_1.01,0.09166719913482665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,8,power_law_1.01,0.13400319814682007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,8,power_law_1.01,0.16108800172805787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,8,balanced,3.766208012898763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.01,0.1800447940826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.01,0.22777600288391114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.01,0.14154239892959594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.01,0.18498560190200805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.01,0.208076810836792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.01,0.2949696063995361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.01,0.38356480598449705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.01,0.5329855918884278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.01,0.6391615867614746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,8,power_law_1.01,0.804319953918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,8,power_law_1.01,1.2250176429748536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,8,power_law_1.01,2.3637567520141602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,8,power_law_1.2,0.10001280307769775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,8,power_law_1.2,0.06995840072631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,8,power_law_1.2,0.0778880000114441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,8,power_law_1.2,0.06709759831428527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,8,power_law_1.2,0.05380480289459229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,8,power_law_1.2,0.09320319890975952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,8,power_law_1.2,0.058220797777175905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,8,power_law_1.2,0.10104320049285889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,8,power_law_1.2,0.06913279891014099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,8,power_law_1.2,0.16136319637298585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,8,power_law_1.2,0.08496639728546143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,8,power_law_1.2,0.15736960172653197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,8,power_law_1.2,0.16997120380401612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,8,power_law_1.2,0.10343680381774903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,8,power_law_1.2,0.17029759883880616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,8,power_law_1.2,0.10980479717254639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,8,power_law_1.2,0.177183997631073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,8,power_law_1.2,0.11304960250854493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,8,power_law_1.2,0.17921279668807982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,8,power_law_1.2,0.11398400068283081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,8,power_law_1.2,0.186572802066803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,8,power_law_1.2,0.12014720439910889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,8,power_law_1.2,0.1893952012062073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,8,power_law_1.2,0.12097280025482178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,8,power_law_1.2,0.19300479888916017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,8,power_law_1.2,0.12190079689025879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,8,power_law_1.2,0.10257920026779174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,8,power_law_1.2,0.03276160061359405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,8,power_law_1.2,0.19352960586547852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,8,power_law_1.2,0.12488319873809814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,8,power_law_1.2,0.20953600406646727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,4,balanced,0.046256000796953835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,8,power_law_1.2,0.133241605758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,8,power_law_1.2,0.07029119729995728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,8,power_law_1.2,0.025574401021003723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,8,power_law_1.2,0.22464001178741455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,8,power_law_1.2,0.14160000085830687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,8,power_law_1.2,0.07364479899406433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,8,power_law_1.2,0.2552896022796631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,8,power_law_1.2,0.03049600124359131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,8,power_law_1.2,0.15003520250320435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,4,balanced,0.051829333106676735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,8,power_law_1.2,0.08958719968795777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,4,balanced,0.05771199862162272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,8,power_law_1.2,0.03219200074672699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,8,power_law_1.2,0.2603264093399048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,8,power_law_1.2,0.1590656042098999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,4,balanced,0.043882668018341064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,8,power_law_1.2,0.09708799719810486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,8,power_law_1.2,0.04223999977111816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,8,power_law_1.2,0.30649600028991697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,4,balanced,0.07096000015735626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,8,power_law_1.2,0.1874943971633911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,4,balanced,0.06249066690603892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,8,power_law_1.2,0.13420159816741944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,8,power_law_1.2,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,8,power_law_1.2,0.33644158840179444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,4,balanced,0.10425600409507751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,8,power_law_1.2,0.1927232027053833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,4,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,4,balanced,0.07459733386834462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,8,power_law_1.2,0.14071680307388307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,8,power_law_1.2,0.42250881195068357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,8,power_law_1.2,0.24579839706420897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,4,balanced,0.15692800283432007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,8,power_law_1.2,0.055795198678970336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,4,balanced,0.05184000233809153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,4,balanced,0.104010671377182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,8,power_law_1.2,0.1471935987472534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,8,power_law_1.2,0.2684031963348389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,8,power_law_1.2,0.47114877700805663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,4,balanced,0.18253332376480103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,8,power_law_1.2,0.3453376054763794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,4,balanced,0.18269334236780801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,4,balanced,0.15038399895032248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,8,power_law_1.2,0.148089599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,8,power_law_1.2,0.5776895999908447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,4,balanced,0.07730133334795634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,8,power_law_1.2,0.06110720038414001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,4,balanced,0.18436266978581747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,8,power_law_1.2,0.41736321449279784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,4,balanced,0.16497600078582764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,8,power_law_1.2,0.7777535915374756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,8,power_law_1.2,0.14200960397720336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,4,balanced,0.10931733250617981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,4,balanced,0.18788800636927286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,8,power_law_1.2,0.06391680240631104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,4,balanced,0.1636319955190023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,8,power_law_1.2,0.6062079906463623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,8,power_law_1.2,1.026252841949463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,4,balanced,0.18682666619618735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,8,power_law_1.2,0.1419327974319458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,4,balanced,0.12186666329701741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,4,balanced,0.16201600432395935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,8,power_law_1.2,0.6990208148956298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,8,power_law_1.2,1.4546303749084473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,8,power_law_1.2,0.06206079721450806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,4,balanced,0.1865760087966919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,8,power_law_1.2,0.14926719665527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,4,balanced,0.12452266613642375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,4,balanced,0.16426133116086325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,8,power_law_1.2,0.9760319709777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,4,balanced,0.18890132506688437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,8,power_law_1.2,1.795404815673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,8,power_law_1.2,0.06223359704017639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,4,balanced,0.1265120009581248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,8,power_law_1.2,0.1527168035507202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,4,balanced,0.023738667368888855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,4,balanced,0.12659733494122824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,8,power_law_1.2,0.15285760164260864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,8,power_law_1.2,1.4116928100585937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,4,balanced,0.1590986649195353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,8,power_law_1.2,2.861420822143555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,8,power_law_1.2,0.06682239770889283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,4,balanced,0.1898826758066813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,4,balanced,0.127210666735967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,8,power_law_1.2,0.1602560043334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,8,power_law_1.2,1.6718656539916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,4,balanced,0.16263999541600546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,8,power_law_1.2,0.06971520185470581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,4,balanced,0.127893328666687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,4,balanced,0.1916053295135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,8,power_law_1.2,0.1566208004951477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,4,balanced,0.025279998779296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,8,power_law_1.2,4.798515319824219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,4,balanced,0.13126933574676514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,4,balanced,0.1605226695537567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,8,power_law_1.2,2.7993728637695314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,8,power_law_1.2,0.07954559922218322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,4,balanced,0.02847466617822647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,8,power_law_1.2,0.17088639736175537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,4,balanced,0.19924267133076987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,4,balanced,0.15788267056147257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,4,balanced,0.13165866335233053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,8,power_law_1.2,0.0838591992855072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,8,power_law_1.2,5.500825500488281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,4,balanced,0.1585706671079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,4,balanced,0.041536000867684685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,8,power_law_1.2,0.1876031994819641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,4,balanced,0.20185067256291708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,4,balanced,0.13640532890955606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,4,balanced,0.056133334835370384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.2,0.20932478904724122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,4,balanced,0.16767466068267822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,8,power_law_1.2,0.08975359797477722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,4,balanced,0.14268267154693604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,4,balanced,0.20611733198165894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.2,0.24246399402618407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,4,balanced,0.06400000055631001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,4,balanced,0.16512533028920492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,8,power_law_1.2,0.1407807946205139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.2,0.26501760482788084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,4,balanced,0.14631999532381693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,4,balanced,0.22489599386850992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,4,balanced,0.06700799862543742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,4,balanced,0.17058134078979492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,8,power_law_1.2,0.16532479524612426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,4,balanced,0.06849066913127899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.2,0.3489023923873901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,4,balanced,0.07047999898592631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,4,balanced,0.15566399693489075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,4,balanced,0.2285226583480835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,4,balanced,0.07600000003973643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.2,0.19594240188598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.2,0.435968017578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,4,balanced,0.07788800199826558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,4,balanced,0.18720000982284546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.2,0.25425920486450193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,4,balanced,0.17160000403722128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,4,balanced,0.08346666892369588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,4,balanced,0.2720853288968404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.2,0.5452415943145752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.2,0.1579519987106323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,4,balanced,0.1893600026766459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,4,balanced,0.18041600783665976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,4,balanced,0.08835200468699138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,4,balanced,0.2752426664034526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.2,0.6718592166900634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,4,balanced,0.08649067083994548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.2,0.193503999710083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,4,balanced,0.21317332983016968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,4,balanced,0.22210667530695596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.2,1.0034048080444335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,4,balanced,0.34564801057179767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,4,balanced,0.09520533680915833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.2,0.23274240493774415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,4,balanced,0.10217600067456563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,4,balanced,0.23520533243815103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.2,0.31560320854187013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.2,1.5161472320556642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,4,balanced,0.10963199536005656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.2,0.4074751853942871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,4,balanced,0.24269866943359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,4,balanced,0.11834667126337688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.2,0.5673088073730469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,8,power_law_1.2,1.9000511169433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,4,balanced,0.3770453135172526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,4,balanced,0.3442986806233724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,4,balanced,0.19527467091878256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.2,0.7591551780700684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,8,power_law_1.2,2.642259216308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,4,balanced,0.3179146647453308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,4,balanced,0.234607994556427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,8,power_law_1.2,0.9248640060424804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,4,balanced,0.3838719924290975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,8,power_law_1.2,1.4691776275634765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,4,balanced,0.19842666387557983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,8,power_law_1.2,7.473926544189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,4,balanced,0.5244799852371216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,4,balanced,0.37299199899037677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,8,power_law_1.2,3.2064640045166017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,4,balanced,0.22825600703557333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,4,balanced,0.5291733344395956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,4,balanced,0.2588319977124532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,4,balanced,0.6134293476740519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,4,balanced,0.5221279859542847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,4,balanced,0.6894026597340902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,4,balanced,0.3899253209431966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,4,balanced,0.6441653172175089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,4,balanced,0.8888800144195557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,4,balanced,0.9026666482289633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,4,balanced,0.4585440158843994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,4,balanced,0.9009280204772949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,4,balanced,1.0671412944793701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,4,balanced,0.6545173327128092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,4,balanced,1.2121439774831135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,4,balanced,1.1652106444040935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,4,balanced,0.8618559837341309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,4,balanced,1.2321813106536865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,4,balanced,1.4199573198954265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,4,balanced,1.392133394877116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,4,balanced,0.991434653600057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,4,balanced,1.9477653503417969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,4,balanced,2.223749319712321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,4,balanced,2.179738680521647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,4,power_law_1.01,0.07249280214309692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,4,power_law_1.01,0.08125439882278443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,4,power_law_1.01,0.06907520294189454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,4,power_law_1.01,0.08929920196533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,4,power_law_1.01,0.10458240509033204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,4,power_law_1.01,0.15322240591049194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,4,balanced,1.604149341583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,4,power_law_1.01,0.1622912049293518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,4,power_law_1.01,0.15164799690246583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,4,power_law_1.01,0.16625920534133912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,4,power_law_1.01,0.1690111994743347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,4,power_law_1.01,0.1746240019798279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,4,power_law_1.01,0.1735360026359558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,4,power_law_1.01,0.18128000497817992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,4,power_law_1.01,0.05120000243186951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,4,power_law_1.01,0.18588160276412963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,4,power_law_1.01,0.19481600522994996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,4,power_law_1.01,0.06072319746017456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,4,power_law_1.01,0.1961856007575989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,4,power_law_1.01,0.05690240263938904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,4,power_law_1.01,0.21188480854034425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,4,power_law_1.01,0.0686847984790802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,4,power_law_1.01,0.23751680850982665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,4,power_law_1.01,0.08673279881477355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,4,power_law_1.01,0.24594559669494628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,4,power_law_1.01,0.10970239639282227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,4,power_law_1.01,0.2897088050842285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,4,power_law_1.01,0.10868480205535888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,4,power_law_1.01,0.32497920989990237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,4,power_law_1.01,0.11895040273666382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,4,power_law_1.01,0.3917311906814575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,4,power_law_1.01,0.12033920288085938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,4,power_law_1.01,0.4716800212860107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,4,power_law_1.01,0.11834239959716797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,4,power_law_1.01,0.6379648208618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,4,power_law_1.01,0.12087680101394653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,4,power_law_1.01,0.7901055812835693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,4,power_law_1.01,0.12272640466690063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,4,power_law_1.01,0.12669440507888793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,4,power_law_1.01,1.0462911605834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,4,power_law_1.01,0.13171839714050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,4,power_law_1.01,1.3239999771118165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,4,power_law_1.01,0.1431488037109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,4,power_law_1.01,0.1458624005317688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,4,power_law_1.01,1.6744575500488281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,4,power_law_1.01,0.1559167981147766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,4,power_law_1.01,2.451686477661133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,4,power_law_1.01,0.18933759927749633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,4,power_law_1.01,4.803449630737305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,4,power_law_1.01,0.20142080783843994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,4,balanced,3.694256146748861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,4,balanced,4.384991963704427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,4,power_law_1.01,0.23895039558410644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,4,power_law_1.01,0.2700095891952515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,4,power_law_1.01,0.35666561126708984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,4,power_law_1.01,0.45218558311462403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,4,power_law_1.01,0.593721580505371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,4,power_law_1.01,0.783238410949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,4,power_law_1.01,1.0580863952636719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,4,balanced,4.235648155212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,4,power_law_1.01,1.4210432052612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,4,power_law_1.01,1.6838592529296874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,4,power_law_1.01,2.6778688430786133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,4,power_law_1.01,0.022988800704479218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,4,power_law_1.01,0.07983999848365783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,4,balanced,3.0990187327067056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,4,power_law_1.01,5.294208145141601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,4,power_law_1.01,0.028832000494003297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,4,power_law_1.01,0.08264960050582885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,4,power_law_1.01,0.026291200518608095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,4,power_law_1.01,0.07385600209236146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,4,power_law_1.01,0.03816959857940674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,4,power_law_1.01,0.0980288028717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,4,power_law_1.01,0.1078336000442505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,4,power_law_1.01,0.042931199073791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,4,power_law_1.01,0.14663039445877074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,4,power_law_1.01,0.05495039820671081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,4,power_law_1.01,0.1506432056427002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,4,power_law_1.01,0.05745279788970947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,4,power_law_1.01,0.1488960027694702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,4,power_law_1.01,0.058348798751831056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,4,power_law_1.01,0.15274879932403565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,4,power_law_1.01,0.0608959972858429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,4,power_law_1.01,0.15230720043182372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,4,power_law_1.01,0.06298879981040954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,4,power_law_1.01,0.15576319694519042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,4,power_law_1.01,0.06503679752349853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,4,power_law_1.01,0.06756479740142822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,4,power_law_1.01,0.15510400533676147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,4,power_law_1.01,0.07310079932212829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,4,power_law_1.01,0.16298240423202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,4,power_law_1.01,0.07011839747428894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,4,power_law_1.01,0.16561919450759888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,4,power_law_1.01,0.07957119941711426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,4,power_law_1.01,0.16866559982299806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,4,power_law_1.01,0.08213120102882385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,4,power_law_1.01,0.16903040409088135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,4,power_law_1.01,0.08101119995117187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,4,power_law_1.01,0.18074239492416383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,4,power_law_1.01,0.09452800154685974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,4,power_law_1.01,0.19816960096359254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.01,0.16904959678649903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.01,0.2183295965194702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.01,0.20433919429779052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.01,0.2489151954650879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.01,0.1536064028739929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.01,0.3081536054611206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.01,0.17928320169448853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.01,0.38585600852966306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.01,0.2092223882675171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.01,0.2796288013458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.01,0.45532798767089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.01,0.598361587524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.01,0.3444159984588623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.01,0.8333439826965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.01,0.4866176128387451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.01,0.6353792190551758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.01,1.1457280158996581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,4,power_law_1.01,0.7074048042297363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.01,1.4123071670532226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,4,power_law_1.01,1.2295104026794434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,4,power_law_1.01,1.7661632537841796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,4,power_law_1.01,2.8516351699829103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,4,power_law_1.01,2.364633560180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,4,power_law_1.01,5.320671844482422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,4,power_law_1.2,0.07311999797821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,4,power_law_1.2,0.04814079999923706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,4,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,4,power_law_1.2,0.07970560193061829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,4,power_law_1.2,0.06798080205917359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,4,power_law_1.2,0.04728319942951202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,4,power_law_1.2,0.08524159789085388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,4,power_law_1.2,0.050310397148132326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,4,power_law_1.2,0.06367999911308289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,4,power_law_1.2,0.10399359464645386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,4,power_law_1.2,0.06753919720649719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,4,power_law_1.2,0.06772480010986329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,4,power_law_1.2,0.15735039710998536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,4,power_law_1.2,0.023244799673557283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,4,power_law_1.2,0.07678719758987426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,4,power_law_1.2,0.09187840223312378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,4,power_law_1.2,0.16078720092773438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,4,power_law_1.2,0.10289280414581299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,4,power_law_1.2,0.10895999670028686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,4,power_law_1.2,0.16076159477233887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,4,power_law_1.2,0.023046399652957916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,4,power_law_1.2,0.10998400449752807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,4,power_law_1.2,0.13941119909286498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,4,power_law_1.2,0.1689471960067749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,4,power_law_1.2,0.025811201333999632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,4,power_law_1.2,0.11218559741973877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,4,power_law_1.2,0.15411839485168458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,4,power_law_1.2,0.17233920097351074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,4,power_law_1.2,0.03351039886474609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,4,power_law_1.2,0.11568640470504761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,4,power_law_1.2,0.15560959577560424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,4,power_law_1.2,0.17601280212402343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,4,power_law_1.2,0.1189695954322815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,4,power_law_1.2,0.15644160509109498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,4,power_law_1.2,0.04268800020217896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,4,power_law_1.2,0.18082560300827027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,4,power_law_1.2,0.12695679664611817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,4,power_law_1.2,0.15953279733657838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,4,power_law_1.2,0.055052798986434934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,4,power_law_1.2,0.18821760416030883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,4,power_law_1.2,0.1279039978981018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,4,power_law_1.2,0.19333759546279908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,4,power_law_1.2,0.056543999910354616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,4,power_law_1.2,0.16405760049819945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,4,power_law_1.2,0.20488319396972657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,4,power_law_1.2,0.12957439422607422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,4,power_law_1.2,0.05852159857749939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,4,power_law_1.2,0.16348799467086791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,4,power_law_1.2,0.2197632074356079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,4,power_law_1.2,0.1651520013809204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,4,power_law_1.2,0.13757439851760864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,4,power_law_1.2,0.061414402723312375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,4,power_law_1.2,0.2151423931121826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,4,power_law_1.2,0.16761599779129027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,4,power_law_1.2,0.14343039989471434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,4,power_law_1.2,0.06297600269317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,4,power_law_1.2,0.2597440004348755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,4,power_law_1.2,0.16640640497207643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,4,power_law_1.2,0.14984960556030275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,4,power_law_1.2,0.0658240020275116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,4,power_law_1.2,0.2563584089279175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,4,power_law_1.2,0.17473920583724975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,4,power_law_1.2,0.1645311951637268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,4,power_law_1.2,0.07008640170097351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,4,power_law_1.2,0.31189761161804197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,4,power_law_1.2,0.1918272018432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,2,balanced,0.04628799855709076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,4,power_law_1.2,0.18647040128707887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,4,power_law_1.2,0.07474560141563416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,2,balanced,0.04572799801826477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,4,power_law_1.2,0.34045441150665284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,4,power_law_1.2,0.20785279273986818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,4,power_law_1.2,0.1994752049446106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,4,power_law_1.2,0.07367680072784424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,4,power_law_1.2,0.4187967777252197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.2,0.23069438934326172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,2,balanced,0.053472002347310386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,4,power_law_1.2,0.24961280822753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,4,power_law_1.2,0.07866880297660828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,2,balanced,0.04839999973773956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,4,power_law_1.2,0.5012159824371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.2,0.2751104116439819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,2,balanced,0.07411733269691467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,4,power_law_1.2,0.2839103937149048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,4,power_law_1.2,0.08453760147094727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,2,balanced,0.05648000041643778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.2,0.31791999340057375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,4,power_law_1.2,0.6755263805389404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,2,balanced,0.10527466734250386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,4,power_law_1.2,0.37091200351715087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,2,balanced,0.08353599905967712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,4,power_law_1.2,0.08288000226020813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.2,0.4227712154388428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,4,power_law_1.2,0.9190400123596192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,2,balanced,0.18149866660435995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,2,balanced,0.1169706682364146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.2,0.4741375923156738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,4,power_law_1.2,0.09596800208091735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,4,power_law_1.2,0.4509568214416504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,4,power_law_1.2,1.2156543731689453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.2,0.6852799892425537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,2,balanced,0.18964266777038574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.2,0.18102400302886962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,2,balanced,0.13303466637929282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,4,power_law_1.2,0.6263807773590088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,2,balanced,0.18955733378728232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.2,0.8435327529907226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,4,power_law_1.2,1.5333951950073241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,2,balanced,0.13617066542307535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.2,0.2154304027557373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,4,power_law_1.2,0.8124223709106445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,2,balanced,0.13776000340779623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.2,0.15892479419708253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,4,power_law_1.2,1.1567872047424317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,2,balanced,0.193722665309906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.2,1.188038444519043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,4,power_law_1.2,1.6497087478637695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,2,balanced,0.13938132921854654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,2,balanced,0.1941173275311788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.2,0.19381120204925537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,4,power_law_1.2,1.4140480041503907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,2,balanced,0.14078399538993835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.2,1.7378879547119142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,2,balanced,0.19235199689865112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.2,0.23315200805664063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,2,balanced,0.14415466785430908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,4,power_law_1.2,2.6231679916381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,2,balanced,0.19820266962051392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,4,power_law_1.2,1.858937644958496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,4,power_law_1.2,1.7861440658569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,2,balanced,0.1469439963499705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.2,0.28652799129486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,2,balanced,0.20117332537968954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,2,balanced,0.15039466818173727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,2,power_law_1.01,0.04893440008163452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.2,0.3710783958435059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,2,balanced,0.20253332455952963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,4,power_law_1.2,3.705452728271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,4,power_law_1.2,2.960576057434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,4,power_law_1.2,6.392083358764649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,2,balanced,0.15722666184107462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,2,balanced,0.20227734247843424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.2,0.5182015895843506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,2,balanced,0.1623146633307139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,2,power_law_1.01,0.06298239827156067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,2,balanced,0.2140373388926188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.2,0.6886655807495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,4,power_law_1.2,5.9346881866455075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,2,balanced,0.1692319909731547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,2,balanced,0.21665066480636597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,4,power_law_1.2,6.217919921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,2,power_law_1.01,0.061849600076675414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,4,power_law_1.2,0.8320192337036133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,2,balanced,0.2266026735305786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,2,power_law_1.01,0.08636159896850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,2,balanced,0.17766932646433511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,2,power_law_1.01,0.10774400234222412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,2,balanced,0.2504319945971171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,4,power_law_1.2,1.2439104080200196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,2,balanced,0.20190932353337607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,2,power_law_1.01,0.1538624048233032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,2,balanced,0.21836266915003458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,2,balanced,0.26306132475535077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,4,power_law_1.2,2.587065505981445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,2,power_law_1.01,0.16305279731750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,2,balanced,0.28598932425181073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,2,power_law_1.01,0.16303999423980714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,2,balanced,0.32419200738271076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,2,power_law_1.01,0.16795519590377808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,2,power_law_1.01,0.04464640021324158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,2,balanced,0.3078719973564148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,2,power_law_1.01,0.17240320444107055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,2,balanced,0.34492266178131104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,2,power_law_1.01,0.04902400076389313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,2,power_law_1.01,0.18195199966430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,2,power_law_1.01,0.05195519924163818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,2,balanced,0.4418239990870158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,2,power_law_1.01,0.18724479675292968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,2,balanced,0.49057066440582275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,2,power_law_1.01,0.06857600212097167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,2,power_law_1.01,0.18815360069274903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,2,power_law_1.01,0.08394240140914917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,2,balanced,0.5125600099563599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,2,balanced,0.518618663152059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,2,power_law_1.01,0.11340160369873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,2,power_law_1.01,0.1926975965499878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,2,power_law_1.01,0.11984000205993653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,2,power_law_1.01,0.122707200050354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,2,power_law_1.01,0.206931209564209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,2,balanced,0.7796533107757568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,2,power_law_1.01,0.12556159496307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,2,balanced,0.7918346722920736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,2,power_law_1.01,0.20955519676208495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,2,power_law_1.01,0.12931840419769286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,2,power_law_1.01,0.22502400875091552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,2,power_law_1.01,0.13904639482498168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,2,power_law_1.01,0.25090560913085935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,2,power_law_1.01,0.14583679437637329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,2,power_law_1.01,0.28023040294647217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,2,power_law_1.01,0.1509951949119568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,2,power_law_1.01,0.349945592880249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,2,power_law_1.01,0.15240319967269897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,2,balanced,0.9078719615936279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,2,power_law_1.01,0.16211199760437012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,2,power_law_1.01,0.39740159511566164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,2,balanced,0.8664053281148275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,2,power_law_1.01,0.16639360189437866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,2,power_law_1.01,0.49303040504455564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,2,power_law_1.01,0.17742079496383667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,2,power_law_1.01,0.6151487827301025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,2,power_law_1.01,0.21111679077148438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,2,power_law_1.01,0.8475263595581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,2,power_law_1.01,0.2378943920135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,2,power_law_1.01,1.042739200592041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,2,power_law_1.01,0.29167358875274657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,2,balanced,1.3550453186035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,2,power_law_1.01,1.4122303962707519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,2,power_law_1.01,0.34629120826721194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,2,power_law_1.01,1.7700288772583008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,2,balanced,1.3508426348368328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,2,power_law_1.01,0.43900160789489745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,2,power_law_1.01,2.1707199096679686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,2,power_law_1.01,0.5980800151824951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,2,power_law_1.01,0.8025664329528809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,2,power_law_1.01,3.3139968872070313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,2,power_law_1.01,1.0340991973876954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,2,power_law_1.01,1.426527976989746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,2,power_law_1.01,5.9436992645263675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,2,power_law_1.01,1.8018560409545898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,2,balanced,1.7125226656595867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,2,power_law_1.01,2.35416316986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,2,balanced,1.578437328338623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,2,power_law_1.01,3.459360122680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,2,power_law_1.01,6.6060737609863285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,2,balanced,2.0414719581604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,2,balanced,1.8500480651855469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,2,balanced,3.2098026275634766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,2,balanced,2.9389546712239585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,2,power_law_1.2,0.0499455988407135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,2,power_law_1.2,0.05055999755859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,2,power_law_1.2,0.058924800157546996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,2,power_law_1.2,0.08097919821739197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,2,power_law_1.2,0.043673598766326906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,2,power_law_1.2,0.10160000324249267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,2,power_law_1.2,0.15176960229873657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,2,power_law_1.2,0.04469119906425476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,2,power_law_1.2,0.16463359594345092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,2,power_law_1.2,0.05203840136528015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,2,power_law_1.2,0.1734015941619873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,2,power_law_1.2,0.17405439615249635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,2,power_law_1.2,0.06577919721603394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,2,power_law_1.2,0.1788480043411255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,2,power_law_1.2,0.07968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,2,power_law_1.2,0.18967679738998414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,2,power_law_1.2,0.10739200115203858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,2,power_law_1.2,0.19517439603805542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,2,power_law_1.2,0.11672320365905761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,2,power_law_1.2,0.19560320377349855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,2,power_law_1.2,0.12407679557800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,2,power_law_1.2,0.20371840000152588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,2,power_law_1.2,0.12670079469680787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,2,power_law_1.2,0.2140415906906128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,2,power_law_1.2,0.1312127947807312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,1,balanced,0.04782933493455251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,2,power_law_1.2,0.2156224012374878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,2,power_law_1.2,0.14161920547485352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,2,power_law_1.2,0.23418879508972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,2,power_law_1.2,0.14947839975357055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,1,balanced,0.049925332268079124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,1,balanced,0.05596266686916351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,2,power_law_1.2,0.2623039960861206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,2,balanced,6.241898854573567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,2,power_law_1.2,0.15348479747772217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,1,balanced,0.07602133353551228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,2,power_law_1.2,0.2830976009368896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,1,balanced,0.05399466554323832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,2,power_law_1.2,0.15266560316085814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,1,balanced,0.11339733004570007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,2,power_law_1.2,0.3498239994049072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,1,balanced,0.06371200084686279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,2,power_law_1.2,0.164300799369812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,1,balanced,0.17523733774820963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,1,balanced,0.09397866328557332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,2,power_law_1.2,0.40958080291748045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,2,power_law_1.2,0.16669440269470215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,1,balanced,0.20708266894022623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,1,balanced,0.1415839989980062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,2,balanced,5.653546651204427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,2,power_law_1.2,0.1818943977355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,2,power_law_1.2,0.5344768047332764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,1,balanced,0.21050133307774863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,1,balanced,0.15572266777356467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,2,power_law_1.2,0.21223680973052977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,1,balanced,0.21489600340525308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,2,power_law_1.2,0.6699711799621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,1,balanced,0.16008533040682474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,2,power_law_1.2,0.23673601150512696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,1,balanced,0.21901333332061768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,2,power_law_1.2,0.859769630432129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,1,balanced,0.1633386711279551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,1,balanced,0.21803200244903564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,2,power_law_1.2,0.29857919216156004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,1,balanced,0.16642666856447855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,2,power_law_1.2,1.152665615081787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,1,balanced,0.22166933616002402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,1,balanced,0.16782933473587036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,2,power_law_1.2,0.35600640773773196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,2,power_law_1.2,1.5974207878112794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,1,balanced,0.22823466857274374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,1,balanced,0.1716266671816508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,1,balanced,0.2356320023536682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,2,power_law_1.2,0.46358399391174315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,2,power_law_1.2,1.7828224182128907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,1,balanced,0.1771199901898702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,1,balanced,0.24045334259668985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,2,power_law_1.2,0.6152063846588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,1,balanced,0.18041600783665976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,1,balanced,0.2503359913825989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,2,power_law_1.2,2.4223936080932615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,1,balanced,0.19138665994008383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,2,power_law_1.2,0.8309951782226562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,1,balanced,0.258842666943868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,1,balanced,0.20102399587631226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,2,power_law_1.2,3.627801513671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,2,power_law_1.2,1.0471424102783202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,1,balanced,0.26813334226608276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,1,balanced,0.2112213373184204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,1,balanced,0.22484799226125082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,2,power_law_1.2,1.4025471687316895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,1,balanced,0.31356799602508545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,1,balanced,0.2671519915262858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,2,power_law_1.2,6.892396545410156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,1,balanced,0.3322719931602478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,2,power_law_1.2,1.9949951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,1,balanced,0.29289599259694415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,2,power_law_1.2,2.4923776626586913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,1,balanced,0.44762134552001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,1,balanced,0.416650652885437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,2,power_law_1.2,3.6316734313964845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,1,balanced,0.46672534942626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,1,balanced,0.44538132349650067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,2,power_law_1.2,7.312287902832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,1,balanced,0.7084906895955404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,1,balanced,0.669274648030599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,1,balanced,0.7646346886952718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,1,balanced,0.7714293003082275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,1,balanced,1.207850694656372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,1,balanced,1.1942773660024006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,1,power_law_1.01,0.045638400316238406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,1,power_law_1.01,0.053260797262191774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,1,power_law_1.01,0.04403199851512909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,1,power_law_1.01,0.06778879761695862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,1,balanced,1.3889813423156738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,1,power_law_1.01,0.09329919815063477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,1,power_law_1.01,0.049804800748825075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,1,power_law_1.01,0.11869440078735352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,1,balanced,1.4328373273213704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,1,power_law_1.01,0.16813440322875978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,1,power_law_1.01,0.18009599447250366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,1,power_law_1.01,0.05799040198326111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,1,power_law_1.01,0.1837440013885498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,1,power_law_1.01,0.19249919652938843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,1,power_law_1.01,0.07797759771347046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,1,power_law_1.01,0.19785599708557128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,1,power_law_1.01,0.09565439820289612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,1,power_law_1.01,0.20475521087646484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,1,power_law_1.01,0.1300927996635437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,1,power_law_1.01,0.21992321014404298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,1,balanced,2.2627414067586265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,1,power_law_1.01,0.13701759576797484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,1,power_law_1.01,0.2296191930770874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,1,power_law_1.01,0.22922239303588868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,1,power_law_1.01,0.14483200311660765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,1,balanced,2.266581376393636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,1,power_law_1.01,0.24570879936218262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,1,power_law_1.01,0.14977279901504517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,1,power_law_1.01,0.2523711919784546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,1,power_law_1.01,0.15714559555053711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,1,power_law_1.01,0.27800960540771485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,1,power_law_1.01,0.32760961055755616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,1,power_law_1.01,0.1647744059562683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,1,power_law_1.01,0.3594367980957031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,1,power_law_1.01,0.17386879920959472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,1,power_law_1.01,0.4428607940673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,1,power_law_1.01,0.18362239599227906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,1,power_law_1.01,0.18549760580062866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,1,power_law_1.01,0.5204415798187256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,1,power_law_1.01,0.19820159673690796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,1,power_law_1.01,0.6883200168609619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,1,balanced,2.6209813753763833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,1,power_law_1.01,0.21067519187927247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,1,power_law_1.01,0.8887104034423828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,1,power_law_1.01,0.22388479709625245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,1,balanced,2.7489493687947593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,1,power_law_1.01,0.27455360889434816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,1,power_law_1.01,1.1781248092651366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,1,power_law_1.01,0.32258560657501223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,1,power_law_1.01,1.6036672592163086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,1,power_law_1.01,0.40940160751342775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,1,power_law_1.01,2.1672767639160155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,1,power_law_1.01,0.4883135795593262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,1,power_law_1.01,2.6421247482299806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,1,power_law_1.01,0.6551424026489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,1,power_law_1.01,0.8586175918579102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,1,power_law_1.01,3.222412872314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,1,power_law_1.01,1.1653440475463868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,1,power_law_1.01,5.00311050415039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,1,power_law_1.01,1.6000831604003907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,1,power_law_1.01,2.2075456619262694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,1,power_law_1.01,9.527398681640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,1,power_law_1.01,2.7524991989135743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,1,balanced,3.082255999247233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,1,power_law_1.01,3.371807861328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,1,balanced,3.3176533381144204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,1,power_law_1.01,5.213145446777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,1,power_law_1.01,10.05962905883789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,1,balanced,5.002357482910156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,1,balanced,5.2208906809488935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,1,power_law_1.2,0.04678399860858917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,1,power_law_1.2,0.04676479995250702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,1,power_law_1.2,0.05221760272979736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,1,power_law_1.2,0.06595839858055115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,1,power_law_1.2,0.05100160241127014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,1,power_law_1.2,0.0866047978401184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,1,power_law_1.2,0.058508801460266116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,1,power_law_1.2,0.1144320011138916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,1,power_law_1.2,0.07550079822540283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,1,power_law_1.2,0.16544640064239502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,1,power_law_1.2,0.0943615972995758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,1,power_law_1.2,0.17991679906845093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,1,power_law_1.2,0.1314687967300415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,1,power_law_1.2,0.18613760471343993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,1,power_law_1.2,0.14060159921646118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,1,power_law_1.2,0.19720959663391113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,1,power_law_1.2,0.1471743941307068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,1,power_law_1.2,0.20508160591125488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,1,power_law_1.2,0.1546880006790161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,1,power_law_1.2,0.21398398876190186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,1,power_law_1.2,0.15918079614639283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,1,power_law_1.2,0.22677760124206542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,1,power_law_1.2,0.16818560361862184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,1,power_law_1.2,0.2332672119140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,1,power_law_1.2,0.1811519980430603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,1,power_law_1.2,0.2402559995651245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,16,balanced,0.05173333485921224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,1,power_law_1.2,0.18857599496841432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,1,power_law_1.2,0.2503488063812256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,16,balanced,0.05297600229581197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,1,power_law_1.2,0.26152958869934084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,1,power_law_1.2,0.18973439931869507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,16,balanced,0.06984533369541168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,1,power_law_1.2,0.2814464092254639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,1,power_law_1.2,0.20699520111083985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,16,balanced,0.09737599889437358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,1,power_law_1.2,0.33607680797576905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,1,power_law_1.2,0.21440000534057618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,16,balanced,0.15044800440470377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,1,power_law_1.2,0.2325376033782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,1,power_law_1.2,0.36801919937133787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,16,balanced,0.24622933069864908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,1,power_law_1.2,0.28153600692749026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,1,power_law_1.2,0.4535232067108154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,16,balanced,0.34567999839782715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,1,power_law_1.2,0.33271040916442873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,1,power_law_1.2,0.525273609161377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,1,power_law_1.2,0.42035841941833496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,16,balanced,0.3473759889602661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,1,power_law_1.2,0.688159990310669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,1,balanced,9.510074615478516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,1,power_law_1.2,0.4990975856781006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,16,balanced,0.35417068004608154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,1,power_law_1.2,0.896121597290039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,1,power_law_1.2,0.6610176086425781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,16,balanced,0.34702932834625244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,16,balanced,0.045781334241231285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,1,power_law_1.2,1.1738431930541993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,1,power_law_1.2,0.8714624404907226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,16,balanced,0.34942400455474854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,1,power_law_1.2,1.1745344161987306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,16,balanced,0.045781334241231285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,1,power_law_1.2,1.625971221923828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,16,balanced,0.3524373372395833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,1,power_law_1.2,1.6247871398925782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,16,balanced,0.35337599118550617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,16,balanced,0.04975466430187225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,1,power_law_1.2,2.180806350708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,1,balanced,10.09018643697103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,1,power_law_1.2,2.2175104141235353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,16,balanced,0.3545120159784953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,16,balanced,0.07254933317502339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,1,power_law_1.2,2.6568319320678713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,16,balanced,0.1011946698029836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,16,balanced,0.36025599638621014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,1,power_law_1.2,2.7573759078979494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,16,balanced,0.15104533235232034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,16,balanced,0.36163198947906494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,16,balanced,0.20244799057642618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,1,power_law_1.2,3.3881153106689452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,1,power_law_1.2,3.230188751220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,16,balanced,0.20055466890335083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,16,balanced,0.37001601854960126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,16,balanced,0.20268267393112183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,1,power_law_1.2,5.280294418334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,1,power_law_1.2,4.982495880126953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,16,balanced,0.201690673828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,16,balanced,0.3832000096638997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,16,balanced,0.20397865772247314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,16,balanced,0.2067520022392273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,16,balanced,0.38301865259806317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,1,power_law_1.2,10.084595489501954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,16,balanced,0.20705066124598184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,1,power_law_1.2,9.507360076904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,16,balanced,0.2116426626841227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,16,balanced,0.40113067626953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,16,balanced,0.216048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,16,balanced,0.21983466545740762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,16,balanced,0.420085350672404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,16,balanced,0.2274613380432129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,16,balanced,0.2405973275502523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,16,balanced,0.25009600321451825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,16,balanced,0.45448001225789386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,16,balanced,0.2762826681137085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,16,balanced,0.4938186804453532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,16,balanced,0.3010666569073995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,16,balanced,0.3466879924138387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,16,balanced,0.5516800085703532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,16,balanced,0.3954720099767049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,16,balanced,0.6678133010864258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,16,balanced,0.4838399887084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,16,balanced,0.05026133358478546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,16,balanced,0.051738664507865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,16,balanced,0.6064533392588297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,16,balanced,0.05162666738033295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,16,balanced,0.06031466523806254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,16,balanced,0.086325337489446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,16,balanced,0.7963626384735107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,16,balanced,0.11629333098729451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,16,balanced,0.1476746698220571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,16,balanced,0.14788800477981567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,16,balanced,0.1497760017712911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,16,balanced,0.14801067113876343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,16,balanced,0.15103466312090555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,16,balanced,0.7836746374766032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,16,balanced,0.15176533659299216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,16,balanced,0.15060266852378845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,16,balanced,0.15575466553370157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,16,balanced,0.1621226668357849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,16,balanced,1.0107359886169434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,16,balanced,0.16320533553759256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,16,balanced,0.1707893411318461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,16,balanced,0.17982399463653564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,16,balanced,1.0114239851633708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,16,balanced,0.1868213415145874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,16,balanced,0.20750399430592856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,16,balanced,0.22728532552719116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,16,balanced,0.2628213365872701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,16,balanced,1.2416213353474934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,16,balanced,0.3037173350652059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,16,balanced,1.2401386896769206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,16,power_law_1.01,0.15527679920196533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,16,balanced,0.37620798746744794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,16,power_law_1.01,0.24308478832244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,16,power_law_1.01,0.3389951944351196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,16,power_law_1.01,0.09608960151672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,16,power_law_1.01,0.1418303966522217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,16,power_law_1.01,0.2113856077194214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,16,balanced,0.478277325630188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,16,power_law_1.01,0.23132801055908203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,16,power_law_1.01,0.2552639961242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,16,power_law_1.01,0.24026238918304443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,16,power_law_1.01,0.27993600368499755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,16,power_law_1.01,0.27491838932037355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,16,power_law_1.01,0.3070143938064575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,16,balanced,1.6903093655904133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,16,power_law_1.01,0.3029567956924438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,16,balanced,0.6202880144119263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,16,power_law_1.01,0.30239999294281006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,16,power_law_1.01,0.3115328073501587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,16,power_law_1.01,0.3200704097747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,16,power_law_1.01,0.33641600608825684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,16,balanced,1.8266347249348958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,16,power_law_1.01,0.3602303981781006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,16,power_law_1.01,0.38858880996704104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,16,power_law_1.01,0.4699711799621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,16,power_law_1.01,0.48848638534545896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,16,power_law_1.01,0.575648021697998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,16,power_law_1.01,0.6840640068054199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,16,balanced,0.7945120334625244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,16,power_law_1.01,0.8489727973937988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,16,power_law_1.01,0.8986944198608399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,16,power_law_1.01,1.230726432800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,16,power_law_1.01,1.385638427734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,16,power_law_1.01,1.8601408004760742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,16,power_law_1.01,0.10362240076065063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,16,power_law_1.01,2.5655935287475584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,16,power_law_1.01,0.14317439794540404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,16,power_law_1.01,0.19930239915847778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,16,power_law_1.01,4.737209701538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,16,power_law_1.01,0.07959039807319641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,16,power_law_1.01,0.10067839622497558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,16,power_law_1.01,0.12011519670486451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,16,power_law_1.01,0.13619840145111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,16,balanced,0.9776000181833903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,16,power_law_1.01,0.1595199942588806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,16,power_law_1.01,0.07897599935531616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,16,power_law_1.2,0.15582720041275025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,16,power_law_1.01,0.16302720308303834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,16,power_law_1.01,0.12840319871902467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,16,power_law_1.01,0.162336003780365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,16,power_law_1.2,0.24392321109771728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,16,power_law_1.01,0.1424831986427307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,16,power_law_1.01,0.17552640438079833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,16,power_law_1.2,0.22598400115966796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,16,power_law_1.01,0.07495679855346679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,16,power_law_1.01,0.1788607954978943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,16,power_law_1.01,0.08606079816818238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,16,power_law_1.2,0.12229119539260865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,16,power_law_1.01,0.10412800312042236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,16,power_law_1.01,0.16978559494018555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,16,power_law_1.2,0.15560319423675537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,16,power_law_1.01,0.10606720447540283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,16,power_law_1.01,0.17844480276107788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,16,power_law_1.2,0.19324159622192383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,16,power_law_1.2,0.10659840106964111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,16,power_law_1.01,0.12332799434661865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,16,power_law_1.01,0.1964095950126648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,16,power_law_1.2,0.1988800048828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,16,power_law_1.01,0.12640000581741334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,16,power_law_1.2,0.144652795791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,16,balanced,3.2797492345174155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,16,power_law_1.01,0.12471679449081421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,16,power_law_1.2,0.2426687955856323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,16,power_law_1.2,0.09807360172271729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,16,power_law_1.01,0.19778560400009154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,16,power_law_1.01,0.1313215970993042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,16,power_law_1.2,0.2660991907119751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,16,power_law_1.2,0.07926399707794189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,16,power_law_1.01,0.21246719360351562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,16,power_law_1.01,0.12896000146865844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,16,power_law_1.2,0.25832319259643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,16,power_law_1.2,0.102566397190094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,16,power_law_1.01,0.23549439907073974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,16,power_law_1.01,0.13594239950180054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,16,balanced,1.4686986605326335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,16,power_law_1.2,0.11882239580154419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,16,balanced,3.5533920923868814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,16,power_law_1.2,0.2877952098846436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,16,power_law_1.01,0.13978879451751708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,16,power_law_1.01,0.25257599353790283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,16,power_law_1.2,0.12197120189666748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,16,power_law_1.2,0.29875199794769286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,16,power_law_1.01,0.14689919948577881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,16,power_law_1.01,0.31286399364471434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,16,power_law_1.2,0.14984320402145385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,16,power_law_1.2,0.2762943983078003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,16,power_law_1.01,0.14970879554748534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,16,power_law_1.01,0.3253632068634033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,16,power_law_1.2,0.14483840465545655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,16,power_law_1.2,0.2948352098464966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,16,power_law_1.01,0.3999295949935913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,16,power_law_1.01,0.15663360357284545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,16,power_law_1.2,0.16043519973754883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,16,power_law_1.2,0.32773120403289796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,16,power_law_1.01,0.1755776047706604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,16,power_law_1.01,0.4854464054107666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,16,power_law_1.2,0.17140480279922485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,16,power_law_1.2,0.317740797996521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,16,power_law_1.01,0.19219839572906494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,16,power_law_1.01,0.6490111827850342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,16,power_law_1.2,0.1714367985725403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,16,power_law_1.01,0.22142720222473145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,16,power_law_1.2,0.3382528066635132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,16,power_law_1.2,0.17658239603042603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,16,power_law_1.01,0.2458751916885376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,16,power_law_1.01,0.7182015895843505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,16,power_law_1.2,0.3637824058532715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,16,power_law_1.2,0.18451199531555176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,16,power_law_1.01,0.3134527921676636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,16,power_law_1.01,1.006719970703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,16,power_law_1.2,0.19319679737091064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,16,power_law_1.2,0.4264319896697998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,16,power_law_1.01,0.34264960289001467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,16,power_law_1.01,1.2424768447875976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,16,power_law_1.2,0.20393600463867187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,16,power_law_1.2,0.5323391914367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,16,power_law_1.01,0.46543359756469727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,16,power_law_1.2,0.2222656011581421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,16,power_law_1.2,0.5236671924591064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,16,power_law_1.01,1.5251392364501952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,16,power_law_1.01,0.5805632114410401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,16,power_law_1.2,0.25214080810546874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,16,power_law_1.01,0.7814527988433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,16,power_law_1.2,0.6698431968688965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,16,power_law_1.01,2.3650943756103517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,16,power_law_1.2,0.26528639793395997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,16,power_law_1.01,0.9888704299926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,16,power_law_1.2,0.7506368160247803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,16,power_law_1.2,0.33527679443359376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,16,power_law_1.01,1.2652544021606444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,16,power_law_1.2,1.0746751785278321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,16,power_law_1.2,0.3478847980499268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,16,power_law_1.01,4.579276657104492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,16,power_law_1.01,1.8414527893066406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,16,power_law_1.2,1.026425552368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,16,power_law_1.2,0.4433599948883057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,16,power_law_1.01,3.65032958984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,16,power_law_1.2,1.380179214477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,16,power_law_1.2,0.5387072086334228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,16,power_law_1.2,0.7275519847869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,16,power_law_1.2,1.639244842529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,16,power_law_1.2,0.8211199760437011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,16,power_law_1.2,2.194745635986328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,16,power_law_1.2,1.1318207740783692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,16,power_law_1.2,2.6770368576049806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,16,power_law_1.2,1.4157183647155762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,16,power_law_1.2,1.7629312515258788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,16,power_law_1.2,5.977222442626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,16,power_law_1.2,2.624787139892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,16,power_law_1.2,5.712243270874024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,16,balanced,2.8347253799438477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,16,power_law_1.2,0.07834879755973816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,16,power_law_1.2,0.1259071946144104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,16,power_law_1.2,0.0840831995010376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,16,power_law_1.2,0.07729920148849487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,16,power_law_1.2,0.08602880239486695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,16,power_law_1.2,0.09812480211257935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,16,power_law_1.2,0.10421760082244873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,16,power_law_1.2,0.11573120355606079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,16,power_law_1.2,0.12268160581588745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,16,power_law_1.2,0.11849600076675415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,16,power_law_1.2,0.12787840366363526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,16,power_law_1.2,0.13030400276184081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,16,power_law_1.2,0.1270143985748291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,16,power_law_1.2,0.13386240005493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,16,power_law_1.2,0.14440959692001343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,16,power_law_1.2,0.15222400426864624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,16,power_law_1.2,0.15968639850616456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,16,power_law_1.2,0.18069119453430177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,16,power_law_1.2,0.20044798851013185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,16,power_law_1.2,0.22938239574432373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,16,power_law_1.2,0.2611776113510132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,16,power_law_1.2,0.34065918922424315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,16,power_law_1.2,0.38306560516357424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,16,power_law_1.2,0.5219136238098144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,8,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,8,balanced,0.0450133333603541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,16,power_law_1.2,0.6259200096130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,8,balanced,0.04738133152325948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,8,balanced,0.045663997530937195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,16,power_law_1.2,0.8889472007751464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,8,balanced,0.06673066814740498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,8,balanced,0.04953599969546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,8,balanced,0.09456533193588257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,16,power_law_1.2,1.1515711784362792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,8,balanced,0.07233599821726482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,8,balanced,0.14756799737612405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,8,balanced,0.10153067111968994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,16,power_law_1.2,1.3334848403930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,8,balanced,0.2451039950052897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,8,balanced,0.15030399958292642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,8,balanced,0.20543466011683145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,8,balanced,0.3455413182576497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,16,power_law_1.2,2.19299201965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,8,balanced,0.2051466703414917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,8,balanced,0.3465813398361206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,8,balanced,0.20589866240819296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,8,balanced,0.34721601009368896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,16,power_law_1.2,4.399078369140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,8,balanced,0.20729066928227743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,8,balanced,0.34865065415700275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,8,balanced,0.2086720069249471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,8,balanced,0.3490080038706462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,8,balanced,0.21317867437998453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,8,balanced,0.35229865709940594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,8,balanced,0.21178666750590006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,8,balanced,0.35337066650390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,8,balanced,0.21619733174641928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,8,balanced,0.3527466853459676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,8,balanced,0.2233440081278483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,8,balanced,0.36213866869608563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,8,balanced,0.22590933243433634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,8,balanced,0.36292799313863117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,8,balanced,0.2349546750386556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,8,balanced,0.3686613241831462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,8,balanced,0.2479626735051473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,8,balanced,0.3825920025507609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,8,balanced,0.25886933008829754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,8,balanced,0.39027198155721027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,8,balanced,0.28962133328119916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,8,balanced,0.05002133548259735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,8,balanced,0.4142986536026001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,8,power_law_1.01,0.09927679896354676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,8,balanced,0.31590932607650757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,8,balanced,0.05394133428732554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,8,power_law_1.01,0.14238719940185546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,8,balanced,0.43642131487528485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,8,balanced,0.054655998945236206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,8,power_law_1.01,0.23721599578857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,8,power_law_1.01,0.06689280271530151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,8,balanced,0.06241600215435028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,8,power_law_1.01,0.10575360059738159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,8,balanced,0.37212268511454266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,8,power_law_1.01,0.0939903974533081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,8,balanced,0.47964266935984295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,8,balanced,0.09059733152389526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,8,power_law_1.01,0.12942080497741698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,8,balanced,0.12272533774375916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,8,power_law_1.01,0.14612480401992797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,8,balanced,0.440170685450236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,8,power_law_1.01,0.1751039981842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,8,balanced,0.5226080020268759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,8,balanced,0.15798399845759073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,8,power_law_1.2,0.09866880178451538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,8,power_law_1.01,0.06392319798469544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,8,power_law_1.01,0.0700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,8,power_law_1.01,0.1974720001220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,8,balanced,0.1575040022532145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,8,balanced,0.5493493477503458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,8,balanced,0.5939893325169882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,8,power_law_1.2,0.14266879558563234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,8,balanced,0.1592693328857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,8,power_law_1.01,0.09544320106506347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,8,power_law_1.01,0.25321600437164304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,8,power_law_1.01,0.08242560029029847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,8,balanced,0.15863999724388123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,8,power_law_1.2,0.12954879999160768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,8,balanced,0.6788586775461832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,8,power_law_1.01,0.2639296054840088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,8,balanced,0.7238240242004395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,8,power_law_1.01,0.11911040544509888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,8,power_law_1.01,0.11484800577163697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,8,balanced,0.16133866707483926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,8,power_law_1.2,0.09563519954681396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,8,power_law_1.01,0.26978559494018556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,8,power_law_1.01,0.13278720378875733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,8,power_law_1.01,0.06759679913520814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,8,power_law_1.01,0.1466879963874817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,8,power_law_1.2,0.12961280345916748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,8,balanced,0.16379732886950174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,8,power_law_1.01,0.27324159145355226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,8,balanced,0.909226655960083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,8,power_law_1.01,0.1578112006187439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,8,power_law_1.01,0.0836031973361969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,8,balanced,0.16364799936612448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,8,power_law_1.2,0.16764800548553466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,8,power_law_1.01,0.27980799674987794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,8,power_law_1.01,0.16075520515441893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,8,balanced,0.8727413018544515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,8,balanced,0.1667626698811849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,8,power_law_1.01,0.10588159561157226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,8,power_law_1.2,0.19823360443115234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,8,power_law_1.01,0.2834815979003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,8,power_law_1.01,0.16741119623184203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,8,balanced,0.1718506614367167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,8,power_law_1.01,0.2974976062774658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,8,power_law_1.01,0.10895359516143799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,8,power_law_1.2,0.23515520095825196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,8,balanced,1.1647573312123616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,8,power_law_1.01,0.1820032000541687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,8,balanced,0.17947200934092203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,8,power_law_1.2,0.2533760070800781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,8,power_law_1.01,0.3236991882324219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,8,power_law_1.01,0.12646399736404418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,8,balanced,0.18519467115402222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,8,power_law_1.01,0.17726720571517945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,8,power_law_1.2,0.25082879066467284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,8,power_law_1.01,0.2979007959365845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,8,power_law_1.01,0.12346240282058715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,8,balanced,0.19640000661214194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,8,power_law_1.01,0.19094400405883788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,8,power_law_1.2,0.2656896114349365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,8,power_law_1.01,0.3265088081359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,8,power_law_1.01,0.12668800354003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,8,balanced,0.20670932531356812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,8,power_law_1.01,0.1911743998527527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,8,power_law_1.2,0.27029120922088623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,8,power_law_1.01,0.35253760814666746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,8,balanced,1.123744010925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,8,power_law_1.01,0.13362560272216797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,8,power_law_1.2,0.28037118911743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,8,power_law_1.01,0.20427520275115968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,8,power_law_1.01,0.14177919626235963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,8,power_law_1.01,0.38502399921417235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,8,balanced,0.23268266518910727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,8,power_law_1.2,0.28309121131896975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,8,power_law_1.01,0.21004800796508788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,8,power_law_1.01,0.14049279689788818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,8,power_law_1.01,0.4673151969909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,8,power_law_1.2,0.30575358867645264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,8,power_law_1.01,0.14651520252227784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,8,power_law_1.01,0.2353600025177002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,8,balanced,0.25489066044489544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,8,power_law_1.2,0.30678400993347166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,8,power_law_1.01,0.45832319259643556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,8,power_law_1.01,0.15461119413375854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,8,power_law_1.01,0.2513472080230713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,8,power_law_1.2,0.3404416084289551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,8,power_law_1.01,0.5313280105590821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,8,power_law_1.01,0.15692800283432007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,8,power_law_1.01,0.3138495922088623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,8,power_law_1.2,0.36821761131286623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,8,balanced,0.30447999636332196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,8,power_law_1.01,0.6425536155700684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,8,power_law_1.2,0.3945535898208618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,8,power_law_1.01,0.16852480173110962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,8,power_law_1.01,0.3332864046096802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,8,power_law_1.01,0.19415040016174318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,8,power_law_1.2,0.4618175983428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,8,power_law_1.01,0.790553617477417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,8,power_law_1.01,0.42106242179870607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,8,balanced,1.4359466234842937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,8,power_law_1.2,0.4722815990447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,8,power_law_1.01,0.20818560123443602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,8,power_law_1.01,0.8669055938720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,8,power_law_1.01,0.5026624202728271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,8,balanced,0.3564053376515706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,8,power_law_1.2,0.597983980178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,8,power_law_1.01,0.23700480461120604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,8,power_law_1.01,0.669491195678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,8,power_law_1.01,1.13504638671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,8,power_law_1.2,0.7070015907287598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,8,power_law_1.01,0.27760000228881837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,8,power_law_1.01,0.8056384086608886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,8,power_law_1.2,0.92608642578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,8,power_law_1.01,1.3668800354003907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,8,power_law_1.01,0.3593024015426636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,8,balanced,1.3726666768391926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,8,power_law_1.01,1.120742416381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,8,power_law_1.2,0.9300352096557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,8,power_law_1.01,0.39689600467681885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,8,balanced,0.44331200917561847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,8,power_law_1.01,1.7101503372192384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,8,power_law_1.2,1.281491184234619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,8,power_law_1.01,0.5481088161468506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,8,power_law_1.01,1.3333248138427733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,8,power_law_1.01,2.4535423278808595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,8,power_law_1.2,1.5295167922973634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,8,power_law_1.01,0.665235185623169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,8,power_law_1.01,1.6465471267700196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,8,power_law_1.2,2.0191104888916014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,8,power_law_1.01,0.8864319801330567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,8,power_law_1.01,4.884492874145508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,8,power_law_1.01,2.5411775588989256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,8,power_law_1.01,1.110105609893799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,8,power_law_1.2,2.6958656311035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,8,balanced,0.5703519980112711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,8,power_law_1.01,1.3682559967041015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,8,power_law_1.01,5.154060745239258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,8,power_law_1.2,5.668511962890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,8,power_law_1.2,0.06691200137138367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,8,power_law_1.01,2.2066623687744142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,8,power_law_1.2,0.09326720237731934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,8,balanced,2.1457600593566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,8,power_law_1.01,4.298092651367187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,8,power_law_1.2,0.10183039903640748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,8,balanced,0.746234655380249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,8,power_law_1.2,0.07230719923973083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,8,power_law_1.2,0.09395840167999267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,8,power_law_1.2,0.1121216058731079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,8,balanced,1.9423519770304363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,8,power_law_1.2,0.1162943959236145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,8,power_law_1.2,0.14655359983444213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,8,power_law_1.2,0.15739519596099855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,8,power_law_1.2,0.15425920486450195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,8,power_law_1.2,0.16146559715270997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,8,power_law_1.2,0.16398719549179078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,8,balanced,0.9660000006357828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,8,power_law_1.2,0.1792512059211731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,8,power_law_1.2,0.18284800052642822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,8,power_law_1.2,0.19059840440750123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,8,power_law_1.2,0.19676159620285033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,8,power_law_1.2,0.21713919639587403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,8,power_law_1.2,0.2488447904586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,8,power_law_1.2,0.27257599830627444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,8,power_law_1.2,0.32819840908050535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,8,power_law_1.2,0.35653760433197024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,8,power_law_1.2,0.4686272144317627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,8,power_law_1.2,0.5450560092926026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,8,power_law_1.2,0.6999872207641602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,8,balanced,1.195087989171346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,8,power_law_1.2,0.8746368408203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,8,power_law_1.2,1.14835844039917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,8,power_law_1.2,1.595967960357666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,8,power_law_1.2,1.8590400695800782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,8,power_law_1.2,3.1764991760253904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,8,power_law_1.2,5.758291244506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,8,balanced,4.18013858795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,8,balanced,3.731189409891764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,8,balanced,1.8193066914876301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,8,balanced,3.5222291946411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,8,power_law_1.2,0.06501759886741638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,8,power_law_1.2,0.0828607976436615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,8,power_law_1.2,0.08227199912071229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,8,power_law_1.2,0.06704000234603882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,8,power_law_1.2,0.0826367974281311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,8,power_law_1.2,0.09294080138206481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,8,power_law_1.2,0.10497280359268188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,8,power_law_1.2,0.12305920124053955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,8,power_law_1.2,0.12537599802017213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,8,power_law_1.2,0.12353279590606689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,8,power_law_1.2,0.13308160305023192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,8,power_law_1.2,0.13537919521331787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,8,power_law_1.2,0.13607679605484008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,8,power_law_1.2,0.14419200420379638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,8,power_law_1.2,0.15322240591049194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,8,power_law_1.2,0.16061439514160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,8,power_law_1.2,0.17196799516677858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,8,power_law_1.2,0.18691200017929077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,8,power_law_1.2,0.21521279811859131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,8,power_law_1.2,0.2546623945236206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,8,power_law_1.2,0.2978559970855713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,8,power_law_1.2,0.3656575918197632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,8,power_law_1.2,0.4596543788909912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,8,power_law_1.2,0.5604288101196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,8,power_law_1.2,0.7040128231048584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,8,power_law_1.2,0.9681344032287598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,8,power_law_1.2,1.2626239776611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,8,power_law_1.2,1.5912384033203124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,8,power_law_1.2,2.4796415328979493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,8,power_law_1.2,4.89306869506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,4,balanced,0.06017066538333893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,4,balanced,0.06035733222961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,4,balanced,0.06400000055631001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,4,balanced,0.07454399764537811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,4,balanced,0.10359467069307964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,4,balanced,0.14618133505185446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,4,balanced,0.1888373295466105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,4,balanced,0.19073599576950073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,4,balanced,0.19194134076436362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,4,balanced,0.1947093407313029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,4,balanced,0.19036799669265747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,4,balanced,0.19585599501927695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,4,balanced,0.19776000579198202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,4,balanced,0.20016533136367798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,4,balanced,0.20674665768941244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,4,balanced,0.2087679902712504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,4,balanced,0.21754666169484457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,4,balanced,0.233952005704244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,4,balanced,0.2510506709416707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,4,balanced,0.27940799792607623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,4,balanced,0.3105066617329915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,4,balanced,0.37482134501139325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,4,balanced,0.4380106528600057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,4,balanced,0.5670080184936523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,4,balanced,0.7546026706695557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,4,power_law_1.01,0.061894398927688596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,4,power_law_1.01,0.0718720018863678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,4,power_law_1.01,0.0995199978351593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,4,power_law_1.01,0.07337599992752075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,4,power_law_1.01,0.09377279877662659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,4,power_law_1.01,0.11298559904098511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,4,balanced,1.0093706448872883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,4,power_law_1.01,0.12332160472869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,4,power_law_1.01,0.1445312023162842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,4,power_law_1.01,0.15075839757919313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,4,power_law_1.01,0.15178240537643434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,4,power_law_1.01,0.15419520139694215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,4,power_law_1.01,0.1651520013809204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,4,power_law_1.01,0.16877440214157105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,4,power_law_1.01,0.17267199754714965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,4,power_law_1.01,0.18517119884490968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,4,power_law_1.01,0.18762880563735962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,4,power_law_1.01,0.20336639881134033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,4,balanced,1.3397919336954753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,4,power_law_1.01,0.21864960193634034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,4,power_law_1.01,0.24460160732269287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,4,power_law_1.01,0.28247039318084716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,4,power_law_1.01,0.3193599939346313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,4,power_law_1.01,0.40499200820922854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,4,power_law_1.01,0.47513599395751954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,4,power_law_1.01,0.6415359973907471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,4,power_law_1.01,0.8380543708801269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,4,power_law_1.01,1.1470144271850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,4,power_law_1.01,1.583622360229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,4,power_law_1.01,1.8234624862670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,4,power_law_1.01,2.8395776748657227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,4,balanced,1.6676534016927083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,4,power_law_1.01,5.130220794677735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,4,balanced,2.521695931752523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,4,balanced,4.819482803344727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,4,power_law_1.2,0.05917440056800842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,4,power_law_1.2,0.07027199864387512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,4,power_law_1.2,0.06704000234603882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,4,power_law_1.2,0.07631360292434693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,4,power_law_1.2,0.09415680170059204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,4,power_law_1.2,0.10568959712982177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,4,power_law_1.2,0.11486719846725464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,4,power_law_1.2,0.14345599412918092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,4,power_law_1.2,0.1484544038772583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,4,power_law_1.2,0.1460927963256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,4,power_law_1.2,0.15395840406417846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,4,power_law_1.2,0.161900794506073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,4,power_law_1.2,0.16308480501174927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,4,power_law_1.2,0.17114880084991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,4,power_law_1.2,0.1825279951095581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,4,power_law_1.2,0.1873088002204895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,4,balanced,0.04798933366934458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,4,power_law_1.2,0.19953919649124147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,4,power_law_1.2,0.22040319442749023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,4,balanced,0.04809066653251648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,4,balanced,0.05417066812515259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,4,power_law_1.2,0.24701440334320068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,4,balanced,0.0763679991165797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,4,power_law_1.2,0.2851263999938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,4,balanced,0.10528533657391866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,4,balanced,0.15722666184107462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,4,power_law_1.2,0.3289599895477295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,4,balanced,0.2084640065828959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,4,power_law_1.2,0.4294591903686523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,4,balanced,0.20757333437601724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,4,balanced,0.213210662206014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,4,power_law_1.2,0.5012928009033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,4,balanced,0.21658132473627725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,4,balanced,0.21413866678873697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,4,power_law_1.2,0.6749760150909424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,4,balanced,0.22230400641759238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,4,power_law_1.2,0.8779328346252442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,4,balanced,0.22615466515223184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,4,balanced,0.2288960019747416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,4,power_law_1.2,1.1904512405395509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,4,balanced,0.23731199900309244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,4,power_law_1.2,1.6064640045166017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,4,balanced,0.2404266595840454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,4,balanced,0.25548267364501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,4,power_law_1.2,2.0718080520629885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,4,balanced,0.2715359926223755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,4,power_law_1.2,2.9609216690063476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,4,balanced,0.2912320097287496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,4,balanced,0.3234399954477946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,4,power_law_1.2,5.450758361816407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,4,balanced,0.052101333936055504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,4,balanced,0.3547573486963908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,4,balanced,0.05468800167242686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,4,balanced,0.07368533313274384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,4,balanced,0.10116266210873921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,4,balanced,0.42851734161376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,4,balanced,0.1514133314291636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,4,balanced,0.24724799394607544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,4,balanced,0.3487306833267212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,4,balanced,0.34827200571695965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,4,balanced,0.3496853510538737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,4,balanced,0.5188746849695841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,4,balanced,0.3497866789499919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,4,balanced,0.35539201895395917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,4,balanced,0.35698668162027997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,4,balanced,0.3630133469899495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,4,balanced,0.36353600025177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,4,balanced,0.3703626791636149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,4,balanced,0.6654186646143595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,4,balanced,0.37404266993204754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,4,balanced,0.3890719811121623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,4,balanced,0.3986719846725464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,4,balanced,0.41103466351826984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,4,balanced,0.8505439758300781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,4,balanced,0.4435093402862549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,4,balanced,0.4678186575571696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,4,balanced,0.5353920062383016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,4,power_law_1.01,0.051686400175094606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,4,power_law_1.01,0.0752128005027771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,4,balanced,0.5930026769638062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,4,balanced,1.1542133490244548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,4,power_law_1.01,0.10259840488433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,4,power_law_1.01,0.07145599722862243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,4,power_law_1.01,0.09352319836616516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,4,power_law_1.01,0.11429120302200317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,4,balanced,0.701754649480184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,4,power_law_1.01,0.12943999767303466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,4,power_law_1.01,0.15492479801177977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,4,power_law_1.01,0.1567296028137207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,4,power_law_1.01,0.159334397315979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,4,power_law_1.01,0.16796799898147582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,4,power_law_1.01,0.18443520069122316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,4,balanced,0.8730879624684652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,4,balanced,1.4796160062154133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,4,power_law_1.01,0.19029760360717773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,4,power_law_1.01,0.19288959503173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,4,power_law_1.01,0.21035521030426024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,4,power_law_1.01,0.21427841186523439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,4,power_law_1.01,0.22347519397735596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,4,power_law_1.01,0.25166079998016355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,4,power_law_1.01,0.27009921073913573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,4,power_law_1.01,0.07331839799880982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,4,power_law_1.01,0.3253567934036255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,4,balanced,1.1189653078715007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,4,power_law_1.01,0.09514880180358887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,4,power_law_1.01,0.36250879764556887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,4,power_law_1.01,0.145798397064209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,4,power_law_1.01,0.46574082374572756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,4,power_law_1.01,0.0913536012172699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,4,power_law_1.01,0.5831103801727295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,4,power_law_1.01,0.13489279747009278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,4,power_law_1.01,0.7672255992889404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,4,balanced,1.830293337504069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,4,power_law_1.01,0.16536959409713745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,4,power_law_1.01,0.9382399559020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,4,power_law_1.01,0.1901952028274536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,4,power_law_1.01,1.3025983810424804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,4,power_law_1.01,0.24797439575195312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,4,power_law_1.01,1.6994943618774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,4,power_law_1.01,0.24776959419250488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,4,power_law_1.01,0.2585024118423462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,4,power_law_1.01,2.0290559768676757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,4,balanced,1.429082711537679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,4,power_law_1.01,0.2680959939956665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,4,power_law_1.01,3.035251235961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,4,power_law_1.01,0.2726399898529053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,4,power_law_1.01,0.2884799957275391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,4,power_law_1.01,6.124480056762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,4,power_law_1.01,0.2916032075881958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,4,power_law_1.01,0.3016063928604126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,4,power_law_1.01,0.3178816080093384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,4,power_law_1.01,0.3323071956634521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,4,power_law_1.01,0.3772991895675659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,4,power_law_1.01,0.38908159732818604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,4,power_law_1.01,0.44935040473937987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,4,power_law_1.01,0.4674111843109131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,4,power_law_1.01,0.5887296199798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,4,balanced,1.7863146464029949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,4,power_law_1.01,0.6657919883728027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,4,balanced,2.7809279759724936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,4,power_law_1.01,0.8697728157043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,4,power_law_1.01,0.9653759956359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,4,power_law_1.01,1.2650431632995605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,4,power_law_1.01,1.6849983215332032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,4,power_law_1.01,1.94586238861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,4,power_law_1.01,2.946745681762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,4,power_law_1.01,5.733062362670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,4,balanced,2.5510613123575845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,4,power_law_1.2,0.07308160066604615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,4,power_law_1.2,0.09508479833602905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,4,power_law_1.2,0.10258560180664063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,4,power_law_1.2,0.0906112015247345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,4,power_law_1.2,0.12277120351791382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,4,power_law_1.2,0.1581439971923828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,4,power_law_1.2,0.18373119831085205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,4,power_law_1.2,0.23022079467773438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,4,power_law_1.2,0.2479680061340332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,4,power_law_1.2,0.254694390296936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,4,power_law_1.2,0.26238079071044923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,4,power_law_1.2,0.27752959728240967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,4,power_law_1.2,0.27842559814453127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,4,power_law_1.2,0.05232639908790589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,4,power_law_1.2,0.29450240135192873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,4,balanced,5.410229365030925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,4,power_law_1.2,0.07492480278015137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,4,power_law_1.2,0.304256010055542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,4,power_law_1.2,0.0725055992603302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,4,power_law_1.2,0.32106239795684816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,4,power_law_1.2,0.34856319427490234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,4,power_law_1.2,0.0705407977104187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,4,power_law_1.2,0.08859519958496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,4,power_law_1.2,0.36883840560913084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,4,power_law_1.2,0.10873600244522094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,4,power_law_1.2,0.4019904136657715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,4,power_law_1.2,0.117958402633667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,4,power_law_1.2,0.48103041648864747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,4,power_law_1.2,0.1555840015411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,4,power_law_1.2,0.5081215858459472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,4,power_law_1.2,0.1568063974380493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,4,power_law_1.2,0.6306111812591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,4,power_law_1.2,0.16027519702911378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,4,power_law_1.2,0.7028351783752441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,4,power_law_1.2,0.17011200189590453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,4,power_law_1.2,0.9327936172485352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,4,power_law_1.2,0.17956479787826538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,4,power_law_1.2,1.108249568939209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,4,power_law_1.2,0.18645119667053223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,4,power_law_1.2,1.365011215209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,4,power_law_1.2,0.19615999460220337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,4,power_law_1.2,0.20648961067199706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,4,power_law_1.2,1.833465576171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,4,power_law_1.2,0.22008318901062013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,4,power_law_1.2,2.192550468444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,4,power_law_1.2,0.23036799430847169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,4,power_law_1.2,0.2643647909164429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,4,power_law_1.2,3.1791296005249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,4,power_law_1.2,0.2864896059036255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,4,power_law_1.2,0.345414400100708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,4,balanced,4.915322621663411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,4,power_law_1.2,6.397945785522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,4,power_law_1.2,0.37813119888305663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,4,power_law_1.2,0.49152002334594724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,4,power_law_1.2,0.6243135929107666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,4,power_law_1.2,0.8331007957458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,4,power_law_1.2,0.9972160339355469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,4,power_law_1.2,1.3339584350585938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,4,power_law_1.2,1.7312192916870117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,4,power_law_1.2,2.144806480407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,4,power_law_1.2,3.274790573120117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,4,power_law_1.2,6.809388732910156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,2,balanced,0.07023466626803081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,2,balanced,0.07428800066312154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,2,balanced,0.07826666533946991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,2,balanced,0.09322667121887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,2,balanced,0.133925328652064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,2,balanced,0.19903467098871866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,2,balanced,0.26132800181706745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,2,balanced,0.25339200099309284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,2,power_law_1.01,0.06639360189437866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,2,balanced,0.2526560028394063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,2,balanced,0.2588640054066976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,2,power_law_1.01,0.07170559763908387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,2,balanced,0.25679999589920044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,2,balanced,0.2595040003458659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,2,power_law_1.01,0.08562560081481933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,2,balanced,0.26180799802144367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,2,power_law_1.01,0.0880895972251892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,2,balanced,0.26709866523742676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,2,power_law_1.01,0.11559679508209228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,2,balanced,0.27550933758417767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,2,power_law_1.01,0.14266239404678344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,2,balanced,0.2796799937884013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,2,power_law_1.01,0.16178560256958008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,2,balanced,0.2892533342043559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,2,power_law_1.01,0.19137920141220094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,2,balanced,0.31465067466100055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,2,power_law_1.01,0.19842560291290284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,2,balanced,0.333296000957489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,2,power_law_1.01,0.2028480052947998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,2,power_law_1.01,0.20869119167327882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,2,balanced,0.3832213481267293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,2,power_law_1.01,0.21891839504241944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,2,power_law_1.01,0.22816638946533202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,2,balanced,0.42479999860127765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,2,power_law_1.01,0.22848639488220215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,2,power_law_1.01,0.24167680740356445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,2,power_law_1.01,0.2541568040847778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,2,balanced,0.517413338025411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,2,power_law_1.01,0.26153600215911865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,2,power_law_1.01,0.2880511999130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,2,power_law_1.01,0.3210432052612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,2,balanced,0.6187893152236938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,2,power_law_1.01,0.37674241065979003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,2,power_law_1.01,0.4381247997283936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,2,power_law_1.01,0.5375487804412842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,2,power_law_1.01,0.6636032104492188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,2,balanced,0.8273226420084635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,2,power_law_1.01,0.9168191909790039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,2,power_law_1.01,1.1700799942016602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,2,power_law_1.01,1.6203264236450194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,2,power_law_1.01,2.059391975402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,2,power_law_1.01,2.5419391632080077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,2,balanced,1.1056640148162842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,2,power_law_1.01,4.015878295898437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,2,power_law_1.01,7.409190368652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,2,balanced,1.518415927886963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,2,balanced,2.0075732866923013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,2,balanced,2.5000319480895996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,2,balanced,3.808160146077474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,2,balanced,7.268149058024089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,2,balanced,0.05403733253479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,2,balanced,0.058378666639328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,2,balanced,0.07663466533025105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,2,balanced,0.10340266426404317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,2,balanced,0.15668800473213196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,2,balanced,0.2555466691652934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,2,balanced,0.3577226797739665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,2,balanced,0.358517328898112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,2,balanced,0.3669919967651367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,2,balanced,0.36237867673238117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,2,balanced,0.3702826499938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,2,balanced,0.3802773157755534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,2,balanced,0.3836959997812907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,2,balanced,0.3787200053532918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,2,balanced,0.3943733374277751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,2,balanced,0.3914666573206584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,2,balanced,0.4134133259455363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,2,balanced,0.43320000171661377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,2,balanced,0.4529866774876912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,2,balanced,0.5013706684112549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,2,power_law_1.2,0.0661952018737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,2,balanced,0.5424426794052124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,2,power_law_1.2,0.07162879705429077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,2,power_law_1.2,0.07625600099563598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,2,power_law_1.2,0.08781440258026123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,2,balanced,0.6326293150583903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,2,power_law_1.2,0.11180800199508667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,2,power_law_1.2,0.13642239570617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,2,power_law_1.2,0.15086079835891725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,2,balanced,0.7432373364766439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,2,power_law_1.2,0.1820736050605774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,2,power_law_1.2,0.1929471969604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,2,power_law_1.2,0.1992640018463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,2,power_law_1.2,0.2035072088241577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,2,power_law_1.2,0.21779839992523192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,2,balanced,0.9392906824747721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,2,power_law_1.2,0.21894400119781493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,2,power_law_1.2,0.22669439315795897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,2,power_law_1.2,0.23966081142425538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,2,power_law_1.2,0.24851839542388915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,2,power_law_1.2,0.26156160831451414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,2,balanced,1.1909653345743816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,2,power_law_1.2,0.2981120109558105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,2,power_law_1.2,0.3187903881072998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,2,power_law_1.2,0.38896639347076417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,2,power_law_1.2,0.44057598114013674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,2,power_law_1.2,0.5628928184509278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,2,power_law_1.2,0.6715583801269531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,2,power_law_1.2,0.9156288146972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,2,power_law_1.2,1.2145983695983886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,2,balanced,1.5747413635253906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,2,power_law_1.2,1.6616704940795899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,2,power_law_1.2,2.0373376846313476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,2,power_law_1.2,2.5370367050170897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,2,power_law_1.2,3.952755355834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,2,power_law_1.2,7.643199920654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,2,balanced,2.0463786125183105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,2,balanced,0.0516480008761088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,2,balanced,0.053690666953722634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,2,balanced,0.059194669127464294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,2,balanced,0.08444266517957051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,2,balanced,0.11434132854143779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,2,balanced,0.17271467049916586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,2,balanced,2.5709546407063804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,2,balanced,0.233130673567454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,2,balanced,0.2274506688117981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,2,balanced,0.22980799277623495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,2,balanced,0.23119999965031943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,2,balanced,0.2331999937693278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,2,balanced,0.23798932631810507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,2,balanced,0.24170666933059692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,2,balanced,0.24515734116236368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,2,balanced,0.2551093300183614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,2,balanced,0.2625760038693746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,2,balanced,0.27491732438405353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,2,balanced,0.30984532833099365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,2,balanced,0.3417066733042399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,2,balanced,0.39398932456970215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,2,balanced,0.4509706497192383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,2,balanced,0.5546773274739584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,2,balanced,0.6771732966105143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,2,balanced,3.6655146280924478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,2,balanced,0.8805973529815674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,2,balanced,1.163925329844157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,2,balanced,1.5822933514912922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,2,balanced,2.1045173009236655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,2,balanced,2.6372373898824057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,2,balanced,7.058773040771484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,2,power_law_1.01,0.05134720206260681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,2,power_law_1.01,0.0717631995677948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,2,power_law_1.01,0.09754239916801452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,2,power_law_1.01,0.08227199912071229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,2,power_law_1.01,0.1261504054069519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,2,power_law_1.01,0.17002880573272705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,2,power_law_1.01,0.19025280475616455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,2,power_law_1.01,0.23963520526885987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,2,power_law_1.01,0.25420799255371096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,2,power_law_1.01,0.2643071889877319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,2,power_law_1.01,0.279040002822876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,2,power_law_1.01,0.2923968076705933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,2,balanced,3.9294986724853516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,2,power_law_1.01,0.3012864112854004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,2,power_law_1.01,0.30316801071166993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,2,power_law_1.01,0.3278847932815552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,2,power_law_1.01,0.3389375925064087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,2,power_law_1.01,0.35404798984527586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,2,power_law_1.01,0.39667840003967286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,2,power_law_1.01,0.4356351852416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,2,power_law_1.01,0.49637761116027834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,2,power_law_1.01,0.5380735874176026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,2,power_law_1.01,0.6718143939971923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,2,power_law_1.01,0.8265472412109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,2,power_law_1.01,1.081158447265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,2,power_law_1.01,1.3587008476257325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,2,power_law_1.01,0.04785279929637909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,2,power_law_1.01,1.8091264724731446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,2,power_law_1.01,2.168819236755371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,2,power_law_1.01,0.05614719986915588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,2,power_law_1.01,0.07868800163269044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,2,power_law_1.01,2.6190912246704103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,2,power_law_1.01,0.06907520294189454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,2,power_law_1.01,3.8970752716064454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,2,power_law_1.01,0.1015936017036438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,2,power_law_1.01,0.12371840476989746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,2,power_law_1.01,0.13930879831314086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,2,power_law_1.01,7.384480285644531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,2,power_law_1.01,0.16751999855041505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,2,power_law_1.01,0.17778559923171997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,2,power_law_1.01,0.1804095983505249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,2,power_law_1.01,0.18944000005722045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,2,power_law_1.01,0.2002687931060791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,2,power_law_1.01,0.2096640110015869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,2,power_law_1.01,0.22053120136260987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,2,power_law_1.01,0.235532808303833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,2,power_law_1.01,0.2512383937835693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,2,power_law_1.01,0.2782912015914917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,2,power_law_1.01,0.32083840370178224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,2,power_law_1.01,0.3428992033004761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,2,power_law_1.01,0.4131904125213623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,2,power_law_1.01,0.4626431941986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,2,power_law_1.01,0.6121664047241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,2,power_law_1.01,0.7247168064117432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,2,balanced,7.606864293416341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,2,power_law_1.01,0.993075180053711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,2,power_law_1.01,1.305568027496338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,2,power_law_1.01,1.791219139099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,2,power_law_1.01,2.121388816833496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,2,power_law_1.01,2.6991424560546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,2,power_law_1.01,4.125196838378907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,2,power_law_1.01,8.117727661132813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,2,power_law_1.2,0.05360640287399292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,2,power_law_1.2,0.04976640045642853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,2,power_law_1.2,0.0748416006565094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,2,power_law_1.2,0.05622400045394897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,2,power_law_1.2,0.07933440208435058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,2,power_law_1.2,0.06356480121612548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,2,power_law_1.2,0.08735359907150268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,2,power_law_1.2,0.0709119975566864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,2,power_law_1.2,0.12099839448928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,2,power_law_1.2,0.09428480267524719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,2,power_law_1.2,0.16003199815750122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,2,power_law_1.2,0.11567360162734985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,2,power_law_1.2,0.1760256052017212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,2,power_law_1.2,0.1335039973258972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,2,power_law_1.2,0.23909120559692382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,2,power_law_1.2,0.16615040302276612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,2,power_law_1.2,0.24495999813079833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,2,power_law_1.2,0.17208319902420044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,2,power_law_1.2,0.26012799739837644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,2,power_law_1.2,0.17512960433959962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,2,power_law_1.2,0.27146880626678466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,2,power_law_1.2,0.187225604057312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,2,power_law_1.2,0.2863231897354126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,2,power_law_1.2,0.19802240133285523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,2,power_law_1.2,0.2927999973297119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,2,power_law_1.2,0.2052288055419922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,2,power_law_1.2,0.3104576110839844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,2,power_law_1.2,0.22416000366210936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,2,power_law_1.2,0.3255039930343628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,2,power_law_1.2,0.23641600608825683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,2,power_law_1.2,0.3471872091293335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,2,power_law_1.2,0.2512063980102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,2,power_law_1.2,0.3681024074554443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,2,power_law_1.2,0.2773632049560547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,2,power_law_1.2,0.39918720722198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,2,power_law_1.2,0.31953279972076415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,2,power_law_1.2,0.42968320846557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,2,power_law_1.2,0.3393984079360962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,2,power_law_1.2,0.5314047813415528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,2,power_law_1.2,0.41984000205993655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,2,power_law_1.2,0.5642687797546386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,2,power_law_1.2,0.48611841201782224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,2,power_law_1.2,0.7063936233520508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,2,power_law_1.2,0.6394624233245849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,2,power_law_1.2,0.8748928070068359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,2,power_law_1.2,0.7338751792907715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,2,power_law_1.2,1.1467328071594238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,2,power_law_1.2,1.0034048080444335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,2,power_law_1.2,1.3444928169250487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,2,power_law_1.2,1.3598719596862794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,2,power_law_1.2,1.7067264556884765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,2,power_law_1.2,1.7946432113647461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,2,power_law_1.2,2.2624256134033205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,2,power_law_1.2,2.245510482788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,2,power_law_1.2,2.804153633117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,2,power_law_1.2,2.704614448547363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,2,power_law_1.2,4.0704193115234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,2,power_law_1.2,4.306022262573242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,2,power_law_1.2,7.9044349670410154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,2,power_law_1.2,8.098892974853516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,1,balanced,0.10328533252080281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,1,balanced,0.11107200384140015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,1,balanced,0.11966400345166524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,1,balanced,0.14616533120473227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,1,balanced,0.20387200514475504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,1,balanced,0.302842656771342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,1,balanced,0.40089066823323566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,1,balanced,0.40034667650858563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,1,balanced,0.4010773499806722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,1,balanced,0.40487468242645264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,1,balanced,0.4102826515833537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,1,balanced,0.4161866505940755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,1,balanced,0.4188213348388672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,1,balanced,0.425493319829305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,1,balanced,0.43693868319193524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,1,balanced,0.44301867485046387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,1,balanced,0.4617866675059001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,1,balanced,0.5107040007909139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,1,balanced,0.5453919967015585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,1,balanced,0.6341066757837931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,1,balanced,0.70469864209493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,1,balanced,0.8774773279825846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,1,balanced,1.065002679824829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,1,balanced,1.4362133344014485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,1,balanced,1.9454347292582195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,1,balanced,2.6815414428710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,1,balanced,3.5925280253092446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,1,balanced,4.485146522521973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,1,power_law_1.01,0.09679999947547913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,1,power_law_1.01,0.10121599435806275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,1,power_law_1.01,0.11291520595550537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,1,power_law_1.01,0.13029119968414307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,1,power_law_1.01,0.1769215941429138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,1,power_law_1.01,0.21668479442596436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,1,power_law_1.01,0.24130558967590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,1,power_law_1.01,0.2859711885452271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,1,power_law_1.01,0.3035583972930908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,1,power_law_1.01,0.31024000644683836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,1,power_law_1.01,0.32036480903625486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,1,balanced,6.899802525838216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,1,power_law_1.01,0.33692159652709963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,1,power_law_1.01,0.34167680740356443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,1,power_law_1.01,0.35587201118469236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,1,power_law_1.01,0.3731264114379883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,1,power_law_1.01,0.3901247978210449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,1,power_law_1.01,0.4100927829742432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,1,power_law_1.01,0.46660480499267576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.01,0.5147391796112061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.01,0.6088064193725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.01,0.6982016086578369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.01,0.9026047706604003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.01,1.0984512329101563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.01,1.499135971069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,1,power_law_1.01,2.0276416778564452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,1,power_law_1.01,2.806937599182129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,1,power_law_1.01,3.5001407623291017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,1,power_law_1.01,4.292678451538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,1,power_law_1.01,6.663750457763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,1,power_law_1.01,12.926725769042969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,1,balanced,13.1527468363444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,1,balanced,0.05821333328882853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,1,balanced,0.06608533362547557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,1,balanced,0.08273600041866302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,1,balanced,0.11276800433794658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,1,balanced,0.17511999607086182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,1,balanced,0.2950773239135742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,1,balanced,0.4155466556549072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,1,balanced,0.4127093156178792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,1,balanced,0.4143893321355184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,1,balanced,0.4184693495432536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,1,balanced,0.4240106741587321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,1,balanced,0.4259519974390666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,1,balanced,0.4289919932683309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,1,balanced,0.4393226703008016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,1,balanced,0.4477333227793376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,1,balanced,0.45431466897328693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,1,balanced,0.4731146494547526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,1,balanced,0.517408013343811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,1,balanced,0.5526453256607056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,1,balanced,0.6360586484273275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,1,balanced,0.7024746735890707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,1,balanced,0.8555520375569662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,1,balanced,1.0562079747517903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,1,balanced,1.3550933202107747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,1,power_law_1.2,0.09325439929962158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,1,power_law_1.2,0.0986624002456665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,1,balanced,1.7934613227844238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,1,power_law_1.2,0.10796159505844116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,1,power_law_1.2,0.13159680366516113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,1,power_law_1.2,0.16692479848861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,1,power_law_1.2,0.2002432107925415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,1,power_law_1.2,0.22123520374298095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,1,power_law_1.2,0.28159360885620116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,1,power_law_1.2,0.29285120964050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,1,power_law_1.2,0.29748480319976806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,1,balanced,2.3994240760803223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,1,power_law_1.2,0.30999040603637695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,1,power_law_1.2,0.32860798835754396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,1,power_law_1.2,0.34269440174102783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,1,power_law_1.2,0.35136001110076903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,1,power_law_1.2,0.37287039756774903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,1,power_law_1.2,0.38108799457550047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,1,power_law_1.2,0.4186560153961182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,1,power_law_1.2,0.4646656036376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.2,0.5165952205657959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.2,0.6235519886016846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.2,0.7254079818725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,1,balanced,3.232842763264974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.2,0.927673625946045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.2,1.1148415565490724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.2,1.5124223709106446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,1,power_law_1.2,2.0620927810668945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,1,power_law_1.2,2.8363967895507813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,1,power_law_1.2,3.481913757324219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,1,power_law_1.2,4.271507263183594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,1,power_law_1.2,6.764915466308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,1,balanced,4.027157465616862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,1,power_law_1.2,13.175558471679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,1,balanced,0.05793599784374237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,1,balanced,0.061343997716903687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,1,balanced,0.07026666899522145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,1,balanced,0.09681600332260132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,1,balanced,0.14565333724021912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,1,balanced,0.22209600607554117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,1,balanced,0.29755733410517377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,1,balanced,0.2997013330459595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,1,balanced,0.30377066135406494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,1,balanced,0.3044053316116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,1,balanced,0.3078453342119853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,1,balanced,0.3183679978052775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,1,balanced,0.319050669670105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,1,balanced,0.3285013238588969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,1,balanced,0.34059735139211017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,1,balanced,5.917045593261719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,1,balanced,0.34674131870269775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,1,balanced,0.3673280080159505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,1,balanced,0.41794665654500324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,1,balanced,0.4524799982706706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,1,balanced,0.5341493288675944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,1,balanced,0.6074986855189005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,1,balanced,0.772432009379069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,1,balanced,0.9778827031453451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,1,balanced,1.2988800207773845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,1,balanced,1.7445173263549805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,1,balanced,2.484469254811605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,1,balanced,11.395306905110678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,1,balanced,3.2421013514200845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,1,balanced,4.0572052001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,1,balanced,6.153104146321614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,32,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,32,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,32,balanced,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,1,power_law_1.01,0.05377280116081238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,32,balanced,0.03566399961709976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,32,balanced,0.056128000219662987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,32,balanced,0.057018667459487915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,1,power_law_1.01,0.05729280114173889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,32,balanced,0.05706666906674703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,1,power_law_1.01,0.07780479788780212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,32,balanced,0.05618133147557577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,32,balanced,0.056218668818473816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,1,power_law_1.01,0.09132159948348999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,32,balanced,0.05807466804981232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,1,power_law_1.01,0.1402176022529602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,32,balanced,0.0581226646900177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,32,balanced,0.0598826656738917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,1,power_law_1.01,0.19226239919662474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,32,balanced,0.06009600063165029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,1,power_law_1.01,0.22417919635772704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,32,balanced,0.0591839998960495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,1,power_law_1.01,0.27882239818572996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,32,balanced,0.06414933502674103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,1,power_law_1.01,0.2996095895767212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,32,balanced,0.06404266754786174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,1,power_law_1.01,0.30732159614562987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,32,balanced,0.06601066887378693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,1,power_law_1.01,0.3172352075576782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,32,balanced,0.06604266663392384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,1,power_law_1.01,0.33060479164123535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,32,balanced,0.07109333574771881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,1,power_law_1.01,0.3396415948867798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,1,power_law_1.01,0.36472320556640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,32,balanced,0.07421866556008656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,1,power_law_1.01,0.3790143966674805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,1,power_law_1.01,0.38516480922698976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,32,balanced,0.07455466687679291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,1,power_law_1.01,0.4041088104248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,1,power_law_1.01,0.47426562309265136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,32,balanced,0.08529067039489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,1,power_law_1.01,0.5198272228240967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,1,power_law_1.01,0.6291647911071777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,1,power_law_1.01,0.6991680145263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,32,balanced,0.09175999959309895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,1,power_law_1.01,0.8937727928161621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,1,power_law_1.01,1.0955327987670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,1,power_law_1.01,1.4378047943115235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,1,power_law_1.01,1.909164810180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,32,balanced,0.10735999544461568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,1,power_law_1.01,2.576255989074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,1,power_law_1.01,3.1543487548828124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,1,power_law_1.01,3.829830551147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,1,balanced,11.904725392659506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,1,power_law_1.01,5.978860855102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,32,balanced,0.1320853332678477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,1,power_law_1.01,0.05514240264892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,1,power_law_1.01,11.343852996826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,1,power_law_1.01,0.05957760214805603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,1,power_law_1.01,0.06602240204811097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,1,power_law_1.01,0.08222079873085023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,32,balanced,0.16387733817100525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,1,power_law_1.01,0.12457599639892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,1,power_law_1.01,0.15575679540634155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,1,power_law_1.01,0.1781440019607544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,1,power_law_1.01,0.21239678859710692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,1,power_law_1.01,0.2229759931564331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,1,power_law_1.01,0.22890880107879638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,1,power_law_1.01,0.23914880752563478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,1,power_law_1.01,0.2621119976043701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,1,power_law_1.01,0.2689663887023926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,32,balanced,0.19593600432078043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,1,power_law_1.01,0.2908031940460205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,1,power_law_1.01,0.31015040874481203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,1,power_law_1.01,0.33055360317230226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,1,power_law_1.01,0.3236288070678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,1,power_law_1.01,0.38458240032196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,1,power_law_1.01,0.43337597846984866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,1,power_law_1.01,0.5376512050628662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,1,power_law_1.01,0.6229504108428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,1,power_law_1.01,0.8184255599975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,1,power_law_1.01,1.0188351631164552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,1,power_law_1.01,1.3895999908447265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,1,power_law_1.01,1.7499200820922851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,32,balanced,0.23164800802866617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,1,power_law_1.01,2.4724416732788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,1,power_law_1.01,3.1872512817382814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,1,power_law_1.01,3.9181758880615236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,1,power_law_1.01,6.208160018920898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,1,power_law_1.01,11.8347900390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,32,balanced,0.3189866741498311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,1,power_law_1.2,0.055027198791503903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,1,power_law_1.2,0.06072319746017456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,32,balanced,0.03533333291610082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,1,power_law_1.2,0.0756991982460022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,32,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,1,power_law_1.2,0.096288001537323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,32,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,32,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,1,power_law_1.2,0.13519359827041627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,32,balanced,0.03939733405907949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,1,power_law_1.2,0.1798784017562866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,32,balanced,0.039605334401130676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,1,power_law_1.2,0.2074431896209717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,32,balanced,0.03988266736268997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,32,balanced,0.03940266619126002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,1,power_law_1.2,0.2678335905075073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,32,balanced,0.03937066594759623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,32,balanced,0.04125866790612539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,1,power_law_1.2,0.28307840824127195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,32,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,1,power_law_1.2,0.29575040340423586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,32,balanced,0.041434665520985924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,1,power_law_1.2,0.31436159610748293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,32,balanced,0.041482667128245033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,32,balanced,0.04176533222198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,1,power_law_1.2,0.3273920059204102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,32,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,1,power_law_1.2,0.3388927936553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,32,balanced,0.04569066564242045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,1,power_law_1.2,0.3562623977661133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,32,balanced,0.04692799846331278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,1,power_law_1.2,0.37256319522857667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,1,power_law_1.2,0.05338240265846252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,32,balanced,0.04795733094215393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,1,power_law_1.2,0.3842751979827881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,1,power_law_1.2,0.058316802978515624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,1,power_law_1.2,0.4152639865875244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,32,balanced,0.05342933535575867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,1,power_law_1.2,0.06544640064239501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,1,power_law_1.2,0.4802559852600098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,1,power_law_1.2,0.08288000226020813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,32,balanced,0.05942399799823761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,1,power_law_1.2,0.5365695953369141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,1,power_law_1.2,0.10807679891586304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,1,power_law_1.2,0.6349696159362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,1,power_law_1.2,0.14069119691848755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,32,balanced,0.06017066538333893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,1,power_law_1.2,0.7250112056732178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,1,power_law_1.2,0.15947519540786742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,1,power_law_1.2,0.9100031852722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,1,power_law_1.2,0.20511999130249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,1,power_law_1.2,1.1204095840454102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,1,power_law_1.2,0.2167743921279907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,32,balanced,0.07855999966462453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,1,power_law_1.2,1.4608768463134765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,1,power_law_1.2,0.21831040382385253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,32,balanced,0.031317333380381264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,1,power_law_1.2,0.23284480571746827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,1,power_law_1.2,1.9525375366210938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,32,balanced,0.03166933357715607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,1,power_law_1.2,0.2558144092559814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,32,balanced,0.08824533224105835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,1,power_law_1.2,2.6217344284057615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,32,balanced,0.03133333226044973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,1,power_law_1.2,0.2649791955947876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,32,balanced,0.031680000325044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,1,power_law_1.2,3.142348861694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,32,balanced,0.03586666782697042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,32,balanced,0.5585066477457682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,1,power_law_1.2,0.29130239486694337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,32,balanced,0.03749333322048187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,32,balanced,0.037765334049860634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,1,power_law_1.2,0.31146240234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,1,power_law_1.2,3.8289535522460936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,32,balanced,0.03760000069936117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,1,power_law_1.2,0.33064320087432864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,32,balanced,0.11051199833552043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,32,balanced,0.03753600021203359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,1,power_law_1.2,6.068454360961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,32,balanced,0.039274667700131737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,1,power_law_1.2,0.3296191930770874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,32,balanced,0.03781333317359289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,32,balanced,0.03956266740957896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,1,power_law_1.2,0.3901952028274536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,1,power_law_1.2,11.449932861328126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,32,balanced,0.03982933362325033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,1,power_law_1.2,0.44505600929260253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,32,balanced,0.04154666761557261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,32,balanced,0.13191999991734824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,1,power_law_1.2,0.5525184154510498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,32,balanced,0.04600533346335093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,32,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,1,power_law_1.2,0.6359104156494141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,32,balanced,0.0490880012512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,1,power_law_1.2,0.8325887680053711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,32,balanced,0.052239999175071716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,1,power_law_1.2,1.0422080039978028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,32,balanced,0.06036800146102905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,1,power_law_1.2,1.4053055763244628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,32,balanced,0.07369600236415863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,32,balanced,0.16677866379419962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,1,power_law_1.2,1.7692480087280273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,1,power_law_1.2,2.493721580505371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,32,balanced,0.08050666749477386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,32,balanced,0.023669332265853882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,1,power_law_1.2,3.18338565826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,32,balanced,0.021418665846188862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,32,power_law_1.01,0.061452800035476686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,32,balanced,0.09826667110125224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,1,power_law_1.2,3.900774383544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,32,balanced,0.021295999487241108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,32,balanced,0.02309333284695943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,32,power_law_1.01,0.03966720104217529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,32,balanced,0.02555199960867564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,1,power_law_1.2,6.252179336547852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,32,power_law_1.01,0.03441919982433319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,32,balanced,0.02959466725587845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,32,power_law_1.01,0.045900800824165346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,32,balanced,0.045968001087506614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,32,balanced,0.11354133486747742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,32,power_law_1.01,0.046342399716377256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,32,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,1,power_law_1.2,11.848857879638672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,32,power_law_1.01,0.05334399938583374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,32,balanced,0.04830400149027506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,32,power_law_1.01,0.054079997539520266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,32,balanced,0.21278399229049683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,32,balanced,0.0462773342927297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,32,power_law_1.01,0.05374079942703247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,32,balanced,0.04502399762471517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,32,power_law_1.01,0.05563520193099976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,32,balanced,0.14430933197339377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,32,balanced,0.04301866888999939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,32,power_law_1.01,0.055219197273254396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,32,balanced,0.03982399900754293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,32,power_law_1.01,0.058931201696395874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,32,balanced,0.03544000039498011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,32,power_law_1.01,0.05930240154266357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,32,balanced,0.058703998724619545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,32,power_law_1.01,0.060172802209854125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,32,balanced,0.05729599793752035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,32,power_law_1.01,0.06312959790229797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,32,power_law_1.01,0.06806399822235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,32,balanced,0.04822933177153269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,32,balanced,0.18435200055440268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,32,power_law_1.01,0.07090560197830201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,32,balanced,0.05597866574923197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,32,power_law_1.01,0.07153279781341552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,32,power_law_1.01,0.07882239818572997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,32,balanced,0.06889600058396657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,32,power_law_1.01,0.08440319895744323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,32,balanced,0.08795733253161113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,32,power_law_1.01,0.0955136001110077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,32,balanced,0.2524106701215108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,32,power_law_1.01,0.10572160482406616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,32,balanced,0.1095360020796458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,32,power_law_1.01,0.14366079568862916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,32,balanced,0.24844799439112344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,32,power_law_1.01,0.15208319425582886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,32,power_law_1.01,0.190009605884552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,32,power_law_1.01,0.26965761184692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,32,balanced,0.14432533582051596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,32,power_law_1.01,0.35773439407348634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,32,power_law_1.01,0.49625601768493655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,32,power_law_1.01,0.5980288028717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,32,balanced,0.18346667289733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,32,power_law_1.01,0.9578432083129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,32,power_law_1.01,1.8427072525024415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,32,balanced,0.29285866022109985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,32,balanced,0.25806933641433716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,32,balanced,0.33351465066274005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,32,power_law_1.01,0.03547520041465759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,32,balanced,0.3629973332087199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,32,power_law_1.01,0.034822401404380796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,32,power_law_1.01,0.031462401151657104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,32,power_law_1.01,0.03374719917774201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,32,power_law_1.01,0.034796801209449765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,32,balanced,0.3593920071919759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,32,power_law_1.01,0.036499199271202085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,32,power_law_1.01,0.03711999952793121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,32,power_law_1.01,0.03726719915866852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,32,power_law_1.01,0.04016000032424927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,32,power_law_1.01,0.04147199988365173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,32,balanced,0.4839093287785848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,32,power_law_1.01,0.045510399341583255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,32,power_law_1.01,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,32,power_law_1.01,0.049542400240898135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,32,power_law_1.01,0.05116159915924072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,32,power_law_1.01,0.055641597509384154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,32,power_law_1.01,0.056979197263717654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,32,power_law_1.01,0.06110079884529114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,32,power_law_1.01,0.07195519804954528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,32,power_law_1.01,0.07583360075950622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,32,power_law_1.01,0.09687680006027222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,32,power_law_1.01,0.11427199840545654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,32,power_law_1.01,0.13506560325622557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,32,power_law_1.01,0.167193603515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,32,balanced,0.6389760176340739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,32,power_law_1.01,0.2178175926208496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,32,power_law_1.01,0.30008320808410643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,32,power_law_1.01,0.45840001106262207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,32,balanced,0.543994665145874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,32,power_law_1.01,0.5163904190063476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,32,power_law_1.01,0.6816383838653565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,32,power_law_1.01,0.9786623954772949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,32,power_law_1.01,2.053171157836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,32,balanced,0.7949706713358561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,32,balanced,0.665450652440389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,32,balanced,1.2535413106282551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,32,power_law_1.01,0.038975998759269714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,32,power_law_1.01,0.024172799289226533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,32,power_law_1.01,0.035078400373458864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,32,power_law_1.01,0.020531199872493744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,32,power_law_1.01,0.03409920036792755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,32,power_law_1.01,0.020883199572563172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,32,power_law_1.01,0.03727999925613403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,32,power_law_1.01,0.022278399765491487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,32,power_law_1.01,0.037036800384521486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,32,power_law_1.01,0.025491198897361754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,32,power_law_1.01,0.03951359987258911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,32,power_law_1.01,0.031699201464653014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,32,power_law_1.01,0.039155200123786926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,32,power_law_1.01,0.043968001008033754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,32,power_law_1.01,0.03925119936466217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,32,power_law_1.01,0.044064000248909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,32,power_law_1.01,0.039705601334571836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,32,power_law_1.01,0.039987200498580934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,32,power_law_1.01,0.049318400025367734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,32,balanced,1.0260319709777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,32,power_law_1.01,0.04030719995498657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,32,power_law_1.01,0.049568000435829165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,32,power_law_1.01,0.04149760007858276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,32,power_law_1.01,0.04933759868144989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,32,power_law_1.01,0.042156800627708435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,32,power_law_1.01,0.04869759976863861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,32,power_law_1.01,0.045311999320983884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,32,power_law_1.01,0.04716159999370575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,32,power_law_1.01,0.04892799854278564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,32,power_law_1.01,0.04483200013637543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,32,power_law_1.01,0.0521664023399353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,32,power_law_1.01,0.042559999227523806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,32,power_law_1.01,0.05550079941749573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,32,power_law_1.01,0.046028798818588255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,32,power_law_1.01,0.060627198219299315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,32,power_law_1.01,0.0462336003780365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,32,power_law_1.01,0.05111680030822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.01,0.07104640007019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.01,0.05578879714012146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.01,0.08131200075149536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.01,0.06888960003852844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.01,0.0869376003742218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.01,0.07740799784660339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.01,0.11690880060195923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.01,0.10510720014572143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.01,0.13203840255737304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.01,0.121561598777771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,32,power_law_1.2,0.06193280220031738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.01,0.17335679531097412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.01,0.217574405670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.01,0.19011199474334717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,32,power_law_1.2,0.03612160086631775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.01,0.3141632080078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.01,0.39594879150390627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.01,0.23854079246520996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,32,power_law_1.2,0.038431999087333676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,32,power_law_1.01,0.4704063892364502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,32,power_law_1.2,0.040006399154663086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.01,0.324070405960083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,32,power_law_1.2,0.04466559886932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,32,power_law_1.01,0.8554368019104004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.01,0.40572161674499513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,32,power_law_1.2,0.05403519868850708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,32,power_law_1.01,0.5779136180877685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,32,power_law_1.01,1.5287872314453126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,32,power_law_1.2,0.05441280007362366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,32,power_law_1.2,0.05560960173606873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,32,power_law_1.01,0.8356672286987304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,32,power_law_1.2,0.056428802013397214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,32,power_law_1.2,0.05678079724311828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,32,power_law_1.2,0.05868160128593445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,32,power_law_1.01,1.676416015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,32,power_law_1.2,0.06060799956321716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,32,power_law_1.2,0.06320000290870667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,32,power_law_1.2,0.06492159962654113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,32,power_law_1.2,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,32,power_law_1.2,0.06976640224456787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,32,power_law_1.2,0.07724800109863281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,32,power_law_1.2,0.030880001187324525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,32,balanced,2.478816032409668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,32,power_law_1.2,0.07880319952964783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,32,power_law_1.2,0.03115519881248474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,32,power_law_1.2,0.08565120100975036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,32,power_law_1.2,0.0328575998544693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,32,power_law_1.2,0.09132800102233887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,32,power_law_1.2,0.03544319868087768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,32,power_law_1.2,0.03590399920940399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,32,power_law_1.2,0.10962560176849365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,32,power_law_1.2,0.03580799996852875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,32,power_law_1.2,0.12394880056381226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,32,power_law_1.2,0.04051199853420258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,32,power_law_1.2,0.1575744032859802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,32,power_law_1.2,0.039673599600791934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,32,power_law_1.2,0.18785279989242554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,32,power_law_1.2,0.04104959964752197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,32,power_law_1.2,0.2676736116409302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,32,power_law_1.2,0.045510399341583255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,32,power_law_1.2,0.4969600200653076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,32,power_law_1.2,0.046988800168037415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,32,power_law_1.2,0.05006719827651977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,32,power_law_1.2,0.6450560092926025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,32,power_law_1.2,0.05074560046195984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,32,power_law_1.2,0.0590719997882843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,32,power_law_1.2,0.8695232391357421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,32,power_law_1.2,0.06044800281524658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,32,power_law_1.2,0.061964797973632815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,32,power_law_1.2,1.425119972229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,32,power_law_1.2,0.07479040026664734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,32,power_law_1.2,0.08269439935684204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,32,power_law_1.2,1.111961555480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,32,power_law_1.2,0.09765120148658753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,32,power_law_1.2,0.12380800247192383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,32,power_law_1.2,4.139449691772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,32,power_law_1.2,0.16580480337142944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,32,power_law_1.2,0.22296319007873536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,32,power_law_1.2,0.297491192817688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,32,power_law_1.2,0.42252798080444337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,32,power_law_1.2,0.6550399780273437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,32,power_law_1.2,0.7221695899963378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,32,power_law_1.2,1.181811237335205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,32,power_law_1.2,1.4721792221069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,32,power_law_1.2,3.8157310485839844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,32,power_law_1.2,0.03899520039558411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,32,power_law_1.2,0.03511039912700653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,32,power_law_1.2,0.03375360071659088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,32,power_law_1.2,0.03694719970226288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,32,power_law_1.2,0.03708159923553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,32,power_law_1.2,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,32,power_law_1.2,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,32,power_law_1.2,0.03940480053424835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,32,power_law_1.2,0.039750400185585025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,32,power_law_1.2,0.03956480026245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,32,power_law_1.2,0.04074240028858185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,32,power_law_1.2,0.042694398760795595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,32,power_law_1.2,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,32,power_law_1.2,0.046982398629188536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,32,power_law_1.2,0.05057280063629151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,32,power_law_1.2,0.022899200022220612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,32,power_law_1.2,0.05265920162200928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,32,power_law_1.2,0.019225600361824035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,32,power_law_1.2,0.020236800611019134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,32,power_law_1.2,0.05785599946975708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,32,power_law_1.2,0.021503999829292297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,32,power_law_1.2,0.06734079718589783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,32,power_law_1.2,0.02462719976902008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,32,power_law_1.2,0.031744000315666196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.2,0.07360000014305115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,32,power_law_1.2,0.043833601474761966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.2,0.09297919869422913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,32,power_law_1.2,0.04318720102310181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.2,0.0939840018749237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,32,power_law_1.2,0.04829440116882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.2,0.13112959861755372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,32,power_law_1.2,0.04696959853172302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.2,0.1610751986503601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,32,power_law_1.2,0.04653440117835998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.2,0.22868480682373046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,32,power_law_1.2,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.2,0.28827519416809083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,32,power_law_1.2,0.045337599515914914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.2,0.38483200073242185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.2,0.49534077644348146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,32,power_law_1.2,0.042444801330566405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,32,power_law_1.2,0.7465023994445801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,32,power_law_1.2,0.04301440119743347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,32,power_law_1.2,1.4513343811035155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,32,power_law_1.2,0.0409280002117157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,32,power_law_1.2,0.042342400550842284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,32,power_law_1.2,2.7556735992431642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,32,power_law_1.2,0.04885759949684143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.2,0.05591679811477661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.2,0.07011839747428894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.2,0.08056319952011108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.2,0.11315840482711792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.2,0.13228800296783447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.2,0.19875839948654175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.2,0.258022403717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.2,0.40111360549926756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.2,0.5082943916320801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,32,power_law_1.2,0.6505727767944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,32,power_law_1.2,1.0035327911376952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,32,power_law_1.2,2.0367807388305663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,16,balanced,0.026543999711672466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,16,balanced,0.023365333676338196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,16,balanced,0.020997333029905956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,16,balanced,0.023007998863856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,16,balanced,0.025018667181332905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,16,balanced,0.02536533276240031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,16,balanced,0.029194665451844532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,16,balanced,0.027295999228954315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,16,balanced,0.03349866718053818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,16,balanced,0.0335413341720899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,16,balanced,0.03327466547489166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,16,balanced,0.040250666439533234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,16,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,16,balanced,0.04601066807905833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,16,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,16,balanced,0.06201066573460897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,16,balanced,0.06227200229962667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,16,balanced,0.037402667105197906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,16,balanced,0.03719466676314672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,16,balanced,0.06628799935181935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,16,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,16,balanced,0.03949866692225138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,16,balanced,0.0888853371143341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,16,balanced,0.04125333329041799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,16,balanced,0.04186666508515676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,16,balanced,0.11005866527557373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,16,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,16,balanced,0.04153066625197729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,16,balanced,0.13980799913406372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,16,balanced,0.04278400043646494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,16,balanced,0.0421066681543986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,16,balanced,0.043840001026789345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,16,balanced,0.0986293355623881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,16,balanced,0.04357333481311798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,16,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,16,balanced,0.049679999550183616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,16,balanced,0.12166399757067363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,16,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,16,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,16,balanced,0.05190399785836538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,16,balanced,0.14412267009417215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,16,balanced,0.05862933397293091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,16,balanced,0.066170667608579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,16,balanced,0.06818133095900218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,16,balanced,0.2083146572113037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,16,balanced,0.09075733025868733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,16,balanced,0.104912002881368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,16,balanced,0.25828800598780316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,16,balanced,0.13262400031089783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,16,balanced,0.3686986764272054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,16,balanced,0.16261333227157593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,16,balanced,0.2131519913673401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,16,power_law_1.01,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,16,balanced,0.48104000091552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,16,power_law_1.01,0.04572800099849701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,16,power_law_1.01,0.04078719913959503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,16,power_law_1.01,0.046003198623657225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,16,power_law_1.01,0.048076799511909483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,16,power_law_1.01,0.052774399518966675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,16,power_law_1.01,0.05432959794998169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,16,power_law_1.01,0.05334399938583374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,16,power_law_1.01,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,16,power_law_1.01,0.05686399936676025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,16,power_law_1.01,0.05628160238265991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,16,power_law_1.01,0.0574720025062561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,16,power_law_1.01,0.056831997632980344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,16,balanced,0.26123199860254925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,16,power_law_1.01,0.05680000185966492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,16,power_law_1.01,0.0607807993888855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,16,power_law_1.01,0.06175360083580017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,16,power_law_1.01,0.0278656005859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,16,power_law_1.01,0.06592000126838685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,16,balanced,0.5962933301925659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,16,power_law_1.01,0.07267839908599853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,16,power_law_1.01,0.02030719965696335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.01,0.08240640163421631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.01,0.09869440197944641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,16,power_law_1.01,0.018956799805164338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.01,0.10907520055770874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,16,power_law_1.01,0.019859200716018675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.01,0.1300160050392151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,16,power_law_1.01,0.02152319997549057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.01,0.1677183985710144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,16,power_law_1.01,0.021887999773025513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.01,0.2041088104248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,16,power_law_1.01,0.022694399952888487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.01,0.2743168115615845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,16,power_law_1.01,0.024646399915218352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,16,power_law_1.01,0.0244159996509552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.01,0.38264319896697996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,16,power_law_1.01,0.026393601298332216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.01,0.5390848159790039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,16,power_law_1.01,0.031276801228523256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,16,power_law_1.01,0.5938496112823486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,16,balanced,0.3145280083020528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,16,power_law_1.01,0.03198719918727875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,16,power_law_1.01,0.03402239978313446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,16,power_law_1.01,0.9576383590698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,16,power_law_1.01,0.044844800233840944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,16,power_law_1.01,0.046316799521446225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,16,power_law_1.01,2.0517759323120117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,16,power_law_1.01,0.05002239942550659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,16,power_law_1.01,0.06247040033340454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,16,power_law_1.01,0.06769919991493226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.01,0.07856000065803528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.01,0.09650560021400452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.01,0.07196159958839417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.01,0.09200000166893005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.01,0.11052800416946411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,16,balanced,0.9425546328226725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.01,0.14524799585342407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.01,0.18744319677352905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,16,balanced,0.03366933266321818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.01,0.2642496109008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,16,balanced,0.03133333226044973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.01,0.33843839168548584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,16,balanced,0.0315733328461647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,16,balanced,0.03469333300987879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,16,power_law_1.01,0.39594879150390627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,16,balanced,0.05648533503214518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,16,balanced,0.055888002117474876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,16,power_law_1.01,0.6516223907470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,16,balanced,0.056789333621660866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,16,balanced,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,16,balanced,0.0565226674079895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,16,balanced,0.056101332108179726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,16,power_law_1.01,1.3173760414123534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,16,balanced,0.058517331878344216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,16,balanced,0.03332799921433131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,16,balanced,0.4713866710662842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,16,balanced,0.05784533421198527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,16,balanced,0.03373866776625315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,16,balanced,0.05682133138179779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,16,balanced,0.033733333150545754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,16,balanced,0.05961066484451294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,16,balanced,0.03882666677236557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,16,balanced,0.05957333246866862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,16,balanced,0.03876800090074539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,16,balanced,0.03965866565704346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,16,balanced,0.06423466900984447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,16,balanced,0.037808001041412354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,16,balanced,0.06406400104363759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,16,balanced,0.04048533240954081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,16,balanced,0.06443200012048085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,16,balanced,0.03940266619126002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,16,balanced,0.04148799926042557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,16,balanced,0.0681386689345042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,16,balanced,0.0417546679576238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,16,balanced,0.04075733323891958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,16,balanced,0.06857599814732869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,16,balanced,0.04358933369318644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,16,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,16,balanced,0.07540266712506612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,16,balanced,0.04770666857560476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,16,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,16,balanced,0.07696000238259633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,16,balanced,0.056202664971351624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,16,balanced,0.061205332477887474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,16,balanced,0.09567999839782715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,16,balanced,0.078575998544693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,16,balanced,0.08705066641171773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,16,balanced,0.10496000448862712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,16,balanced,0.11314133803049724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,16,balanced,0.1249066690603892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,16,balanced,0.12779733538627625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,16,balanced,1.838175932566325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,16,balanced,0.16341867049535116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,16,balanced,0.1565600037574768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,16,power_law_1.01,0.05618559718132019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,16,balanced,0.20957867304484049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,16,power_law_1.01,0.04933759868144989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,16,power_law_1.01,0.0440064013004303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,16,power_law_1.01,0.03592320084571839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,16,balanced,0.1975626746813456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,16,power_law_1.01,0.03917439877986908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,16,power_law_1.01,0.03252480030059814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,16,power_law_1.01,0.04261760115623474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,16,balanced,0.8885599772135416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,16,power_law_1.01,0.03157120048999786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,16,balanced,0.282368004322052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,16,power_law_1.01,0.0541375994682312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,16,power_law_1.01,0.03240320086479187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,16,power_law_1.01,0.03430399894714355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,16,power_law_1.01,0.05392640233039856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,16,power_law_1.01,0.03641600012779236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,16,power_law_1.01,0.03759360015392303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,16,power_law_1.01,0.05614719986915588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,16,power_law_1.01,0.03828479945659637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,16,power_law_1.01,0.055353599786758426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,16,balanced,0.23332800467809042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,16,power_law_1.01,0.03925119936466217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,16,power_law_1.01,0.05630720257759094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,16,power_law_1.01,0.03934719860553741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,16,power_law_1.01,0.05633919835090637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,16,power_law_1.01,0.04178560078144074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,16,power_law_1.01,0.05804799795150757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,16,power_law_1.01,0.047219198942184445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,16,power_law_1.01,0.048416000604629514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,16,power_law_1.01,0.05916799902915955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,16,power_law_1.01,0.046726399660110475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,16,power_law_1.01,0.06119040250778198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,16,balanced,0.3516960144042969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,16,power_law_1.01,0.05564799904823303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,16,power_law_1.01,0.06541439890861511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,16,power_law_1.01,0.06119040250778198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,16,power_law_1.01,0.0631168007850647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,16,power_law_1.01,0.06656640172004699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,16,power_law_1.01,0.0688256025314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,16,power_law_1.01,0.07137920260429383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,16,power_law_1.01,0.07991679906845092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,16,power_law_1.01,0.07808640003204345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,16,power_law_1.01,0.09676799774169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,16,power_law_1.01,0.08108159899711609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,16,power_law_1.01,0.10605440139770508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,16,power_law_1.01,0.09978880286216736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,16,power_law_1.01,0.14135680198669434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,16,power_law_1.01,0.10255359411239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,16,balanced,0.2820853392283122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,16,power_law_1.01,0.17448960542678832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,16,power_law_1.01,0.1350399971008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,16,balanced,0.4294666846593221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,16,power_law_1.01,0.24293119907379152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,16,power_law_1.01,0.16450560092926025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,16,power_law_1.01,0.3175168037414551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,16,power_law_1.01,0.44136958122253417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,16,power_law_1.01,0.20586879253387452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,16,power_law_1.01,0.5980095863342285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,16,power_law_1.01,0.2676800012588501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,16,power_law_1.01,0.6369855880737305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,16,power_law_1.01,0.3351680040359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,16,power_law_1.01,0.9722623825073242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,16,power_law_1.01,0.49059200286865234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,16,power_law_1.01,2.0602176666259764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,16,power_law_1.01,0.5710591793060302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,16,power_law_1.01,0.7647232055664063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,16,power_law_1.01,1.5116415977478028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,16,power_law_1.2,0.05235199928283692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,16,balanced,0.40492268403371173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,16,power_law_1.2,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,16,power_law_1.2,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,16,power_law_1.2,0.03700479865074158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,16,power_law_1.2,0.04038400053977966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,16,power_law_1.2,0.05340800285339355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,16,balanced,0.6555999914805094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,16,power_law_1.2,0.05554559826850891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,16,power_law_1.2,0.05608320236206055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,16,power_law_1.2,0.056435197591781616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,16,power_law_1.2,0.05675519704818725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,16,power_law_1.2,0.05697280168533325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,16,power_law_1.2,0.0585919976234436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,16,power_law_1.2,0.05953279733657837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,16,power_law_1.2,0.061926400661468504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,16,power_law_1.2,0.06594560146331788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,16,power_law_1.2,0.0688256025314331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,16,power_law_1.2,0.07587199807167053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,16,power_law_1.2,0.08121600151062011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,16,power_law_1.2,0.0897599995136261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,16,power_law_1.2,0.10335359573364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,16,power_law_1.2,0.11544959545135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,16,power_law_1.2,0.14969600439071656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,16,power_law_1.2,0.17907840013504028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,16,power_law_1.2,0.21989760398864747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,16,power_law_1.2,0.04028800129890442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,16,power_law_1.2,0.29374079704284667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,16,power_law_1.2,0.5198207855224609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,16,power_law_1.2,0.030368000268936157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,16,power_law_1.2,0.6234303951263428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,16,power_law_1.2,0.764851188659668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,16,power_law_1.2,0.029542401432991028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,16,power_law_1.2,0.03172479867935181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,16,power_law_1.2,1.4449407577514648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,16,power_law_1.2,0.03409920036792755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,16,power_law_1.2,2.589049530029297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,16,power_law_1.2,0.03699840009212494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,16,power_law_1.2,0.03749119937419891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,16,power_law_1.2,0.039315199851989745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,16,power_law_1.2,0.04031359851360321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,16,power_law_1.2,0.041119998693466185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,16,power_law_1.2,0.04428159892559051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,16,power_law_1.2,0.04442239999771118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,16,power_law_1.2,0.05010560154914856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,16,power_law_1.2,0.051769602298736575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,16,balanced,0.7363519668579102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,16,power_law_1.2,0.05870720148086548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,16,power_law_1.2,0.06076800227165222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,16,power_law_1.2,0.06276479959487916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,16,power_law_1.2,0.07207040190696716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,16,power_law_1.2,0.08135679960250855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,16,balanced,1.2449386914571126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,16,power_law_1.2,0.10047359466552734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,16,power_law_1.2,0.13189760446548462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,16,power_law_1.2,0.18410240411758422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,16,power_law_1.2,0.20023679733276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,16,power_law_1.2,0.2735487937927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,16,power_law_1.2,0.3647423982620239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,16,power_law_1.2,0.535097599029541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,16,power_law_1.2,0.7072896003723145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,16,power_law_1.2,0.9104640007019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,16,power_law_1.2,1.7399936676025392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,16,power_law_1.2,0.05666559934616089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,16,power_law_1.2,2.921561622619629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,16,power_law_1.2,0.039647999405860904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,16,power_law_1.2,0.042105600237846375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,16,power_law_1.2,0.04586879909038544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,16,power_law_1.2,0.04867840111255646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,16,power_law_1.2,0.05628799796104431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,16,power_law_1.2,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,16,power_law_1.2,0.05444480180740356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,16,power_law_1.2,0.05591040253639221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,16,power_law_1.2,0.056518399715423585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,16,power_law_1.2,0.057574397325515746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,16,power_law_1.2,0.0586624026298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,16,power_law_1.2,0.028460800647735596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,16,power_law_1.2,0.058329600095748904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,16,power_law_1.2,0.05846400260925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,16,power_law_1.2,0.020479999482631683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,16,power_law_1.2,0.06272000074386597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,16,power_law_1.2,0.020364800095558168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,16,power_law_1.2,0.06501759886741638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,16,power_law_1.2,0.020070399343967437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,16,power_law_1.2,0.06677119731903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,16,power_law_1.2,0.07850880026817322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,16,power_law_1.2,0.021247999370098115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.2,0.08728960156440735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,16,power_law_1.2,0.021612800657749176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.2,0.09986559748649597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,16,power_law_1.2,0.024051199853420257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.2,0.11731200218200684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,16,power_law_1.2,0.02462719976902008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.2,0.1640128016471863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,16,power_law_1.2,0.025625601410865784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,16,power_law_1.2,0.027590399980545043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.2,0.1957759976387024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,16,power_law_1.2,0.031385600566864014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.2,0.25681281089782715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,16,power_law_1.2,0.03311359882354736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.2,0.3509376049041748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,16,power_law_1.2,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.2,0.5571455955505371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,16,power_law_1.2,0.04575999975204468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,16,power_law_1.2,0.04987519979476929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.2,0.728550386428833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,16,power_law_1.2,0.05532159805297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,16,power_law_1.2,0.827507209777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,16,power_law_1.2,0.0633791983127594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,16,power_law_1.2,0.06869120001792908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,16,power_law_1.2,1.3707327842712402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.2,0.08176640272140503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.2,0.10083839893341065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,16,power_law_1.2,2.909561538696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.2,0.0722432017326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.2,0.09418879747390747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.2,0.1232640027999878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.2,0.15708800554275512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.2,0.2036288022994995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.2,0.27268478870391843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.2,0.3802623987197876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,16,power_law_1.2,0.47837438583374026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,16,power_law_1.2,0.7324800014495849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,16,power_law_1.2,1.4868224143981934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,8,balanced,0.021338666478792827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,8,balanced,0.020586666961510975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,8,balanced,0.019978666057189304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,8,balanced,0.0210506667693456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,8,balanced,0.036415999134381614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,8,balanced,0.023631999890009563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,8,balanced,0.02922666569550832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,8,balanced,0.03817066550254822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,8,balanced,0.03158933420976003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,8,balanced,0.037615999579429626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,8,balanced,0.029498666524887085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,8,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,8,balanced,0.03126399964094162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,8,balanced,0.0469760000705719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,8,balanced,0.033520000676314034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,8,balanced,0.04621333380540212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,8,balanced,0.03437866767247518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,8,balanced,0.047637333472569786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,8,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,8,balanced,0.047653332352638245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,8,balanced,0.039477333426475525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,8,balanced,0.04756799836953481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,8,balanced,0.047930667797724404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,8,balanced,0.04197333256403605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,8,balanced,0.048026666045188904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,8,balanced,0.043562665581703186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,8,balanced,0.0498933345079422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,8,balanced,0.04458666841189066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,8,balanced,0.04991999765237173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,8,balanced,0.06688533226648967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,8,balanced,0.05134400228659312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,8,balanced,0.055786664287249245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,8,balanced,0.07045333087444305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,8,balanced,0.054005334774653115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,8,balanced,0.07776533563931783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,8,balanced,0.057914664347966514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,8,balanced,0.10514133175214131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,8,balanced,0.06237333516279856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,8,balanced,0.07052266597747803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,8,balanced,0.08937600255012512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,8,balanced,0.07966933151086171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,8,balanced,0.11018133163452148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,8,balanced,0.08678399523099263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,8,balanced,0.11685333649317424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,8,balanced,0.12993599971135458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,8,power_law_1.01,0.05045120120048523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,8,balanced,0.14034666617711386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,8,power_law_1.01,0.048102399706840514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,8,power_law_1.01,0.04336639940738678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,8,balanced,0.18900267283121744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,8,power_law_1.01,0.04844799935817719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,8,power_law_1.01,0.05322239995002746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,8,power_law_1.01,0.0599295973777771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,8,power_law_1.01,0.06172159910202026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,8,balanced,0.18453333775202432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,8,power_law_1.01,0.05987200140953064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,8,power_law_1.01,0.06170880198478699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,8,balanced,0.22246400515238443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,8,power_law_1.01,0.061913597583770755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,8,power_law_1.01,0.06289920210838318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,8,power_law_1.01,0.06394879817962647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,8,power_law_1.01,0.06295040249824524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,8,power_law_1.01,0.06563839912414551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,8,balanced,0.22037333250045776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,8,power_law_1.01,0.06941440105438232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,8,power_law_1.01,0.07111679911613464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,8,power_law_1.01,0.0201664000749588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,8,power_law_1.01,0.07713279724121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,8,power_law_1.01,0.08261119723320007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,8,power_law_1.01,0.018963199853897095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.01,0.09319040179252625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,8,balanced,0.3141760031382243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,8,power_law_1.01,0.018822400271892546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.01,0.1160256028175354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,8,power_law_1.01,0.01940480023622513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.01,0.12424319982528687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,8,power_law_1.01,0.020127999782562255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.01,0.16392960548400878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,8,power_law_1.01,0.022316800057888032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.01,0.20046720504760743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,8,balanced,0.30689066648483276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.01,0.2626431941986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,8,power_law_1.01,0.024991999566555022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,8,power_law_1.01,0.022784000635147093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.01,0.34698240756988524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,8,power_law_1.01,0.026041600108146667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.01,0.5095680236816407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,8,power_law_1.01,0.02887679934501648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.01,0.6490560054779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,8,power_law_1.01,0.025939199328422546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,8,power_law_1.01,0.7462528228759766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,8,power_law_1.01,0.027718400955200194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,8,power_law_1.01,0.030432000756263733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,8,power_law_1.01,1.2280960083007812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,8,power_law_1.01,0.03450239896774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,8,balanced,0.407968004544576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,8,power_law_1.01,0.03875199854373932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,8,power_law_1.01,0.04131839871406555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,8,power_law_1.01,2.438380813598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,8,power_law_1.01,0.0510591983795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,8,power_law_1.01,0.06857600212097167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.01,0.07461119890213012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,8,balanced,0.3909440040588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.01,0.09286400079727172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.01,0.07280640006065368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.01,0.0911616027355194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.01,0.10675200223922729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.01,0.14282239675521852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.01,0.17825280427932738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.01,0.2509567975997925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.01,0.3266304016113281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,8,balanced,0.5074559847513834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,8,power_law_1.01,0.39310081005096437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,8,power_law_1.01,0.6184256076812744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,8,power_law_1.01,1.1541055679321288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,8,balanced,0.47890667120615643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,8,balanced,0.02942399928967158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,8,balanced,0.031184000273545582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,8,balanced,0.033258666594823204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,8,balanced,0.035360001027584076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,8,balanced,0.05782400071620941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,8,balanced,0.057760000228881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,8,balanced,0.059194669127464294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,8,balanced,0.8024000326792399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,8,balanced,0.05985066791375478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,8,balanced,0.057775999108950295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,8,balanced,0.060090666015942894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,8,balanced,0.05787733197212219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,8,balanced,0.05929600199063619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,8,balanced,0.061850666999816895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,8,balanced,0.7344319820404053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,8,balanced,0.062218666076660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,8,balanced,0.06598400076230367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,8,balanced,0.06778666873772939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,8,balanced,0.029418667157491047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,8,balanced,0.06828799843788147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,8,balanced,0.03161066770553589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,8,balanced,0.07222400108973186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,8,balanced,0.03332266708215078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,8,balanced,0.03370666752258936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,8,balanced,0.0816480020682017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,8,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,8,balanced,0.09180266658465068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,8,balanced,0.03974399964014689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,8,balanced,0.04174399872620901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,8,balanced,0.09598400195439656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,8,balanced,0.04159999887148539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,8,balanced,0.041152000427246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,8,balanced,0.041802664597829185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,8,balanced,0.11029866337776184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,8,balanced,0.04190933207670847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,8,balanced,0.04428266485532125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,8,balanced,0.04563199977080027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,8,balanced,0.04598933458328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,8,balanced,0.12892267107963562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,8,balanced,0.053802669048309326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,8,balanced,0.051738664507865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,8,balanced,0.0540533314148585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,8,balanced,0.05991999804973602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,8,balanced,0.16261866688728333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,8,balanced,0.07492800056934357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,8,balanced,0.09273599584897359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,8,balanced,0.09896533687909444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,8,balanced,0.19765333334604898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,8,balanced,0.1274720033009847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,8,balanced,0.1585760017236074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,8,balanced,0.26306132475535077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,8,balanced,0.2056480050086975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,8,balanced,1.5786773363749187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,8,power_law_1.01,0.03699840009212494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,8,power_law_1.01,0.03530240058898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,8,balanced,0.2563146750132243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,8,balanced,1.4253652890523274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,8,power_law_1.01,0.03336319923400879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,8,power_law_1.01,0.03731200098991394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,8,power_law_1.01,0.03978239893913269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,8,balanced,0.32443199555079144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,8,power_law_1.01,0.05182719826698303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,8,power_law_1.01,0.05257599949836731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,8,power_law_1.01,0.05505920052528381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,8,power_law_1.01,0.05539199709892273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,8,power_law_1.01,0.05546240210533142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,8,power_law_1.01,0.05785599946975708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,8,balanced,0.3554399808247884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,8,power_law_1.01,0.05959039926528931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,8,power_law_1.01,0.060288000106811526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,8,power_law_1.01,0.06207360029220581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,8,power_law_1.01,0.06647040247917176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,8,power_law_1.01,0.07038080096244811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,8,power_law_1.01,0.0720192015171051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,8,power_law_1.01,0.08193920254707336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,8,power_law_1.01,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,8,power_law_1.01,0.08124799728393554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,8,power_law_1.01,0.10621440410614014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,8,balanced,0.3980213403701782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,8,power_law_1.01,0.031040000915527343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,8,power_law_1.01,0.11811840534210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,8,power_law_1.01,0.031231999397277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,8,power_law_1.01,0.1456447958946228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,8,power_law_1.01,0.034643200039863584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,8,power_law_1.01,0.17341439723968505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,8,power_law_1.01,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,8,power_law_1.01,0.22382719516754152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,8,balanced,0.45856531461079914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,8,power_law_1.01,0.038412800431251524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,8,power_law_1.01,0.03989759981632233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,8,power_law_1.01,0.284006404876709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,8,power_law_1.01,0.039801600575447085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,8,power_law_1.01,0.37827839851379397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,8,power_law_1.01,0.04202240109443665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,8,power_law_1.01,0.5116735935211182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,8,power_law_1.01,0.0433023989200592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,8,power_law_1.01,0.5515007972717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,8,power_law_1.01,0.045126399397850035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,8,power_law_1.01,0.04783360064029694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,8,power_law_1.01,1.0756223678588868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,8,power_law_1.01,0.05438079833984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,8,power_law_1.01,0.04917120039463043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,8,power_law_1.01,1.9431039810180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,8,power_law_1.01,0.05431680083274841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,8,power_law_1.01,0.05678719878196716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,8,power_law_1.01,0.063155198097229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,8,power_law_1.2,0.03730559945106506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,8,power_law_1.01,0.07376639842987061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,8,power_law_1.01,0.08263040184974671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,8,power_law_1.01,0.09864959716796876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,8,power_law_1.2,0.03129599988460541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,8,power_law_1.2,0.03349120020866394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,8,power_law_1.01,0.11934080123901367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,8,balanced,0.5647626717885336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,8,power_law_1.01,0.15701760053634645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,8,power_law_1.2,0.038540801405906676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,8,power_law_1.01,0.1903807997703552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,8,power_law_1.2,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,8,power_law_1.01,0.2667072057723999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,8,balanced,0.5941439867019653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,8,power_law_1.2,0.0553600013256073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,8,power_law_1.01,0.3583872079849243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,8,power_law_1.2,0.055232000350952146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,8,power_law_1.01,0.4632575988769531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,8,power_law_1.2,0.057183998823165896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,8,power_law_1.01,0.6176383972167969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,8,power_law_1.2,0.056480002403259275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,8,power_law_1.2,0.0573311984539032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,8,power_law_1.01,0.7461696147918702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,8,power_law_1.2,0.05848960280418396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,8,power_law_1.01,1.1545151710510253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,8,power_law_1.2,0.061312001943588254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,8,power_law_1.2,0.06193280220031738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,8,power_law_1.2,0.06419839859008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,8,power_law_1.01,2.4600576400756835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,8,power_law_1.2,0.06951680183410644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,8,power_law_1.2,0.07268480062484742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,8,power_law_1.2,0.07564160227775574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,8,power_law_1.2,0.08300799727439881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,8,power_law_1.2,0.031814399361610415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,8,power_law_1.2,0.09715200066566468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,8,power_law_1.2,0.1131775975227356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,8,power_law_1.2,0.029094401001930236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,8,power_law_1.2,0.13322240114212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,8,balanced,0.8679786523183187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,8,power_law_1.2,0.029651200771331786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,8,power_law_1.2,0.16692479848861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,8,power_law_1.2,0.21427199840545655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,8,power_law_1.2,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,8,power_law_1.2,0.2811903953552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,8,power_law_1.2,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,8,power_law_1.2,0.34053759574890136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,8,power_law_1.2,0.040531200170516965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,8,power_law_1.2,0.5065855979919434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,8,power_law_1.2,0.039980798959732056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,8,power_law_1.2,0.7527743816375733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,8,power_law_1.2,0.04208639860153198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,8,power_law_1.2,0.8555520057678223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,8,power_law_1.2,0.043263998627662656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,8,power_law_1.2,1.1401280403137206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,8,power_law_1.2,0.044223999977111815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,8,power_law_1.2,0.04764159917831421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,8,power_law_1.2,3.2753726959228517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,8,power_law_1.2,0.052908802032470705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,8,power_law_1.2,0.05482879877090454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,8,power_law_1.2,0.053574401140213015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,8,power_law_1.2,0.059494400024414064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,8,power_law_1.2,0.06231039762496948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,8,power_law_1.2,0.0686784029006958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,8,power_law_1.2,0.07715200185775757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,8,power_law_1.2,0.09724799990653991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,8,power_law_1.2,0.10990079641342163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,8,power_law_1.2,0.14531199932098388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,8,power_law_1.2,0.1708351969718933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,8,balanced,1.1053706804911296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,8,power_law_1.2,0.20647680759429932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,8,power_law_1.2,0.30989439487457277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,8,power_law_1.2,0.36693758964538575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,8,power_law_1.2,0.5519807815551758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,8,power_law_1.2,0.7257599830627441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,8,power_law_1.2,1.1730303764343262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,8,power_law_1.2,1.5269824028015138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,8,power_law_1.2,3.992979049682617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,8,balanced,1.665818691253662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,8,power_law_1.2,0.04952960014343262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,8,power_law_1.2,0.041196799278259276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,8,power_law_1.2,0.043315199017524716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,8,power_law_1.2,0.021196800470352172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,8,power_law_1.2,0.0473471999168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,8,power_law_1.2,0.05051519870758057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,8,power_law_1.2,0.0190528005361557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,8,power_law_1.2,0.0615231990814209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,8,power_law_1.2,0.019097599387168884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,8,power_law_1.2,0.061894398927688596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,8,power_law_1.2,0.06087679862976074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,8,power_law_1.2,0.019859200716018675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,8,power_law_1.2,0.06281599998474122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,8,power_law_1.2,0.021299199759960176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,8,power_law_1.2,0.06314880251884461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,8,power_law_1.2,0.02263679951429367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,8,power_law_1.2,0.06264320015907288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,8,power_law_1.2,0.027481600642204285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,8,power_law_1.2,0.06499199867248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,8,power_law_1.2,0.06416000127792358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,8,power_law_1.2,0.022944000363349915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,8,power_law_1.2,0.06674559712409973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,8,power_law_1.2,0.025222399830818178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,8,power_law_1.2,0.0711359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,8,power_law_1.2,0.02826879918575287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,8,power_law_1.2,0.026553601026535034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,8,power_law_1.2,0.07286400198936463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,8,power_law_1.2,0.030368000268936157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,8,power_law_1.2,0.07869439721107482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,8,power_law_1.2,0.033062401413917544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,8,power_law_1.2,0.08701440095901489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,8,power_law_1.2,0.03558399975299835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.2,0.10019199848175049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,8,power_law_1.2,0.04032639861106872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.2,0.1164736032485962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,8,power_law_1.2,0.04399999976158142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.2,0.13122559785842897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,8,power_law_1.2,0.053625601530075076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.2,0.16915199756622315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,8,power_law_1.2,0.06977279782295227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.2,0.19613440036773683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.2,0.07439360022544861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.2,0.29068799018859864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.2,0.09612159729003907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.2,0.39377920627593993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.2,0.07431679964065552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.2,0.5368192195892334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.2,0.09042559862136841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.2,0.7518208026885986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.2,0.1122431993484497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,8,power_law_1.2,1.108512020111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.2,0.1464192032814026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,8,power_law_1.2,1.775040054321289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.2,0.1897536039352417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.2,0.26659839153289794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,8,power_law_1.2,3.475564956665039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.2,0.3492608070373535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,8,power_law_1.2,0.42044801712036134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,8,power_law_1.2,0.6822015762329101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,8,power_law_1.2,1.347212791442871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,4,balanced,0.029498666524887085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,4,balanced,0.03130666663249334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,4,balanced,0.033386667569478355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,4,balanced,0.039605334401130676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,4,balanced,0.059903999169667564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,4,balanced,0.05994133154551188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,4,balanced,0.05994133154551188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,4,balanced,0.06196266909440359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,4,balanced,0.061994666854540505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,4,balanced,0.060506666700045265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,4,balanced,0.06228800117969513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,4,balanced,0.062309334675470986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,4,balanced,0.06437333424886067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,4,balanced,0.06677866478761037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,4,balanced,0.07147199908892314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,4,balanced,0.07342400153477986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,4,balanced,0.07559466858704884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,4,balanced,0.08239999910195668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,4,balanced,0.08666132887204488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,4,balanced,0.10506133238474528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,4,balanced,0.11413333813349406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,4,balanced,0.1465013325214386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,4,balanced,0.17733333508173624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,4,balanced,0.2299626668294271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,4,balanced,0.02958400050799052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,4,balanced,0.03146133323510488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,4,balanced,0.0335413341720899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,4,balanced,0.29313600063323975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,4,power_law_1.01,0.038335999846458434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,4,balanced,0.037418665985266365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,4,balanced,0.04554666578769684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,4,balanced,0.047338664531707764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,4,power_law_1.01,0.031744000315666196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,4,balanced,0.045738667249679565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,4,power_law_1.01,0.036211198568344115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,4,balanced,0.04628799855709076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,4,power_law_1.01,0.03871360123157501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,4,balanced,0.04590400060017904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,4,power_law_1.01,0.04161919951438904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,4,balanced,0.04772266745567322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,4,power_law_1.01,0.05612159967422485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,4,balanced,0.047925333182017006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,4,balanced,0.049728001157442726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,4,power_law_1.01,0.05502079725265503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,4,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,4,balanced,0.41152532895406085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,4,balanced,0.05226133267084757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,4,power_law_1.01,0.058271998167037965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,4,balanced,0.06461333235104878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,4,power_law_1.01,0.05862399935722351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,4,balanced,0.06193066636721293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,4,power_law_1.01,0.05912320017814636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,4,balanced,0.06431999802589417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,4,power_law_1.01,0.06170240044593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,4,power_law_1.01,0.06406400203704835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,4,balanced,0.07446399827798207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,4,power_law_1.01,0.06641280055046081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,4,balanced,0.0844586690266927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,4,power_law_1.01,0.07013760209083557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,4,power_law_1.01,0.077702397108078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,4,balanced,0.1104906698067983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,4,power_law_1.01,0.08229119777679443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,4,power_law_1.01,0.08860160112380981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,4,power_law_1.01,0.09071999788284302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,4,balanced,0.12403200070063274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,4,power_law_1.01,0.10243840217590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,4,balanced,0.5241706768671671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,4,power_law_1.01,0.1183359980583191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,4,power_law_1.01,0.03312000036239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,4,power_law_1.2,0.034694400429725644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,4,power_law_1.01,0.1351359963417053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,4,balanced,0.16849066813786825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,4,power_law_1.01,0.0322816014289856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,4,power_law_1.01,0.1778880000114441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,4,power_law_1.01,0.03223679959774017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,4,power_law_1.2,0.03059200048446655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,4,power_law_1.01,0.20464000701904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,4,power_law_1.01,0.036294400691986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,4,power_law_1.2,0.03550080060958862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,4,power_law_1.01,0.28557438850402833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,4,power_law_1.01,0.03841919898986816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,4,power_law_1.2,0.036287999153137206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,4,power_law_1.01,0.04469119906425476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,4,power_law_1.01,0.3459007978439331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,4,balanced,0.20258132616678873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,4,power_law_1.2,0.04151679873466492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,4,power_law_1.01,0.045363199710845944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,4,power_law_1.2,0.055641597509384154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,4,power_law_1.01,0.5615488052368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,4,power_law_1.01,0.045516800880432126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,4,power_law_1.2,0.05623040199279785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,4,power_law_1.01,0.048102399706840514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,4,power_law_1.2,0.05841919779777527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,4,balanced,0.28069865703582764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,4,power_law_1.01,0.6883776187896729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,4,power_law_1.01,0.048614400625228885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,4,power_law_1.2,0.05849599838256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,4,power_law_1.01,0.05259519815444946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,4,power_law_1.2,0.060096001625061034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,4,power_law_1.01,0.8354047775268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,4,power_law_1.01,0.05556480288505554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,4,balanced,0.6440373261769613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,4,power_law_1.2,0.06376320123672485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,4,power_law_1.01,0.059596800804138185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,4,power_law_1.01,1.2349696159362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,4,power_law_1.2,0.032979199290275575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,4,power_law_1.2,0.06479359865188598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,4,power_law_1.01,0.05868160128593445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,4,power_law_1.2,0.06529279947280883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,4,power_law_1.01,0.06436480283737182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,4,power_law_1.2,0.03175039887428284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,4,power_law_1.2,0.06906239986419678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,4,power_law_1.01,2.5872320175170898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,4,power_law_1.01,0.06800000071525573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,4,power_law_1.2,0.03258880078792572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,4,power_law_1.2,0.07493759989738465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,4,power_law_1.01,0.07390720248222352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,4,power_law_1.2,0.03455359935760498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,4,power_law_1.2,0.08115199804306031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,4,balanced,0.360149343808492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,4,power_law_1.01,0.09185280203819275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,4,power_law_1.2,0.03805440068244934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,4,power_law_1.01,0.09886080026626587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,4,power_law_1.2,0.04472320079803467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,4,power_law_1.01,0.12296960353851319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,4,power_law_1.2,0.08227840065956116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,4,power_law_1.01,0.14880000352859496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,4,power_law_1.2,0.0446399986743927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,4,power_law_1.01,0.19283839464187622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,4,power_law_1.2,0.09244800209999085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,4,power_law_1.2,0.046623998880386354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,4,power_law_1.01,0.24187519550323486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,4,power_law_1.2,0.048895999789237976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,4,power_law_1.01,0.3594367980957031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,4,power_law_1.2,0.10004479885101318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,4,power_law_1.2,0.04960640072822571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,4,power_law_1.01,0.4251264095306396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,4,power_law_1.2,0.12815359830856324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,4,power_law_1.2,0.05495679974555969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,4,power_law_1.01,0.6235072135925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,4,power_law_1.2,0.05707520246505737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,4,power_law_1.2,0.14179840087890624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,4,power_law_1.01,0.8780415534973145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,4,power_law_1.2,0.06119679808616638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,4,balanced,0.5227359930674235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,4,power_law_1.2,0.17322239875793458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,4,power_law_1.01,1.0663935661315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,4,power_law_1.2,0.05978879928588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,4,power_law_1.2,0.22605440616607667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,4,power_law_1.01,1.6076799392700196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,4,power_law_1.2,0.0673792004585266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,4,power_law_1.2,0.3832128047943115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,4,power_law_1.2,0.07026559710502625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,4,power_law_1.2,0.41071357727050783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,4,power_law_1.01,3.3289726257324217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,4,power_law_1.2,0.0758463978767395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,4,power_law_1.2,0.6849728107452393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,4,power_law_1.2,0.0920639991760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,4,power_law_1.2,0.7981823921203614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,4,power_law_1.2,0.1028607964515686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,4,power_law_1.2,0.12712960243225097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,4,power_law_1.2,1.2087807655334473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,4,power_law_1.2,0.15456639528274535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,4,balanced,0.9780746301015218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,4,power_law_1.2,1.5557056427001954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,4,power_law_1.2,0.2073728084564209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,4,power_law_1.2,0.29217278957366943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,4,power_law_1.2,3.245907211303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,4,power_law_1.2,0.39792640209198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,4,power_law_1.2,0.5308479785919189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,4,balanced,0.6757013003031412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,4,power_law_1.2,0.7578623771667481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,4,power_law_1.2,0.9616000175476074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,4,power_law_1.2,1.1998784065246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,4,power_law_1.2,2.211212730407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,4,power_law_1.2,3.734912109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,4,balanced,0.8443093299865723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,4,balanced,1.3109599749247234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,4,balanced,1.8241599400838215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,2,balanced,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,2,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,2,balanced,0.037802666425704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,2,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,2,balanced,0.07019199927647908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,2,balanced,0.07050133248170216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,2,balanced,0.07230933507283528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,2,balanced,0.0721973329782486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,2,balanced,0.07083199918270111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,2,balanced,0.07246399919191997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,2,balanced,0.07431999842325847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,2,balanced,0.07500266532103221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,2,balanced,0.07797333101431529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,2,balanced,0.07840000092983246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,2,balanced,0.08472533027331035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,2,balanced,0.08678932984670003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,2,balanced,0.0892639954884847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,2,balanced,0.10215999682744344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,2,balanced,0.11562666296958923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,2,balanced,0.03525333354870478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,2,balanced,0.1378773351510366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,2,balanced,0.035786665976047516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,2,balanced,0.0377813329299291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,2,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,2,balanced,0.1535306672255198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,2,balanced,0.056032001972198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,2,balanced,0.056048000852266945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,2,balanced,0.0547626664241155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,2,balanced,0.05604266623655955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,2,balanced,0.21573867400487265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,2,balanced,0.05817066629727682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,2,balanced,0.058229332168896995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,2,balanced,0.060191998879114784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,2,balanced,0.0620959997177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,2,balanced,0.261957327524821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,2,balanced,0.06494399905204773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,2,balanced,0.06939200063546498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,2,balanced,0.07453866799672444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,2,balanced,0.07861333092053731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,2,balanced,0.08520533641179402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,2,balanced,0.37455467383066815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,2,balanced,0.1029866635799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,2,balanced,0.11989866693814595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,2,balanced,0.144186665614446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,4,balanced,2.5402933756510415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,2,balanced,0.1693013310432434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,2,balanced,0.4663306474685669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,2,balanced,0.23883734146753946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,2,balanced,0.29825599988301593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,2,balanced,0.6746133168538412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,2,balanced,0.43406931559244794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,2,balanced,0.5490666627883911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,2,power_law_1.01,0.03213439881801605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,2,power_law_1.01,0.032601600885391234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,2,balanced,0.8663307030995687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,2,power_law_1.01,0.034790399670600894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,2,power_law_1.01,0.04221439957618713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,2,power_law_1.01,0.04704639911651611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,2,power_law_1.01,0.06413440108299255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,2,balanced,0.8100533485412598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,2,power_law_1.01,0.06906880140304565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,2,power_law_1.01,0.0695039987564087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,2,power_law_1.01,0.0711679995059967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,2,power_law_1.01,0.07193599939346314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,2,power_law_1.01,0.0731328010559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,2,power_law_1.01,0.07719039916992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,2,power_law_1.01,0.08046079874038696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,2,power_law_1.01,0.082259202003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,2,power_law_1.01,0.0901311993598938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,2,power_law_1.01,0.09342719912528992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,2,power_law_1.01,0.09447680115699768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,2,balanced,1.070186694463094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,2,power_law_1.01,0.11086080074310303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,2,power_law_1.2,0.03127039968967438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,2,power_law_1.01,0.03088639974594116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,2,power_law_1.01,0.12659840583801268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,2,balanced,1.0562667051951091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,2,power_law_1.2,0.031046399474143983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,2,power_law_1.01,0.0324864000082016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,2,power_law_1.01,0.15640319585800172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,2,power_law_1.01,0.033228799700737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,2,power_law_1.01,0.2012415885925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,2,power_law_1.2,0.0341376006603241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,2,power_law_1.01,0.24878079891204835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,2,power_law_1.01,0.039103999733924866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,2,power_law_1.2,0.03950720131397247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,2,power_law_1.2,0.030924800038337707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,2,power_law_1.01,0.3098047971725464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,2,power_law_1.01,0.04238080084323883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,2,power_law_1.01,0.4186111927032471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,2,power_law_1.2,0.04268800020217896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,2,power_law_1.01,0.0542527973651886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,2,power_law_1.01,0.5200640201568604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,2,power_law_1.2,0.031116798520088196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,2,power_law_1.2,0.0651199996471405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,2,power_law_1.01,0.0542527973651886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,2,power_law_1.01,0.7075263977050781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,2,power_law_1.2,0.06761599779129028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,2,power_law_1.01,0.056383997201919556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,2,power_law_1.2,0.032979199290275575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,2,power_law_1.01,0.9558015823364258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,2,power_law_1.2,0.07009279727935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,2,power_law_1.01,0.06087039709091187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,2,power_law_1.2,0.037510401010513304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,2,power_law_1.01,1.1399552345275878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,2,power_law_1.2,0.07171840071678162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,2,power_law_1.01,0.06177279949188232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,2,power_law_1.2,0.04193280041217804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,2,power_law_1.01,1.7123647689819337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,2,power_law_1.2,0.07232000231742859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,2,power_law_1.2,0.05448319911956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,2,power_law_1.01,0.06282879710197449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,2,power_law_1.2,0.07416960000991821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,2,power_law_1.01,3.5094463348388674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,2,balanced,1.306949297587077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,2,power_law_1.2,0.05633280277252197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,2,power_law_1.01,0.06721919775009155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,2,balanced,1.6431999206542969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,2,power_law_1.2,0.07719039916992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,2,power_law_1.2,0.08278399705886841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,2,power_law_1.2,0.05857279896736145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,2,power_law_1.01,0.07093120217323304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,2,power_law_1.2,0.08490880131721497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,2,power_law_1.2,0.061267197132110596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,2,power_law_1.01,0.07132800221443177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,2,power_law_1.2,0.089683198928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,2,power_law_1.2,0.06230400204658508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,2,power_law_1.01,0.0800704002380371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,2,power_law_1.2,0.09623680114746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,2,power_law_1.2,0.06521599888801574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,2,power_law_1.01,0.08410879969596863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,2,power_law_1.2,0.09609599709510804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,2,power_law_1.2,0.06887680292129517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,2,power_law_1.01,0.09539200067520141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,2,power_law_1.2,0.11176320314407348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,2,power_law_1.2,0.07230079770088196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,2,power_law_1.01,0.11063679456710815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,2,power_law_1.2,0.12867200374603271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,2,power_law_1.01,0.13401600122451782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,2,power_law_1.2,0.07352319955825806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,2,power_law_1.2,0.16186879873275756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,2,power_law_1.2,0.07976959943771363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,2,power_law_1.01,0.15909759998321532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,2,power_law_1.2,0.08650879859924317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,2,power_law_1.01,0.20016000270843506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,2,power_law_1.2,0.19067519903182983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,2,power_law_1.2,0.09633280038833618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,2,power_law_1.01,0.27022080421447753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,2,power_law_1.2,0.251692795753479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,2,power_law_1.2,0.11111680269241334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,2,power_law_1.01,0.3434623956680298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,2,power_law_1.2,0.32739839553833006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,2,power_law_1.2,0.13519999980926514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,2,power_law_1.01,0.4829696178436279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,2,power_law_1.2,0.16565120220184326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,2,power_law_1.2,0.4226687908172607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,2,power_law_1.2,0.2041408061981201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,2,power_law_1.01,0.6039743900299073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,2,power_law_1.2,0.5591743946075439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,2,power_law_1.2,0.26803200244903563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,2,power_law_1.01,0.8755135536193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,2,power_law_1.2,0.8150336265563964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,2,balanced,2.0409119923909507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,2,power_law_1.2,0.3488447904586792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,2,power_law_1.01,1.2042431831359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,2,power_law_1.2,1.0857728004455567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,2,power_law_1.2,0.4947711944580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,2,power_law_1.01,1.4651391983032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,2,power_law_1.2,1.160204792022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,2,power_law_1.2,0.635097599029541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,2,power_law_1.01,2.189414405822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,2,power_law_1.2,0.908358383178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,2,power_law_1.2,2.1557504653930666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,2,power_law_1.2,1.1827520370483398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,2,power_law_1.01,4.403155136108398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,2,power_law_1.2,4.015840148925781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,2,power_law_1.2,1.6089727401733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,2,power_law_1.2,2.6029375076293944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,2,power_law_1.2,5.089324951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,2,balanced,3.1566667556762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,2,balanced,4.004026730855306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,1,balanced,0.03606399893760681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,1,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,1,balanced,0.04295999805132548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,1,balanced,0.053488001227378845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,1,balanced,0.08469333251317342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,1,balanced,0.08859200278917949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,1,balanced,0.08861333131790161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,1,balanced,0.08909866213798523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,1,balanced,0.08962133526802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,1,balanced,0.09091732899347942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,1,balanced,0.09303999940554301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,1,balanced,0.09579199552536011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,1,balanced,0.0974133312702179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,1,balanced,0.10110400120417277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,1,balanced,0.10704533259073894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,1,balanced,0.10983999570210774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,1,balanced,0.1239946683247884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,1,balanced,0.1387999951839447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,1,balanced,0.17178134123484293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,1,balanced,0.2048906683921814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,1,balanced,0.24970134099324545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,1,balanced,0.37278934319814044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,1,balanced,0.4274666706720988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,1,balanced,0.616592009862264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,1,balanced,0.7977706591288248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,1,balanced,1.1617226600646973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,1,balanced,1.497477372487386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,1,balanced,0.03570133447647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,1,balanced,0.037578667203585304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,1,balanced,0.042447999119758606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,1,balanced,0.05376000205675761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,1,balanced,0.06971199810504913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,1,balanced,0.07235733171304067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,1,balanced,1.873301347096761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,1,balanced,0.07283199826876323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,1,balanced,0.07422400017579396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,1,balanced,0.07584533095359802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,1,balanced,0.07659733295440674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,32,balanced,0.0810346653064092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,1,balanced,0.08040533463160197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,32,balanced,0.04773333172003428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,1,balanced,0.08385599652926128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,32,balanced,0.05387733379999796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,32,balanced,0.04972266654173533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,1,balanced,0.0885706643263499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,32,balanced,0.05866666634877523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,32,balanced,0.04879466692606608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,1,balanced,0.09862400094668071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,32,balanced,0.08149866759777069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,32,balanced,0.053786665201187134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,32,balanced,0.07856533428033192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,1,balanced,0.10622933506965637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,32,balanced,0.10755200187365214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,32,balanced,0.08490133285522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,32,balanced,0.11727466185887654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,1,balanced,0.11263466874758403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,32,balanced,0.08543466528256734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,32,balanced,0.11754666765530904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,32,balanced,0.08646933237711589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,1,balanced,0.12290133039156596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,32,balanced,0.11905066172281902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,32,balanced,0.08810133735338847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,32,balanced,0.11753599842389424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,32,balanced,0.08731733759244283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,1,balanced,0.1474453310171763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,32,balanced,0.11962667107582092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,32,balanced,0.08895466725031535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,32,balanced,0.1195146640141805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,32,balanced,0.08931733171145122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,1,balanced,0.17484800020853677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,32,balanced,0.09013332923253377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,32,balanced,0.11962667107582092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,32,balanced,0.0926080048084259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,32,balanced,0.1213653286298116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,1,balanced,0.22403732935587564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,32,balanced,0.09778133034706116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,32,balanced,0.12198932965596516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,32,balanced,0.09922132889429729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,32,balanced,0.1244533360004425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,1,power_law_1.01,0.03375999927520752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,1,balanced,0.26393065849939984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,32,balanced,0.10214933753013611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,32,balanced,0.12746133406956991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,32,balanced,0.11145599683125813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,1,power_law_1.01,0.03624320030212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,32,balanced,0.12959466377894083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,1,power_law_1.01,0.04129279851913452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,32,balanced,0.13571199774742126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,1,balanced,0.40593600273132324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,32,balanced,0.11493866642316182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,1,power_law_1.01,0.04849919974803925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,32,balanced,0.1365493337313334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,1,power_law_1.01,0.05727999806404114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,32,balanced,0.13940266768137613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,1,power_law_1.01,0.0791167974472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,1,balanced,2.8978614807128906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,32,balanced,0.16750399271647134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,1,power_law_1.01,0.08062719702720642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,1,balanced,0.4925440152486165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,1,power_law_1.01,0.08327040076255798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,32,balanced,0.14595199624697366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,1,power_law_1.01,0.08615040183067321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,32,balanced,0.1564479966958364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,1,power_law_1.01,0.08748800158500672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,1,power_law_1.01,0.09173759818077087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,32,balanced,0.19100266695022583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,1,power_law_1.01,0.09409279823303222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,32,balanced,0.21854400634765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,1,balanced,0.7375360329945883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,1,power_law_1.01,0.09905920028686524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,1,power_law_1.01,0.10594559907913208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,1,power_law_1.01,0.11554559469223022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,32,balanced,0.208570659160614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,1,power_law_1.01,0.12178560495376586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,32,balanced,0.20457599560419717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,1,power_law_1.01,0.1283519983291626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,1,power_law_1.01,0.15133440494537354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,1,power_law_1.01,0.17785600423812867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,1,balanced,0.9648586908976237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,1,power_law_1.01,0.2232192039489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,1,power_law_1.01,0.2838399887084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,32,balanced,0.30215466022491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,32,balanced,0.3095039923985799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,1,power_law_1.01,0.38005120754241944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,1,power_law_1.01,0.463756799697876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,1,power_law_1.01,0.6308928012847901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,1,power_law_1.01,0.8060928344726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,1,power_law_1.01,1.1675775527954102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,1,power_law_1.01,1.5049599647521972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,32,balanced,0.33165866136550903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,32,balanced,0.3001813292503357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,1,balanced,1.431104024251302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,1,power_law_1.01,1.867692756652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,1,power_law_1.01,2.893497657775879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,1,power_law_1.01,5.672768020629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,32,balanced,0.47898133595784503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,32,balanced,0.46061865488688153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,1,balanced,1.842458724975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,1,power_law_1.2,0.03326080143451691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,32,balanced,0.5893173217773438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,32,balanced,0.49274134635925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,1,power_law_1.2,0.03537279963493347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,1,power_law_1.2,0.03999359905719757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,1,power_law_1.2,0.045407998561859134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,1,balanced,5.654805501302083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,1,power_law_1.2,0.055379199981689456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,1,power_law_1.2,0.07739520072937012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,1,balanced,2.330117384592692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,1,power_law_1.2,0.07982720136642456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,1,power_law_1.2,0.08276479840278625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,1,power_law_1.2,0.08646399974822998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,1,power_law_1.2,0.08728960156440735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,1,power_law_1.2,0.09183359742164612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,1,power_law_1.2,0.09633920192718506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,1,power_law_1.2,0.09872639775276185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,1,power_law_1.2,0.10777599811553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,1,power_law_1.2,0.11528960466384888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,32,balanced,0.7031306425730387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,1,power_law_1.2,0.12154879570007324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,32,balanced,0.5622560183207194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,1,power_law_1.2,0.1295232057571411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,1,power_law_1.2,0.15219839811325073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,1,power_law_1.2,0.1765247941017151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,1,power_law_1.2,0.22100479602813722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,1,power_law_1.2,0.28635520935058595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,1,power_law_1.2,0.3769023895263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,1,power_law_1.2,0.45659518241882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,1,power_law_1.2,0.6238656044006348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,1,power_law_1.2,0.7967232227325439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,1,power_law_1.2,1.1578304290771484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,1,power_law_1.2,1.5019904136657716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,1,balanced,3.590757369995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,1,power_law_1.2,1.8527616500854491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,1,power_law_1.2,2.8780927658081055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,1,power_law_1.2,5.643910217285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,32,balanced,1.0715573628743489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,32,balanced,0.852346658706665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,1,power_law_1.01,0.03274880051612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,1,power_law_1.2,0.033497598767280576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,1,power_law_1.01,0.03594880104064942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,1,power_law_1.01,0.03947519958019256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,1,power_law_1.2,0.03528960049152374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,1,power_law_1.01,0.045132800936698914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,1,power_law_1.2,0.03898240029811859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,1,power_law_1.01,0.05464320182800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,1,power_law_1.2,0.044659200310707095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,1,power_law_1.01,0.06659200191497802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,1,power_law_1.2,0.05377920269966126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,1,power_law_1.2,0.06674559712409973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,1,power_law_1.01,0.0700543999671936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,1,power_law_1.2,0.07106559872627258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,1,power_law_1.01,0.0714303970336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,1,power_law_1.2,0.07393919825553893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,1,power_law_1.01,0.07511680126190186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,1,power_law_1.2,0.076883202791214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,1,power_law_1.01,0.07780479788780212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,1,power_law_1.2,0.07932800054550171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,1,power_law_1.01,0.08456959724426269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,1,power_law_1.01,0.09283840060234069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,1,power_law_1.2,0.08614400029182434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,1,power_law_1.01,0.09745919704437256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,1,power_law_1.2,0.09363840222358703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,1,power_law_1.01,0.09806720018386841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,1,power_law_1.2,0.09848319888114929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,1,power_law_1.01,0.10898560285568237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,1,power_law_1.2,0.10005120038986207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,1,power_law_1.01,0.11525119543075561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,1,power_law_1.2,0.10960639715194702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,1,power_law_1.01,0.12801920175552367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,1,power_law_1.2,0.11685760021209717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,1,power_law_1.2,0.13011200428009034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,1,power_law_1.01,0.16021759510040284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,1,power_law_1.2,0.1610303997993469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,1,power_law_1.01,0.1873471975326538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,1,power_law_1.2,0.18910720348358154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,1,power_law_1.01,0.246124792098999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,1,power_law_1.2,0.24571518898010253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,1,power_law_1.01,0.2943871974945068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,1,power_law_1.2,0.2969343900680542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,1,power_law_1.01,0.41025919914245607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,1,power_law_1.2,0.4097407817840576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,1,power_law_1.01,0.5191743850708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,1,power_law_1.2,0.5198527812957764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,1,power_law_1.01,0.7524928092956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,1,power_law_1.2,0.751308822631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,1,power_law_1.01,0.9703231811523437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,1,power_law_1.2,0.9682239532470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,1,power_law_1.2,1.409280014038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,1,power_law_1.01,1.4166015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,1,power_law_1.2,1.8822463989257812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,1,power_law_1.2,2.320524787902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,1,power_law_1.01,1.8824127197265625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,1,power_law_1.2,3.591206359863281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,1,balanced,7.101845423380534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,1,power_law_1.01,2.31530876159668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,1,power_law_1.2,7.086924743652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,1,power_law_1.01,3.584940719604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,32,balanced,2.0530187288920083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,1,power_law_1.01,7.099910736083984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,32,balanced,1.6005333264668782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,32,balanced,0.04645866652329763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,32,balanced,0.047050664822260536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,32,balanced,0.04308266441027323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,32,balanced,0.041333332657814026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,32,balanced,0.052069331208864846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,32,balanced,0.043466667334238686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,32,balanced,0.04761599997679392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,32,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,32,balanced,0.05314666529496511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,32,balanced,0.05026133358478546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,32,balanced,0.051776001850763954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,32,balanced,0.061946665247281395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,32,balanced,0.05398400127887726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,32,power_law_1.01,0.11857919692993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,32,balanced,0.06205866734186808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,32,balanced,0.07020266850789388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,32,balanced,0.0832533339659373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,32,balanced,0.07538133362929027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,32,power_law_1.01,0.11364480257034301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,32,balanced,0.07834133505821228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,32,balanced,0.11218133568763733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,32,balanced,0.07834666470686595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,32,power_law_1.01,0.0758463978767395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,32,balanced,0.11122133334477742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,32,balanced,0.07840533554553986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,32,power_law_1.01,0.09215999841690063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,32,balanced,0.0783679982026418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,32,balanced,0.11284266908963521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,32,power_law_1.01,0.10325119495391846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,32,balanced,0.08053333560625713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,32,balanced,0.13980266451835632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,32,power_law_1.01,0.10621440410614014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,32,balanced,0.07886399825414021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,32,balanced,0.14181333780288696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,32,power_law_1.01,0.11300480365753174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,32,balanced,0.08055999875068665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,32,balanced,0.17045332988103232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,32,power_law_1.01,0.11795200109481811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,32,balanced,0.0827893316745758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,32,power_law_1.01,0.11619199514389038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,32,balanced,0.2510133385658264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,32,balanced,0.08655466636021932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,32,power_law_1.01,0.11994240283966065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,32,balanced,0.31359465916951496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,32,balanced,0.0865760048230489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,32,power_law_1.01,0.12055039405822754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,32,balanced,0.09032533566157024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,32,power_law_1.01,0.1296064019203186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,32,balanced,0.4516106843948364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,32,power_law_1.01,0.13066879510879517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,32,balanced,0.09641066193580627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,32,power_law_1.01,0.12214399576187134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,32,balanced,0.2145493427912394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,32,balanced,0.10058133800824483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,32,power_law_1.01,0.1300160050392151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,32,power_law_1.01,0.13390079736709595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,32,power_law_1.01,0.14165760278701783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,32,balanced,0.2938239971796672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,32,balanced,0.11745599905649821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,32,power_law_1.01,0.1587839961051941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,32,power_law_1.01,0.1920896053314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,32,balanced,0.12359999616940816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,32,power_law_1.01,0.24924159049987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,32,balanced,0.3803466558456421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,32,power_law_1.01,0.24710400104522706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,32,power_law_1.01,0.3451904058456421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,32,balanced,0.14665599664052328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,32,power_law_1.01,0.3601599931716919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,32,power_law_1.01,0.45235838890075686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,32,power_law_1.01,0.5732607841491699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,32,balanced,0.5506399869918823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,32,power_law_1.01,0.8023872375488281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,32,balanced,0.16780799627304077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,32,power_law_1.01,1.0180543899536132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,32,power_law_1.01,1.2820096015930176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,32,power_law_1.01,2.4388671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,32,balanced,0.7195893128712972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,32,balanced,0.23024000724156699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,32,power_law_1.01,3.736012649536133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,32,balanced,0.2596106727917989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,32,balanced,1.0599520206451416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,32,balanced,0.3691200017929077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,32,balanced,1.4035785992940266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,32,power_law_1.01,0.044038400053977966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,32,power_law_1.01,0.07745919823646545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,1,32,power_law_1.2,0.11792000532150268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,1,32,power_law_1.2,0.07845759987831116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,32,power_law_1.01,0.04454399943351746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,32,power_law_1.01,0.07620480060577392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,32,power_law_1.01,0.11249920129776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,32,power_law_1.01,0.053337597846984865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,32,power_law_1.01,0.032601600885391234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,1,32,power_law_1.2,0.07592960000038147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,32,power_law_1.01,0.06223999857902527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,1,32,power_law_1.2,0.06704639792442321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,32,power_law_1.01,0.10046080350875855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,32,power_law_1.01,0.03685120046138764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,32,power_law_1.01,0.06359040141105651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,1,32,power_law_1.2,0.07831680178642272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,32,power_law_1.01,0.04167680144309997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,32,power_law_1.01,0.07754240036010743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,32,power_law_1.01,0.07809280157089234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,32,balanced,0.4394986629486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,1,32,power_law_1.2,0.058169597387313844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,1,32,power_law_1.2,0.08701440095901489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,32,power_law_1.01,0.04517120122909546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,32,power_law_1.01,0.08112000226974488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,32,power_law_1.01,0.08595839738845826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,1,32,power_law_1.2,0.07026559710502625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,32,power_law_1.01,0.045337599515914914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,32,power_law_1.01,0.08280320167541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,1,32,power_law_1.2,0.09400320053100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,32,power_law_1.01,0.09613440036773682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,32,power_law_1.01,0.08188160061836243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,32,power_law_1.01,0.05279359817504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,1,32,power_law_1.2,0.11075839996337891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,1,32,power_law_1.2,0.07448319792747497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,32,power_law_1.01,0.08439679741859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,32,power_law_1.01,0.10636160373687745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,32,power_law_1.01,0.052985602617263795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,1,32,power_law_1.2,0.11084799766540528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,32,power_law_1.01,0.08586239814758301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,32,power_law_1.01,0.05251839756965637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,32,power_law_1.01,0.08755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,1,32,power_law_1.2,0.08075519800186157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,32,power_law_1.01,0.11075199842453003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,1,32,power_law_1.2,0.10767999887466431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,32,power_law_1.01,0.0908735990524292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,32,power_law_1.01,0.08406400084495544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,1,32,power_law_1.2,0.08417919874191285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,1,32,power_law_1.2,0.11546239852905274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,32,power_law_1.01,0.10758399963378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,32,balanced,1.7356106440226238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,32,power_law_1.01,0.09433599710464477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,1,32,power_law_1.2,0.0832319974899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,32,power_law_1.01,0.10276479721069336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,32,power_law_1.01,0.08273919820785522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,1,32,power_law_1.2,0.1238144040107727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,32,power_law_1.01,0.10573439598083496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,32,power_law_1.01,0.10899840593338013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,1,32,power_law_1.2,0.0833728015422821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,1,32,power_law_1.2,0.12547199726104735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,32,power_law_1.01,0.08442879915237426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,32,power_law_1.01,0.11226240396499634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,32,power_law_1.01,0.10679680109024048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,32,power_law_1.01,0.12442879676818848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,1,32,power_law_1.2,0.08516479730606079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,1,32,power_law_1.2,0.12987519502639772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,32,power_law_1.01,0.08479359745979309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,32,power_law_1.01,0.14130560159683228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,1,32,power_law_1.2,0.13102079629898072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,32,power_law_1.01,0.10430719852447509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,32,power_law_1.01,0.09880959987640381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,32,power_law_1.01,0.1860159993171692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,1,32,power_law_1.2,0.08936960101127625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,1,32,power_law_1.2,0.13050880432128906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,32,power_law_1.01,0.10487680435180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,32,power_law_1.01,0.19396480321884155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,32,power_law_1.01,0.11063040494918823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,1,32,power_law_1.2,0.14083199501037597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,32,power_law_1.01,0.24913280010223388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,32,power_law_1.01,0.10711679458618165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,32,balanced,0.5213119983673096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,32,power_law_1.01,0.10366719961166382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,1,32,power_law_1.2,0.09189760088920593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,32,power_law_1.01,0.2969599962234497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,1,32,power_law_1.2,0.13933440446853637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,32,power_law_1.01,0.1466752052307129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,32,power_law_1.01,0.10954240560531617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,32,power_law_1.01,0.404307222366333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,1,32,power_law_1.2,0.15319679975509642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.01,0.19658880233764647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,1,32,power_law_1.2,0.09322879910469055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,32,power_law_1.01,0.49609599113464353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,32,power_law_1.01,0.10759040117263793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,1,32,power_law_1.2,0.17463040351867676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.01,0.24734079837799072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,32,power_law_1.01,0.7479040145874023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,1,32,power_law_1.2,0.0945855975151062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,1,32,power_law_1.2,0.2129728078842163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.01,0.1276800036430359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,32,power_law_1.01,0.11390719413757325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,32,power_law_1.01,0.9078911781311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,1,32,power_law_1.2,0.2945280075073242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.01,0.16710400581359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,1,32,power_law_1.2,0.10813440084457397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,32,power_law_1.01,0.12685439586639405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,32,power_law_1.01,1.1503232002258301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.01,0.20174720287322997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,1,32,power_law_1.2,0.2978111982345581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,32,power_law_1.01,0.1291391968727112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,1,32,power_law_1.2,0.1120959997177124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,32,power_law_1.01,1.740563201904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.01,0.3183552026748657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,1,32,power_law_1.2,0.3641088008880615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.01,0.1468224048614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,1,32,power_law_1.2,0.11673599481582642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.01,0.4026495933532715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,1,32,power_law_1.2,0.4122176170349121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,32,power_law_1.01,3.3145729064941407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.01,0.16948479413986206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,1,32,power_law_1.2,0.5142335891723633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,1,32,power_law_1.2,0.13362560272216797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.01,0.633241605758667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.01,0.19356160163879393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,1,32,power_law_1.2,0.8837120056152343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.01,0.24788479804992675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,1,32,power_law_1.2,0.15216000080108644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.01,0.7497536182403565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.01,0.317740797996521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,1,32,power_law_1.2,1.197555160522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,1,32,power_law_1.2,0.19363839626312257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,32,balanced,2.772864023844401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.01,0.41479039192199707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,32,power_law_1.01,1.133241558074951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.01,0.5446464061737061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,1,32,power_law_1.2,0.214355206489563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,1,32,power_law_1.2,1.8547136306762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,32,power_law_1.01,1.6770368576049806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.01,0.7750207901000976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,1,32,power_law_1.2,0.2899775981903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,1,32,power_law_1.2,2.022271919250488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.01,1.0362496376037598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,1,32,power_law_1.2,0.33844480514526365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,32,power_law_1.01,3.0692352294921874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,1,32,power_law_1.2,0.4535359859466553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,32,power_law_1.01,1.1376192092895507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,32,balanced,0.7861706415812174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,1,32,power_law_1.2,3.2653057098388674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,1,32,power_law_1.2,0.6804992198944092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,32,power_law_1.01,1.9505535125732423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,1,32,power_law_1.2,1.017612838745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,1,32,power_law_1.2,7.269612884521484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,32,power_law_1.01,3.765798568725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,1,32,power_law_1.2,1.121126365661621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,1,32,power_law_1.2,1.7423744201660156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,1,32,power_law_1.2,2.431679916381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,1,32,power_law_1.2,5.057132720947266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,32,balanced,5.488949457804362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,32,balanced,1.4991733233133953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,1,32,power_law_1.2,0.10854400396347046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,1,32,power_law_1.2,0.08067200183868409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,1,32,power_law_1.2,0.07434239983558655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,1,32,power_law_1.2,0.07992960214614868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,1,32,power_law_1.2,0.08916479945182801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,1,32,power_law_1.2,0.0987775981426239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,1,32,power_law_1.2,0.1016319990158081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,1,32,power_law_1.2,0.10298240184783936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,1,32,power_law_1.2,0.10202239751815796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,1,32,power_law_1.2,0.11038080453872681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,1,32,power_law_1.2,0.10766719579696656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,1,32,power_law_1.2,0.1048192024230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,1,32,power_law_1.2,0.10760960578918458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,1,32,power_law_1.2,0.10699520111083985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,1,32,power_law_1.2,0.11347839832305909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,1,32,power_law_1.2,0.11783039569854736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,1,32,power_law_1.2,0.12636159658432006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,1,32,power_law_1.2,0.1487231969833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.2,0.15011199712753295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.2,0.18982399702072145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.2,0.21795198917388917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.2,0.28268799781799314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.2,0.35155839920043946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.2,0.5483327865600586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.2,0.6038591861724854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.2,0.8570879936218262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.2,1.4121600151062013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,1,32,power_law_1.2,1.466470432281494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,1,32,power_law_1.2,2.411731147766113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,1,32,power_law_1.2,5.248870468139648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,1,32,power_law_1.2,0.045184001326560974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,1,32,power_law_1.2,0.03598720133304596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,1,32,power_law_1.2,0.027910399436950683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,1,32,power_law_1.2,0.03500800132751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,1,32,power_law_1.2,0.03916800022125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,1,32,power_law_1.2,0.045542401075363156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,1,32,power_law_1.2,0.04508799910545349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,1,32,power_law_1.2,0.05296639800071716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,1,32,power_law_1.2,0.05270400047302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,1,32,power_law_1.2,0.054176002740859985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,1,32,power_law_1.2,0.08339200019836426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,1,32,power_law_1.2,0.08377599716186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,1,32,power_law_1.2,0.08409600257873535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,1,32,power_law_1.2,0.08533120155334473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,1,32,power_law_1.2,0.09488000273704529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,1,32,power_law_1.2,0.11196160316467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,1,32,power_law_1.2,0.12312320470809937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,1,32,power_law_1.2,0.16312320232391359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.2,0.19093120098114014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.2,0.25822720527648924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.2,0.14229120016098024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.2,0.1911936044692993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.2,0.2376255989074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.2,0.37034881114959717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.2,0.45413761138916015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.2,0.6591872215270996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.2,0.9641216278076172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,1,32,power_law_1.2,1.2620863914489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,1,32,power_law_1.2,1.9248895645141602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,1,32,power_law_1.2,3.781497573852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,16,balanced,0.03225066761175791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,16,balanced,0.02515200028816859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,16,balanced,0.027562665442625683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,16,balanced,0.02712533374627431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,16,balanced,0.03941333293914795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,16,balanced,0.04514666895071665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,16,balanced,0.047184000412623085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,16,balanced,0.05109333495299021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,16,balanced,0.05413866539796194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,16,balanced,0.05446400245030721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,16,balanced,0.05435200035572052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,16,balanced,0.06368533273537953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,16,balanced,0.049653331438700356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,16,balanced,0.0643093337615331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,16,balanced,0.08641067147254944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,16,balanced,0.05175999800364176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,16,balanced,0.11173333724339803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,16,balanced,0.04993600149949392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,16,balanced,0.05384533107280731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,16,balanced,0.11363200346628825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,16,balanced,0.06630933284759521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,16,balanced,0.13962666193644205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,16,balanced,0.07446933289368947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,16,balanced,0.0774186650911967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,16,balanced,0.17441066106160483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,16,balanced,0.07778133451938629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,16,balanced,0.07851199805736542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,16,balanced,0.2071466644605001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,16,balanced,0.07832000156243642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,16,balanced,0.08030933141708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,16,balanced,0.07811200122038524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,16,balanced,0.2709653377532959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,16,balanced,0.08146666487058003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,16,balanced,0.08418132861455281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,16,balanced,0.16609066724777222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,16,balanced,0.08712533116340637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,16,balanced,0.08873599767684937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,16,balanced,0.09057600299517314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,16,balanced,0.21659733851750693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,16,balanced,0.09893866380055745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,16,balanced,0.105103999376297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,16,balanced,0.2586666742960612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,16,balanced,0.12444800138473511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,16,balanced,0.13024000326792398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,16,balanced,0.3753173351287842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,16,balanced,0.15585600336392721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,16,balanced,0.17903467019399008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,16,balanced,0.4676906665166219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,16,balanced,0.2526773413022359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,16,power_law_1.01,0.09542400240898133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,16,power_law_1.01,0.10112639665603637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,16,balanced,0.6896320184071859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,16,balanced,0.293013334274292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,16,power_law_1.01,0.06346880197525025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,16,power_law_1.01,0.0809984028339386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,16,power_law_1.01,0.08419200181961059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,16,power_law_1.01,0.09864320158958435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,16,power_law_1.01,0.10015360116958619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,16,power_law_1.01,0.10414079427719117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,16,power_law_1.01,0.10398080348968505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,16,power_law_1.01,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,16,power_law_1.01,0.10029439926147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,16,power_law_1.01,0.10343040227890014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,16,balanced,0.40053868293762207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,16,power_law_1.01,0.10291199684143067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,16,balanced,0.9037653605143229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,16,power_law_1.01,0.10698239803314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,16,power_law_1.01,0.11059199571609497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,16,power_law_1.01,0.11276160478591919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,16,power_law_1.01,0.11349120140075683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,16,power_law_1.01,0.12978559732437134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.01,0.14058239459991456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.01,0.16951680183410645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.01,0.19648000001907348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.01,0.24698879718780517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.01,0.28019840717315675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.01,0.39749760627746583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,16,balanced,0.4995306730270386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.01,0.4946879863739014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.01,0.7290112018585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.01,0.8949055671691895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,16,power_law_1.01,1.0581055641174317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,16,balanced,1.0962186654408772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,16,power_law_1.01,1.8564416885375976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,16,power_law_1.01,3.1243007659912108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,16,balanced,0.5972373485565186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,16,balanced,1.7569173177083333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,16,power_law_1.01,0.045516800880432126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,16,balanced,0.9102773666381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,16,power_law_1.01,0.03544319868087768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,16,power_law_1.01,0.026745599508285523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,16,power_law_1.01,0.027475199103355406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,16,power_law_1.01,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,16,power_law_1.01,0.040652799606323245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,16,power_law_1.01,0.040729600191116336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,16,power_law_1.01,0.045184001326560974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,16,power_law_1.01,0.044019201397895814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,16,power_law_1.01,0.04575360119342804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,16,power_law_1.01,0.05623040199279785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,16,power_law_1.01,0.0561024010181427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,16,power_law_1.01,0.05742080211639404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,16,power_law_1.01,0.08533759713172913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,16,power_law_1.01,0.08541439771652222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,16,power_law_1.01,0.0910975992679596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,16,power_law_1.01,0.09299200177192687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,16,power_law_1.01,0.11723519563674926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.01,0.13068159818649291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.01,0.17958400249481202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.01,0.1128383994102478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.01,0.14454400539398193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.01,0.1696447968482971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.01,0.23906559944152833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.01,0.29679360389709475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.01,0.45148801803588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.01,0.5878911972045898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,16,power_law_1.01,0.7751039981842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,16,power_law_1.01,1.0522047996520996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,16,power_law_1.01,2.1764671325683596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,16,balanced,3.450607935587565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,16,balanced,1.7471893628438313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,16,balanced,0.0473333348830541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,16,balanced,0.04910400013128916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,16,balanced,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,16,balanced,0.05374933282534281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,16,balanced,0.07654933134714763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,16,balanced,0.08455999692281087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,16,balanced,0.08681600292523702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,16,balanced,0.0855466624101003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,16,balanced,0.08732266227404277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,16,balanced,0.08763733506202698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,16,balanced,0.08893866340319316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,16,balanced,0.09088533123334248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,16,balanced,0.0913866659005483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,16,balanced,0.09238400061925252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,16,balanced,0.09799466530481975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,16,balanced,0.04580800235271454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,16,balanced,0.10094400246938069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,16,balanced,0.04772266745567322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,16,balanced,0.10264000296592712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,16,balanced,0.050474668542544045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,16,balanced,0.11590400338172913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,16,balanced,0.07635200023651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,16,balanced,0.10316266616185506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,16,balanced,0.11947733163833618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,16,balanced,0.1141813298066457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,16,balanced,0.11698133746782939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,16,balanced,0.14366933703422546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,16,balanced,0.1170240044593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,16,balanced,0.11946133772532146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,16,balanced,0.11512533823649089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,16,balanced,0.15689067045847574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,16,balanced,0.11969066659609477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,16,balanced,0.11806933085123698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,16,balanced,0.11817600329717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,16,balanced,0.11938666303952535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,16,balanced,0.20028799772262573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,16,balanced,0.12574932972590128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,16,balanced,0.12424533565839131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,16,balanced,0.12894933422406515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,16,balanced,0.22243734200795492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,16,balanced,0.14079999923706055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,16,balanced,0.13615467151006064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,16,balanced,0.16583466529846191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,16,balanced,0.3290239969889323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,16,balanced,0.1574986676375071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,16,power_law_1.01,0.07799680233001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,16,balanced,0.22220800320307413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,16,power_law_1.01,0.0796992003917694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,16,power_law_1.01,0.05742080211639404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,16,balanced,0.37121065457661945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,16,power_law_1.01,0.057920002937316896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,16,power_law_1.2,0.10264960527420045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,16,power_law_1.01,0.06477439999580384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,16,balanced,0.20965866247812906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,16,power_law_1.01,0.07839999794960022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,16,power_law_1.2,0.055129599571228025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,16,power_law_1.01,0.08262400031089782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,16,power_law_1.2,0.05555199980735779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,16,power_law_1.01,0.0836736023426056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,16,power_law_1.2,0.07518079876899719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,16,power_law_1.01,0.1081663966178894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,16,power_law_1.01,0.08349440097808838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,16,power_law_1.2,0.09528959989547729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,16,power_law_1.01,0.08339840173721313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,16,power_law_1.2,0.10399999618530273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,16,power_law_1.01,0.11265920400619507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,16,power_law_1.01,0.08748160004615783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,16,balanced,0.362277348836263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,16,power_law_1.2,0.10771199464797973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,16,power_law_1.01,0.08983680009841918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,16,power_law_1.01,0.06492800116539002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,16,power_law_1.2,0.1108672022819519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,16,power_law_1.01,0.09167360067367554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,16,power_law_1.01,0.08067839741706848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,16,power_law_1.2,0.11406079530715943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,16,balanced,0.5521546602249146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,16,power_law_1.01,0.08984959721565247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,16,power_law_1.01,0.08605440258979798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,16,power_law_1.2,0.11433600187301636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,16,power_law_1.01,0.10298880338668823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,16,power_law_1.01,0.09869440197944641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,16,power_law_1.2,0.11246720552444459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,16,power_law_1.01,0.10679680109024048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,16,power_law_1.01,0.10357120037078857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,16,power_law_1.2,0.1151360034942627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,16,power_law_1.01,0.11099519729614257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,16,power_law_1.01,0.10996479988098144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,16,power_law_1.2,0.1150272011756897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,16,balanced,0.31563733021418255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,16,power_law_1.01,0.13201279640197755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,16,power_law_1.01,0.10909440517425537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,16,power_law_1.2,0.12537599802017213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,16,power_law_1.01,0.13663359880447387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,16,power_law_1.01,0.10938880443572999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,16,power_law_1.2,0.13081599473953248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,16,power_law_1.01,0.17132799625396727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,16,power_law_1.2,0.1373695969581604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,16,power_law_1.01,0.11262719631195069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,16,power_law_1.2,0.1487615942955017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,16,power_law_1.01,0.20103681087493896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,16,power_law_1.01,0.11561599969863892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,16,power_law_1.2,0.1766592025756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,16,power_law_1.01,0.2684416055679321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,16,power_law_1.01,0.11872639656066894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,16,power_law_1.2,0.17878400087356566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,16,power_law_1.01,0.11770880222320557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,16,power_law_1.01,0.31674239635467527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,16,power_law_1.2,0.2346816062927246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,16,power_law_1.01,0.1296895980834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,16,power_law_1.01,0.41564159393310546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,16,power_law_1.2,0.2829375982284546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,16,power_law_1.01,0.12887040376663209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,16,power_law_1.01,0.5309696197509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,16,power_law_1.2,0.32778239250183105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,16,balanced,0.656549334526062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,16,power_law_1.01,0.14017280340194702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,16,power_law_1.2,0.4170368194580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,16,power_law_1.01,0.7488255977630616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,16,balanced,0.4803893168767293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,16,power_law_1.01,0.16858880519866942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,16,power_law_1.2,0.5695615768432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,16,power_law_1.01,0.18071039915084838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,16,power_law_1.01,0.9513855934143066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,16,power_law_1.2,0.6060031890869141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,16,power_law_1.01,0.21896960735321044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,16,power_law_1.01,1.1083840370178222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,16,power_law_1.2,0.8270144462585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,16,power_law_1.01,0.21998720169067382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,16,power_law_1.2,1.3409343719482423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,16,power_law_1.01,1.6648895263671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,16,power_law_1.01,0.3108992099761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,16,power_law_1.2,1.435155200958252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,16,power_law_1.01,0.3495296001434326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,16,power_law_1.01,3.3098175048828127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,16,power_law_1.01,0.46528639793396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,16,power_law_1.2,2.059404754638672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,16,power_law_1.01,0.5276735782623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,16,power_law_1.2,4.314822387695313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,16,power_law_1.01,0.6501503944396972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,16,power_law_1.01,0.9079423904418945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,16,power_law_1.01,1.1660415649414062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,16,balanced,0.5249013503392538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,16,power_law_1.01,1.5307519912719727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,16,balanced,0.7740639845530192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,16,power_law_1.01,2.9191808700561523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,16,power_law_1.2,0.07197440266609192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,16,power_law_1.2,0.04932479858398438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,16,power_law_1.2,0.0496832013130188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,16,power_law_1.2,0.058323198556900026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,16,balanced,0.6057759920756022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,16,power_law_1.2,0.06249600052833557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,16,power_law_1.2,0.07861760258674622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,16,power_law_1.2,0.08101760149002075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,16,power_law_1.2,0.08160640001296997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,16,power_law_1.2,0.08069760203361512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,16,power_law_1.2,0.08330879807472229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,16,power_law_1.2,0.08763520121574402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,16,power_law_1.2,0.08999680280685425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,16,power_law_1.2,0.09098240137100219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,16,power_law_1.2,0.09765759706497193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,16,power_law_1.2,0.09722239971160888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,16,balanced,1.200698693593343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,16,power_law_1.2,0.06152960062026978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,16,power_law_1.2,0.10392320156097412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,16,power_law_1.2,0.06602240204811097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,16,power_law_1.2,0.11626240015029907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,16,power_law_1.2,0.07641599774360656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,16,power_law_1.2,0.11323519945144653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,16,power_law_1.2,0.08044800162315369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,16,power_law_1.2,0.1279039978981018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,16,power_law_1.2,0.09575679898262024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,16,power_law_1.2,0.14872959852218628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,16,power_law_1.2,0.10381439924240113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,16,power_law_1.2,0.18202240467071534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,16,power_law_1.2,0.10266879796981812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,16,power_law_1.2,0.20151040554046631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,16,power_law_1.2,0.1022271990776062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,16,power_law_1.2,0.2687743902206421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,16,power_law_1.2,0.10154240131378174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,16,power_law_1.2,0.3478208065032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,16,power_law_1.2,0.04500479996204376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,16,power_law_1.2,0.10492160320281982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,16,power_law_1.2,0.48749442100524903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,16,power_law_1.2,0.10720000267028809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,16,power_law_1.2,0.023270399868488313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,16,power_law_1.2,0.6013760089874267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,16,balanced,0.9230453173319498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,16,power_law_1.2,0.10623359680175781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,16,power_law_1.2,0.026681599020957947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,16,power_law_1.2,0.10301439762115479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,16,power_law_1.2,0.8648832321166993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,16,power_law_1.2,0.025670400261878966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,16,power_law_1.2,1.220787239074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,16,power_law_1.2,0.11422719955444335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,16,power_law_1.2,0.03342719972133636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,16,power_law_1.2,1.4714303970336915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,16,power_law_1.2,0.11164159774780273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,16,power_law_1.2,0.038624000549316403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,16,power_law_1.2,2.3336063385009767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,16,power_law_1.2,0.1223423957824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,16,power_law_1.2,0.04115839898586273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,16,power_law_1.2,0.1325376033782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,16,power_law_1.2,0.04426240026950836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,16,power_law_1.2,5.083929443359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.2,0.14554879665374756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,16,power_law_1.2,0.044870400428771974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.2,0.17614079713821412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,16,power_law_1.2,0.044409599900245664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.2,0.20096640586853026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,16,power_law_1.2,0.054764801263809205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.2,0.2642751932144165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,16,power_law_1.2,0.05605120062828064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.2,0.313811206817627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,16,power_law_1.2,0.056543999910354616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,16,power_law_1.2,0.08624640107154846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.2,0.4131775856018066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,16,power_law_1.2,0.08499199748039246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.2,0.5375679969787598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,16,power_law_1.2,0.09859840273857116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.2,0.8609087944030762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,16,power_law_1.2,0.10132479667663574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,16,power_law_1.2,0.12250239849090576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.2,1.0537983894348144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.2,0.13704320192337036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,16,power_law_1.2,1.2728192329406738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.2,0.19521280527114868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,16,power_law_1.2,2.5151744842529298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.2,0.1126207947731018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.2,0.1429695963859558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,16,power_law_1.2,5.599110412597656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.2,0.19347840547561646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.2,0.2658112049102783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.2,0.3315840005874634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.2,0.502726411819458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.2,0.6558400154113769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,16,balanced,2.3128320376078286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,16,power_law_1.2,0.8108223915100098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,16,power_law_1.2,1.376576042175293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,16,power_law_1.2,2.6187456130981444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,16,balanced,1.712959925333659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,8,balanced,0.050111999114354454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,8,balanced,0.023221333821614582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,8,balanced,0.05198400219281515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,8,balanced,0.05399466554323832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,8,balanced,0.05590933561325073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,8,balanced,0.02513066679239273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,8,balanced,0.07188799977302551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,8,balanced,0.025120000044504803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,8,balanced,0.08241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,8,balanced,0.02712533374627431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,8,balanced,0.08248533308506012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,8,balanced,0.03977599988381068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,8,balanced,0.0844533344109853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,8,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,8,balanced,0.08477866649627686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,8,balanced,0.052890668312708534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,8,balanced,0.08478400111198425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,8,balanced,0.051818668842315674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,8,balanced,0.08489066362380981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,8,balanced,0.05388266841570536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,8,balanced,0.08668266733487447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,8,balanced,0.06614399949709575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,8,balanced,0.08821333448092143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,8,balanced,0.06328533093134563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,8,balanced,0.09001599748929341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,8,balanced,0.06849599877993266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,8,balanced,0.09487467010815938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,8,balanced,0.08995200196901958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,8,balanced,0.09750933448473613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,8,balanced,0.08966933687527974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,8,balanced,0.10057066877683003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,8,balanced,0.09133866429328918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,8,balanced,0.1029813289642334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,8,balanced,0.11083199580510457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,8,balanced,0.09910399715105693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,8,balanced,0.1200213332970937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,8,balanced,0.10981333255767822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,8,balanced,0.15029866496721903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,8,balanced,0.12403733531634013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,8,balanced,0.1325866679350535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,8,balanced,0.15413332978884378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,8,balanced,0.15424000223477682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,8,balanced,0.19510400295257568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,8,balanced,0.18451199928919473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,8,balanced,0.22688533862431845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,8,balanced,0.2107413411140442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,8,power_law_1.01,0.05459840297698974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,8,balanced,0.3240373333295186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,8,power_law_1.01,0.022976000607013703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,8,power_law_1.01,0.05454720258712768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,8,power_law_1.01,0.04797439873218536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,8,balanced,0.31566399335861206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,8,power_law_1.01,0.02444159984588623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,8,power_law_1.01,0.05685120224952698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,8,power_law_1.01,0.023603199422359465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,8,power_law_1.01,0.05945600271224975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,8,power_law_1.01,0.026137599349021913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,8,power_law_1.01,0.07490559816360473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,8,power_law_1.01,0.028166401386260986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,8,power_law_1.01,0.07391999959945679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,8,power_law_1.01,0.040320000052452086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,8,balanced,0.3803733189900716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,8,power_law_1.01,0.07834240198135375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,8,power_law_1.01,0.046854400634765626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,8,power_law_1.01,0.0797439992427826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,8,power_law_1.01,0.04521600008010864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,8,balanced,0.3866080045700073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,8,power_law_1.01,0.07986559867858886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,8,power_law_1.01,0.041433599591255185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,8,power_law_1.01,0.08101760149002075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,8,power_law_1.01,0.05573760271072388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,8,power_law_1.01,0.08439679741859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,8,power_law_1.01,0.05446400046348572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,8,power_law_1.01,0.08348159790039063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,8,power_law_1.01,0.057036799192428586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,8,power_law_1.01,0.08833280205726624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,8,power_law_1.01,0.0855679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,8,power_law_1.01,0.09793279767036438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,8,power_law_1.01,0.08565760254859925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,8,power_law_1.01,0.0998848021030426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,8,power_law_1.01,0.08529279828071594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,8,balanced,0.5226879914601644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,8,power_law_1.01,0.10572799444198608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,8,power_law_1.01,0.09279999732971192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,8,power_law_1.01,0.09730560183525086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,8,power_law_1.01,0.12253439426422119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,8,balanced,0.5481119950612386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,8,power_law_1.01,0.08810880184173583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.01,0.13432960510253905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.01,0.09125120043754578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.01,0.17093119621276856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.01,0.09534720182418824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.01,0.1777151942253113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.01,0.11203839778900146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.01,0.2363136053085327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.01,0.1274623990058899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.01,0.30110719203948977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.01,0.16202880144119264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.01,0.40032639503479006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,8,balanced,0.6736266613006592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.01,0.2222208023071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.01,0.4883135795593262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.01,0.2829440116882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.01,0.7134655952453614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,8,balanced,0.7222560246785482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.01,0.39112319946289065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.01,0.8852352142333985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.01,0.4684864044189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,8,power_law_1.01,1.076959991455078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,8,power_law_1.01,0.5793216228485107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,8,power_law_1.01,1.7308544158935546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,8,power_law_1.01,0.9830207824707031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,8,power_law_1.01,3.2707073211669924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,8,power_law_1.01,1.9351999282836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,8,balanced,0.8031840324401855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,8,balanced,0.8584213256835938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,8,balanced,1.241871992746989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,8,balanced,1.377994696299235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,8,balanced,2.4085493087768555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,8,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,8,balanced,2.6882187525431314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,8,balanced,0.04460800190766653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,8,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,8,balanced,0.053786665201187134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,8,balanced,0.0793333351612091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,8,balanced,0.08455466230710347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,8,balanced,0.08689066767692566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,8,balanced,0.08925333619117737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,8,balanced,0.08716799815495808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,8,balanced,0.04770666857560476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,8,balanced,0.08962133526802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,8,balanced,0.09072533249855042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,8,balanced,0.04974933465321859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,8,balanced,0.09104532996813457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,8,balanced,0.05382933219273885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,8,balanced,0.09267733494440715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,8,balanced,0.07337599992752075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,8,balanced,0.09706133604049683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,8,balanced,0.10546666383743286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,8,balanced,0.10186666250228882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,8,balanced,0.11563733220100403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,8,balanced,0.10538132985432942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,8,balanced,0.11653866370519002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,8,balanced,0.11821867028872173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,8,balanced,0.11006399989128113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,8,balanced,0.11990933616956075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,8,balanced,0.11900267004966736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,8,balanced,0.11821867028872173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,8,balanced,0.11993599931399028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,8,balanced,0.13062933087348938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,8,balanced,0.12035199999809265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,8,balanced,0.12144533793131511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,8,balanced,0.15820266803105673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,8,balanced,0.12212266524632771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,8,balanced,0.12867732842763266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,8,balanced,0.17257599035898843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,8,balanced,0.12753599882125854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,8,balanced,0.1318986713886261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,8,balanced,0.14088533322016397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,8,balanced,0.22995734214782715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,8,balanced,0.1462453305721283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,8,balanced,0.17293334007263184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,8,balanced,0.26418666044871014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,8,balanced,0.17908799648284912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,8,balanced,0.23549334208170572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,8,balanced,0.3791360060373942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,8,balanced,0.24367467562357584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,8,balanced,0.44412799676259357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,8,balanced,0.38170135021209717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,8,power_law_1.2,0.07217280268669128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,8,power_law_1.2,0.049491199851036075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,8,power_law_1.2,0.04815999865531921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,8,power_law_1.2,0.04307200014591217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,8,power_law_1.2,0.05432320237159729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,8,balanced,0.38180267810821533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,8,power_law_1.2,0.045849600434303285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,8,power_law_1.2,0.06705920100212097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,8,balanced,0.620037317276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,8,power_law_1.2,0.05315200090408325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,8,power_law_1.2,0.07394559979438782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,8,power_law_1.2,0.060703998804092406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,8,power_law_1.2,0.0985535979270935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,8,power_law_1.2,0.07893760204315185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,8,power_law_1.2,0.10334080457687378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,8,power_law_1.2,0.08147839903831482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,8,power_law_1.2,0.10959999561309815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,8,power_law_1.2,0.08197759985923767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,8,power_law_1.2,0.11093759536743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,8,power_law_1.2,0.08465279936790467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,8,power_law_1.2,0.11189119815826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,8,power_law_1.2,0.0830784022808075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,8,power_law_1.2,0.11364480257034301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,8,balanced,0.5272053480148315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,8,power_law_1.2,0.08821120262145996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,8,power_law_1.2,0.11374720335006713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,8,power_law_1.2,0.09093760251998902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,8,power_law_1.2,0.09397760033607483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,8,power_law_1.2,0.12083200216293336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,8,power_law_1.2,0.10071040391921997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,8,balanced,0.797386646270752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,8,power_law_1.2,0.12303999662399293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,8,power_law_1.2,0.10650240182876587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,8,power_law_1.2,0.1306048035621643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,8,power_law_1.2,0.1097599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,8,power_law_1.2,0.13265279531478882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,8,power_law_1.2,0.11399680376052856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,8,power_law_1.2,0.1385983943939209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,8,power_law_1.2,0.12789119482040406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,8,power_law_1.2,0.15635199546813966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,8,power_law_1.2,0.14844160079956054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,8,power_law_1.2,0.1913599967956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,8,power_law_1.2,0.18195199966430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,8,power_law_1.2,0.2157439947128296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,8,power_law_1.2,0.21514880657196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,8,power_law_1.2,0.2377023935317993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,8,balanced,0.6647306680679321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,8,power_law_1.2,0.2768127918243408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,8,power_law_1.01,0.05175039768218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,8,power_law_1.2,0.31153919696807864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,8,power_law_1.2,0.3847807884216309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,8,power_law_1.2,0.3979840040206909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,8,power_law_1.2,0.5021440029144287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,8,power_law_1.01,0.0728384017944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,8,power_law_1.2,0.5634304046630859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,8,power_law_1.2,0.6106304168701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,8,power_law_1.01,0.0775488018989563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,8,power_law_1.01,0.05194240212440491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,8,power_law_1.01,0.05611519813537598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,8,power_law_1.2,0.5584447860717774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,8,power_law_1.2,0.819264030456543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,8,power_law_1.01,0.06444799900054932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,8,power_law_1.01,0.0884223997592926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,8,balanced,0.9502507050832113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,8,power_law_1.2,0.8912192344665527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,8,power_law_1.2,1.1196160316467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,8,power_law_1.01,0.0796288013458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,8,power_law_1.01,0.05980160236358643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,8,power_law_1.01,0.0819711983203888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,8,power_law_1.2,1.1696576118469237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,8,power_law_1.2,1.6577856063842773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,8,power_law_1.01,0.07256320118904114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,8,power_law_1.01,0.084307199716568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,8,power_law_1.01,0.08312320113182067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,8,power_law_1.01,0.08361600041389465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,8,power_law_1.2,2.495334434509277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,8,power_law_1.2,1.2939007759094239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,8,power_law_1.01,0.08591359853744507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,8,power_law_1.01,0.1031999945640564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,8,power_law_1.01,0.08917760252952575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,8,power_law_1.01,0.10709120035171509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,8,power_law_1.2,2.2417024612426757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,8,power_law_1.01,0.09117439985275269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,8,power_law_1.2,4.903392028808594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,8,power_law_1.01,0.1086400032043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,8,power_law_1.01,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,8,power_law_1.01,0.11153279542922974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,8,power_law_1.01,0.0959231972694397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,8,power_law_1.2,4.110015869140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,8,power_law_1.01,0.10920959711074829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,8,power_law_1.01,0.10587519407272339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,8,balanced,0.7639146645863851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,8,power_law_1.01,0.10880000591278076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,8,power_law_1.01,0.11279360055923462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,8,power_law_1.01,0.1155135989189148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,8,power_law_1.01,0.11573760509490967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,8,power_law_1.01,0.13032959699630736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,8,power_law_1.01,0.11747839450836181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,8,power_law_1.01,0.1390015959739685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,8,power_law_1.01,0.1235584020614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,8,power_law_1.01,0.17882879972457885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,8,power_law_1.01,0.21107840538024902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,8,power_law_1.01,0.13147519826889037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,8,power_law_1.01,0.26352639198303224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,8,power_law_1.01,0.1349184036254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,8,power_law_1.01,0.3336575984954834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,8,power_law_1.01,0.1364416003227234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,8,power_law_1.01,0.4641024112701416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,8,power_law_1.01,0.15365760326385497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,8,power_law_1.01,0.5420544147491455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,8,power_law_1.01,0.16227200031280517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,8,power_law_1.01,0.7613183975219726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,8,power_law_1.01,0.18487679958343506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,8,power_law_1.01,1.0317567825317382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,8,power_law_1.01,0.22006399631500245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,8,power_law_1.01,1.227667236328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,8,balanced,1.4786186218261719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,8,power_law_1.01,0.27260799407958985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,8,power_law_1.01,1.874835205078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,8,power_law_1.01,0.3641664028167725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,8,power_law_1.01,0.4773888111114502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,8,power_law_1.01,3.993804931640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,8,power_law_1.01,0.5090432167053223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,8,power_law_1.01,0.6553984165191651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,8,power_law_1.01,0.9744959831237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,8,power_law_1.01,1.1388416290283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,8,power_law_1.01,1.7460287094116211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,8,balanced,1.1654187043507893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,8,power_law_1.01,3.4196414947509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,8,power_law_1.2,0.05689600110054016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,8,power_law_1.2,0.023500800132751465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,8,power_law_1.2,0.05090559720993042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,8,power_law_1.2,0.024198399484157564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,8,power_law_1.2,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,8,power_law_1.2,0.025363200902938844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,8,power_law_1.2,0.05676800012588501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,8,power_law_1.2,0.02526719868183136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,8,power_law_1.2,0.029766398668289184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,8,power_law_1.2,0.03793280124664307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,8,power_law_1.2,0.05990399718284607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,8,power_law_1.2,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,8,power_law_1.2,0.0709824025630951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,8,power_law_1.2,0.04287360012531281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,8,power_law_1.2,0.07571840286254883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,8,power_law_1.2,0.04428159892559051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,8,power_law_1.2,0.080595201253891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,8,power_law_1.2,0.05331839919090271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,8,power_law_1.2,0.0809664011001587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,8,power_law_1.2,0.05505920052528381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,8,power_law_1.2,0.08193920254707336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,8,power_law_1.2,0.05448319911956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,8,power_law_1.2,0.08659200072288513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,8,power_law_1.2,0.08300160169601441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,8,power_law_1.2,0.08658559918403626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,8,power_law_1.2,0.08544639945030212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,8,power_law_1.2,0.08691840171813965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,8,power_law_1.2,0.0870848000049591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,8,power_law_1.2,0.08986240029335021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,8,power_law_1.2,0.08996480107307434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,8,power_law_1.2,0.09387519955635071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,8,power_law_1.2,0.08697599768638611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,8,power_law_1.2,0.09738879799842834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,8,balanced,2.8641173044840493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.2,0.0911296010017395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,8,power_law_1.2,0.1041983962059021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,8,power_law_1.2,0.10739840269088745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.2,0.10586880445480347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,8,power_law_1.2,0.12665599584579468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.2,0.11342079639434814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.2,0.1431615948677063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.2,0.14793599843978883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.2,0.18257919549942017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.2,0.17719039916992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.2,0.2038271903991699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.2,0.22979199886322021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.2,0.2569983959197998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.2,0.28499200344085696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.2,0.31400959491729735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,8,balanced,2.1900107065836587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.2,0.4375936031341553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.2,0.3988735914230347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.2,0.5555583953857421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.2,0.5513279914855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.2,0.7714623928070068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,8,power_law_1.2,0.7148672103881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.2,1.0848383903503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,8,power_law_1.2,1.106015968322754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,8,power_law_1.2,1.171513557434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,8,power_law_1.2,2.118764877319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,8,power_law_1.2,2.1556991577148437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,8,power_law_1.2,4.054399871826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,4,balanced,0.04577066500981649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,4,balanced,0.04537599782148997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,4,balanced,0.05295999844868978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,4,balanced,0.0780213326215744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,4,balanced,0.10389332969983418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,4,balanced,0.12099200487136841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,4,balanced,0.12411733468373616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,4,balanced,0.12123200297355652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,4,balanced,0.12098667025566101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,4,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,4,balanced,0.12331733107566833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,4,balanced,0.12364799777666728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,4,balanced,0.1237493356068929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,4,balanced,0.1251306633154551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,4,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,4,balanced,0.0499946673711141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,4,balanced,0.12847999731699625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,4,balanced,0.06028800209363302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,4,balanced,0.13427199920018515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,4,balanced,0.08272533118724823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,4,balanced,0.1374506652355194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,4,balanced,0.09252799550692241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,4,balanced,0.1421440045038859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,4,balanced,0.09361599882443745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,4,balanced,0.09422933061917622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,4,balanced,0.15449066956837973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,4,balanced,0.09498133261998494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,4,balanced,0.09331199526786804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,4,balanced,0.16702399651209512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,4,balanced,0.0965333382288615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,4,balanced,0.10033599535624187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,4,balanced,0.2091253399848938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,4,balanced,0.10051733255386353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,4,balanced,0.105295995871226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,4,power_law_1.01,0.053651201725006106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,4,balanced,0.2176533341407776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,4,balanced,0.11204266548156738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,4,balanced,0.11512533823649089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,4,power_law_1.01,0.0665727972984314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,4,balanced,0.12169599533081055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,4,balanced,0.2852213382720947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,4,power_law_1.01,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,4,balanced,0.13528533776601157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,4,power_law_1.01,0.07022719979286193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,4,balanced,0.14822933077812195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,4,balanced,0.3108160098393758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,4,power_law_1.01,0.07982720136642456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,4,power_law_1.01,0.10362240076065063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,4,balanced,0.18887466192245483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,4,power_law_1.01,0.10935039520263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,4,power_law_1.01,0.11134719848632812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,4,power_law_1.01,0.11694719791412353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,4,balanced,0.21110934019088745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,4,balanced,0.4513546625773112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,4,power_law_1.01,0.11354240179061889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,4,power_law_1.01,0.1161344051361084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,4,power_law_1.01,0.1205888032913208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,4,balanced,0.2929760018984477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,4,power_law_1.01,0.04948480129241943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,4,power_law_1.01,0.12154239416122437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,4,power_law_1.01,0.13076479434967042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,4,power_law_1.01,0.05611519813537598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,4,power_law_1.01,0.13722239732742308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,4,power_law_1.01,0.0541375994682312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,4,power_law_1.01,0.1385599970817566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,4,balanced,0.5313599904378256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,4,power_law_1.01,0.05985919833183288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,4,balanced,0.3316319982210795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,4,power_law_1.01,0.14610559940338136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,4,power_law_1.01,0.06535040140151978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,4,power_law_1.01,0.16168320178985596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,4,power_law_1.01,0.08343679904937744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,4,power_law_1.01,0.1802559971809387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,4,power_law_1.01,0.08558080196380616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,4,power_law_1.01,0.21818239688873292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,4,power_law_1.01,0.09000319838523865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,4,power_law_1.01,0.2515968084335327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,4,power_law_1.01,0.09157119989395142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,4,power_law_1.01,0.30145280361175536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,4,power_law_1.01,0.09264640212059021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,4,balanced,0.4771893421808879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,4,power_law_1.01,0.09318400025367737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,4,power_law_1.01,0.3977407932281494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,4,power_law_1.01,0.1022528052330017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,4,power_law_1.01,0.5243135929107666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,4,power_law_1.01,0.10686719417572021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,4,power_law_1.01,0.6852672100067139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,4,power_law_1.01,0.10606080293655396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,4,balanced,0.788383960723877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,4,power_law_1.01,0.9153599739074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,4,power_law_1.01,0.11352959871292115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,4,power_law_1.01,1.2174528121948243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,4,power_law_1.01,0.11958400011062623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,4,power_law_1.01,0.12540160417556762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,4,power_law_1.01,1.443609619140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,4,balanced,0.5967733462651571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,4,power_law_1.01,0.1420415997505188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,4,power_law_1.01,2.2345407485961912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,4,power_law_1.01,0.16542719602584838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,4,power_law_1.2,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,4,power_law_1.01,0.21310720443725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,4,power_law_1.01,0.24595201015472412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,4,power_law_1.01,4.102495956420898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,4,power_law_1.2,0.049497601389884946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,4,power_law_1.01,0.3113600015640259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,4,power_law_1.2,0.05560960173606873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,4,power_law_1.01,0.3980992078781128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,4,power_law_1.2,0.06289280056953431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,4,power_law_1.01,0.5695615768432617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,4,power_law_1.2,0.07780479788780212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,4,power_law_1.01,0.656550407409668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,4,power_law_1.2,0.1027519941329956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,4,balanced,0.9311359723409017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,4,power_law_1.01,1.003654384613037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,4,balanced,0.8400746981302897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,4,power_law_1.2,0.10803200006484985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,4,power_law_1.01,1.2774720191955566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,4,power_law_1.2,0.10986239910125732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,4,power_law_1.2,0.11342719793319703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,4,power_law_1.01,1.5220607757568358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,4,power_law_1.2,0.11340800523757935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,4,power_law_1.01,2.76300163269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,4,power_law_1.2,0.11597440242767335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,4,power_law_1.2,0.12205439805984497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,4,power_law_1.2,0.1232192039489746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,4,power_law_1.01,4.785772705078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,4,power_law_1.2,0.13082239627838135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,4,power_law_1.2,0.1365504026412964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,4,power_law_1.2,0.13690240383148194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,4,power_law_1.2,0.05070080161094666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,4,power_law_1.2,0.15253119468688964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,4,balanced,1.0811200141906738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,4,power_law_1.2,0.17375999689102173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,4,power_law_1.2,0.04754559993743897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,4,power_law_1.2,0.19095040559768678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,4,balanced,1.0911093552907307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,4,power_law_1.2,0.05454080104827881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,4,power_law_1.2,0.23607680797576905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,4,power_law_1.2,0.06037120223045349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,4,power_law_1.2,0.2641855955123901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,4,power_law_1.2,0.06282879710197449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,4,power_law_1.2,0.08305280208587647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,4,power_law_1.2,0.34057600498199464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,4,power_law_1.2,0.08683519959449768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,4,power_law_1.2,0.4161087989807129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,4,power_law_1.2,0.08944640159606934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,4,power_law_1.2,0.5521344184875489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,4,power_law_1.2,0.09247999787330627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,4,power_law_1.2,0.7806399822235107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,4,power_law_1.2,0.09637759923934937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,4,power_law_1.2,1.0881728172302245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,4,power_law_1.2,0.09930880069732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,4,power_law_1.2,0.10279680490493774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,4,power_law_1.2,1.3016575813293456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,4,power_law_1.2,0.10685440301895141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,4,power_law_1.2,1.4067008018493652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,4,power_law_1.2,0.10944000482559205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,4,balanced,1.3035306930541992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,4,power_law_1.2,0.11392639875411988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,4,power_law_1.2,2.5938112258911135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,4,power_law_1.2,0.11991039514541627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,4,power_law_1.2,0.12709759473800658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,4,power_law_1.2,5.143558502197266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,4,power_law_1.2,0.15040639638900757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,4,power_law_1.2,0.17847039699554443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,4,power_law_1.2,0.21631360054016113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,4,power_law_1.2,0.2566272020339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,4,power_law_1.2,0.34050559997558594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,4,power_law_1.2,0.4484416007995605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,4,balanced,1.7315467198689778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,4,power_law_1.2,0.6034495830535889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,4,power_law_1.2,0.7609983921051026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,4,power_law_1.2,1.128326416015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,4,power_law_1.2,1.4482687950134276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,4,power_law_1.2,1.8389055252075195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,4,power_law_1.2,2.8969600677490233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,4,power_law_1.2,5.3191169738769535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,4,balanced,2.042479991912842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,2,balanced,0.04595200220743815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,2,balanced,0.05083733300367991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,2,balanced,0.0572266678015391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,2,balanced,0.0831413318713506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,2,balanced,0.1165120005607605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,2,balanced,0.13363200426101685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,2,balanced,0.1346453328927358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,2,balanced,0.1325226624806722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,2,balanced,0.13331199685732523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,2,balanced,0.1349066694577535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,2,balanced,0.13726933797200522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,2,balanced,0.1386560002962748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,2,balanced,0.1428000032901764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,2,balanced,0.14760532975196838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,2,balanced,0.1546933352947235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,2,balanced,0.15798933307329813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,2,balanced,0.17015467087427774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,2,balanced,0.1858773430188497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,2,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,2,balanced,0.20082666476567587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,2,balanced,0.04771733283996582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,2,balanced,0.2644266684850057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,2,balanced,0.051872000098228455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,4,balanced,3.3036108016967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,2,balanced,0.06781333188215892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,2,balanced,0.28592000404993695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,2,balanced,0.09450133641560872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,2,balanced,0.10983467102050781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,2,balanced,0.11292800307273865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,2,balanced,0.420965313911438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,2,balanced,0.10636799534161885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,2,balanced,0.10604266325632732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,2,balanced,0.10987200339635213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,2,balanced,0.44531198342641193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,2,balanced,0.11084266503651936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,2,balanced,0.11424533526102702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,2,balanced,0.1157973309357961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,2,balanced,0.6787839730580648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,2,balanced,0.12546666463216147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,2,balanced,0.13184000054995218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,2,balanced,0.13692800203959146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,4,balanced,3.971461296081543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,2,balanced,0.14708266655604044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,2,balanced,0.16777600844701132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,2,balanced,0.7826346556345621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,2,balanced,0.18337599436442056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,2,balanced,0.2532266577084859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,2,balanced,0.2707680066426595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,2,balanced,1.2139519850413005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,2,balanced,0.3983413378397624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,2,power_law_1.01,0.04659200012683869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,2,power_law_1.01,0.05135999917984009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,2,balanced,0.4680960178375244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,2,power_law_1.01,0.052262401580810545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,2,power_law_1.01,0.06920959949493408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,2,power_law_1.01,0.08300799727439881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,2,power_law_1.01,0.11098239421844483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,2,power_law_1.01,0.11636480093002319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,2,power_law_1.01,0.12323839664459228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,2,power_law_1.01,0.12300159931182861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,2,balanced,0.7051626841227213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,2,power_law_1.01,0.12717440128326415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,2,power_law_1.01,0.13012479543685912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,2,power_law_1.01,0.13527679443359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,2,balanced,1.45360533396403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,2,power_law_1.01,0.14130560159683228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,2,power_law_1.01,0.04570879936218262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,2,power_law_1.01,0.1477952003479004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,2,power_law_1.01,0.15607039928436278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,2,balanced,0.8527200222015381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,2,power_law_1.01,0.046911999583244324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,2,power_law_1.01,0.15846400260925292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,2,power_law_1.01,0.1747712016105652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,2,power_law_1.01,0.050918400287628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,2,power_law_1.01,0.19431040287017823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,2,power_law_1.01,0.06213759779930115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,2,power_law_1.01,0.22704639434814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,2,power_law_1.01,0.07316480278968811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,2,power_law_1.01,0.2685631990432739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,2,power_law_1.01,0.09333119988441467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,2,power_law_1.01,0.09919999837875366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,2,power_law_1.01,0.33892478942871096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,2,power_law_1.01,0.10190080404281616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,2,power_law_1.01,0.432755184173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,2,power_law_1.01,0.10575360059738159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,2,power_law_1.01,0.5284031867980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,2,power_law_1.01,0.10607359409332276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,2,power_law_1.01,0.692467212677002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,2,power_law_1.01,0.11227519512176513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,2,balanced,1.282597303390503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,2,power_law_1.01,0.9211392402648926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,2,power_law_1.01,0.1181439995765686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,2,balanced,1.699445406595866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,2,power_law_1.01,0.12245759963989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,2,power_law_1.01,1.2644288063049316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,2,power_law_1.01,0.12341760396957398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,2,power_law_1.01,1.5819968223571776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,2,power_law_1.01,0.1342463970184326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,2,power_law_1.01,0.1419648051261902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,2,power_law_1.01,1.9080255508422852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,2,power_law_1.01,0.1543936014175415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,2,power_law_1.01,3.0698944091796876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,2,power_law_1.01,0.18042240142822266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,2,power_law_1.01,0.20776960849761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,2,power_law_1.01,0.2533695936203003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,2,power_law_1.01,5.8010303497314455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,2,power_law_1.01,0.3090240001678467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,2,power_law_1.01,0.4178112030029297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,2,power_law_1.01,0.5402304172515869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,2,balanced,1.6103839874267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,2,power_law_1.01,0.7217984199523926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,2,power_law_1.01,0.9736448287963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,2,power_law_1.01,1.3242239952087402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,2,power_law_1.01,1.7276351928710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,2,power_law_1.01,2.1213760375976562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,2,power_law_1.2,0.04508799910545349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,2,power_law_1.01,3.5090049743652343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,2,power_law_1.2,0.047635200619697574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,2,power_law_1.2,0.050911998748779295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,2,balanced,2.7114133834838867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,2,power_law_1.01,6.6459907531738285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,2,power_law_1.2,0.06521599888801574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,2,power_law_1.2,0.08116480112075805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,2,power_law_1.2,0.11036800146102906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,2,power_law_1.2,0.1171072006225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,2,power_law_1.2,0.1235200047492981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,2,balanced,1.9448213577270508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,2,power_law_1.2,0.12275840044021606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,2,power_law_1.2,0.1287935972213745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,2,power_law_1.2,0.13212159872055054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,2,power_law_1.2,0.1381824016571045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,2,power_law_1.2,0.14250240325927735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,2,power_law_1.2,0.149126398563385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,2,power_law_1.2,0.156876802444458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,2,power_law_1.2,0.046003198623657225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,2,power_law_1.2,0.1603327989578247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,2,power_law_1.2,0.17820800542831422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,2,power_law_1.2,0.04611839950084686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,2,power_law_1.2,0.20647039413452148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,2,power_law_1.2,0.05118719935417175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,2,power_law_1.2,0.2204416036605835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,2,power_law_1.2,0.29747838973999025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,2,power_law_1.2,0.06111999750137329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,2,power_law_1.2,0.3600640058517456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,2,power_law_1.2,0.06903039813041686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,2,power_law_1.2,0.46543359756469727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,2,power_law_1.2,0.09631999731063842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,2,power_law_1.2,0.09907839894294738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,2,power_law_1.2,0.5324927806854248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,2,power_law_1.2,0.10307199954986572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,2,power_law_1.2,0.7361216068267822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,2,power_law_1.2,0.10658559799194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,2,power_law_1.2,0.10986239910125732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,2,power_law_1.2,0.9829183578491211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,2,power_law_1.2,0.11382399797439575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,2,power_law_1.2,1.3380928039550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,2,power_law_1.2,0.11879680156707764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,2,power_law_1.2,1.66943359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,2,power_law_1.2,0.1253056049346924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,2,balanced,3.043247858683268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,2,power_law_1.2,0.12579840421676636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,2,power_law_1.2,2.1441152572631834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,2,power_law_1.2,0.13757439851760864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,2,power_law_1.2,0.14426239728927612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,2,power_law_1.2,3.338700866699219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,2,power_law_1.2,0.15947519540786742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,2,power_law_1.2,0.1899135947227478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,2,power_law_1.2,5.880684661865234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,2,power_law_1.2,0.21409280300140382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,2,power_law_1.2,0.27114880084991455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,2,power_law_1.2,0.32750720977783204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,2,power_law_1.2,0.42629117965698243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,2,power_law_1.2,0.5316287994384765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,2,power_law_1.2,0.7492159843444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,2,power_law_1.2,1.0171392440795899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,2,balanced,5.249578793843587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,2,power_law_1.2,1.3833984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,2,power_law_1.2,1.8952064514160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,2,power_law_1.2,2.342099189758301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,2,power_law_1.2,3.434694290161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,2,power_law_1.2,6.802508544921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,2,balanced,5.914981206258138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,32,balanced,0.08266666531562805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,32,balanced,0.08261866867542267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,32,balanced,0.08291733264923096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,32,balanced,0.09750933448473613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,32,balanced,0.12576533357302347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,32,balanced,0.19142399231592813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,32,balanced,0.2558773358662923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,32,balanced,0.24500799179077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,32,balanced,0.2548639973004659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,32,balanced,0.26259734233220416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,32,balanced,0.26816000541051227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,32,balanced,0.2557973265647888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,32,balanced,0.2582826614379883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,32,balanced,0.06323733429114024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,32,balanced,0.2785759965578715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,32,balanced,0.06420266628265381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,32,balanced,0.262661337852478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,32,balanced,0.06348266700903575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,32,balanced,0.2552693287531535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,32,balanced,0.076773335536321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,32,balanced,0.27509866158167523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,32,balanced,0.09435199697812398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,32,balanced,0.1325440009435018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,32,balanced,0.2813760042190552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,32,balanced,0.16618133584658304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,32,balanced,0.16004266341527304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,32,balanced,0.28329066435496014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,32,balanced,0.15583466490109762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,32,balanced,0.15962666273117065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,32,balanced,0.2945706645647685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,32,balanced,0.1591200033823649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,32,balanced,0.1670560042063395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,32,balanced,0.16665599743525186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,32,balanced,0.305402676264445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,32,balanced,0.1715679963429769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,32,balanced,0.17449599504470825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,32,balanced,0.1732106606165568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,32,balanced,0.3326080044110616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,32,balanced,0.19446933269500732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,32,balanced,0.18490666151046753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,32,balanced,0.19889066616694132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,32,balanced,0.3531839847564697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,32,balanced,0.22482667366663614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,32,balanced,0.2411466638247172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,32,balanced,0.4142293135325114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,32,balanced,0.2876586715380351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,32,balanced,0.33157867193222046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,32,balanced,0.6067200104395548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,32,balanced,0.41898131370544434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,32,balanced,0.5623519817988077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,32,balanced,0.6698613166809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,32,balanced,0.7159573237101237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,32,balanced,0.8736000061035156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,32,balanced,0.9355839888254801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,32,balanced,1.165023962656657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,32,balanced,1.1964373588562012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,32,balanced,1.6982186635335286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,32,balanced,1.8146506945292156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,32,balanced,3.036709467569987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,32,balanced,3.431898752848307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,1,balanced,0.05014933149019877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,1,balanced,0.056133334835370384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,1,balanced,0.06845333178838094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,1,balanced,0.09403199950853984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,1,balanced,0.13778666655222574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,1,balanced,0.16065067052841187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,1,balanced,0.16055466731389365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,1,balanced,0.16404799620310465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,1,balanced,0.1646346648534139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,1,balanced,0.1688800056775411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,1,balanced,0.1687999963760376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,1,balanced,0.1752906640370687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,1,balanced,0.17637866735458374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,1,balanced,0.18345065911610922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,1,balanced,0.19169066349665323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,1,balanced,0.1991893251736959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,1,balanced,0.2084160049756368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,1,balanced,0.24999467531840006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,1,balanced,0.2741706569989522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,1,balanced,0.39268799622853595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,1,balanced,0.41758398214975995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,1,balanced,0.6265759865442911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,1,balanced,0.6930987040201823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,1,balanced,1.0379626750946045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,1,balanced,0.04799999793370565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,1,balanced,0.052015999952952065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,1,balanced,0.06084799766540527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,1,balanced,0.07868266602357228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,1,balanced,0.11655466755231221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,32,balanced,0.06425599753856659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,1,balanced,0.13355732957522073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,1,balanced,0.13458133737246195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,32,balanced,0.06420266628265381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,1,balanced,1.2744853496551514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,1,balanced,0.13952533404032388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,32,balanced,0.06436799963315327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,1,balanced,0.1418239971001943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,32,balanced,0.07444266478220622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,1,balanced,0.14281599720319113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,32,balanced,0.09715732932090759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,1,balanced,0.1474026640256246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,32,balanced,0.12971199552218118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,1,balanced,0.15130133430163065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,32,balanced,0.1699519952138265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,1,balanced,0.15704533457756042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,32,balanced,0.15823466579119363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,1,balanced,0.162581334511439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,1,power_law_1.01,0.04917759895324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,32,balanced,0.16370667020479837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,1,balanced,0.17961066961288452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,32,balanced,0.16637333234151205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,1,balanced,0.2213546633720398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,1,power_law_1.01,0.05490559935569763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,32,balanced,0.16618133584658304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,1,balanced,0.19744000832239786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,32,balanced,0.1581493318080902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,1,power_law_1.01,0.06298239827156067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,1,balanced,0.23970667521158853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,32,balanced,0.16586132844289145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,1,power_law_1.01,0.07955840229988098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,32,balanced,0.15877866744995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,1,power_law_1.01,0.09501439929008484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,1,balanced,0.2659306724866231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,1,balanced,1.9240907033284504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,32,balanced,0.16755733887354532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,1,power_law_1.01,0.1341055989265442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,32,power_law_1.01,0.19335039854049682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,1,balanced,0.3957013289133708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,1,power_law_1.01,0.1400383949279785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,32,balanced,0.1702400048573812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,32,power_law_1.01,0.2599744081497192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,1,power_law_1.01,0.14138879776000976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,1,balanced,0.41306134064992267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,32,balanced,0.17696533600489298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,1,power_law_1.01,0.14869120121002197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,32,power_law_1.01,0.2520064115524292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,1,power_law_1.01,0.1510591983795166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,32,balanced,0.18297600746154785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,1,power_law_1.01,0.15493760108947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,1,balanced,0.6309866507848104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,32,power_law_1.01,0.11712640523910522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,1,balanced,2.381829261779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,1,power_law_1.01,0.16721919775009156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,32,balanced,0.19338132937749228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,32,power_law_1.01,0.14172159433364867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,1,power_law_1.01,0.16764800548553466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,1,power_law_1.01,0.17525119781494142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,32,power_law_1.01,0.18695679903030396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,1,balanced,0.735765298207601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,32,balanced,0.20069867372512817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,1,power_law_1.01,0.18724479675292968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,32,power_law_1.01,0.18617600202560425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,1,power_law_1.01,0.19660799503326415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,32,power_law_1.01,0.20726399421691893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,1,power_law_1.01,0.21052160263061523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,32,balanced,0.21311465899149576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,32,power_law_1.01,0.20965759754180907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,1,power_law_1.01,0.25724799633026124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,1,balanced,1.119909365971883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,1,power_law_1.01,0.29703679084777834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,32,power_law_1.01,0.2131200075149536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,1,power_law_1.01,0.38375039100646974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,32,balanced,0.24580266078313193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,32,power_law_1.01,0.23739519119262695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,1,power_law_1.01,0.45853438377380373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,32,power_law_1.01,0.24027519226074218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,1,power_law_1.01,0.6122560024261474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,1,power_law_1.01,0.7542208194732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,32,power_law_1.01,0.24828801155090333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,1,balanced,1.3609066009521484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,1,power_law_1.01,1.0195199966430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,32,balanced,0.27480532725652057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,32,power_law_1.01,0.25627520084381106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,1,power_law_1.01,1.3753151893615723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,32,power_law_1.01,0.23408000469207763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,1,balanced,2.8472213745117188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,1,power_law_1.01,1.8917055130004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,32,power_law_1.01,0.23363840579986572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,1,power_law_1.01,2.3122943878173827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,32,power_law_1.01,0.2596800088882446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,32,power_law_1.01,0.2677119970321655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,32,balanced,0.3375893433888753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,1,power_law_1.01,2.843680000305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,1,power_law_1.01,0.048102399706840514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,32,power_law_1.01,0.26975359916687014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,1,power_law_1.01,4.4290111541748045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,1,power_law_1.01,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,32,power_law_1.01,0.31359999179840087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,1,power_law_1.2,0.04714879989624023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,1,balanced,2.0814080238342285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,1,power_law_1.01,0.062003201246261595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,1,power_law_1.01,8.661567687988281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,32,power_law_1.01,0.33935999870300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,32,balanced,0.48325332005818683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,1,power_law_1.2,0.052537602186203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,1,power_law_1.01,0.07499520182609558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,1,balanced,2.6119786898295083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,32,power_law_1.01,0.3984064102172852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,1,balanced,4.497386614481608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,1,power_law_1.2,0.06023039817810059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,1,power_law_1.01,0.0883903980255127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,32,power_law_1.01,0.4954048156738281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,1,power_law_1.2,0.07453439831733703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,1,power_law_1.01,0.11331199407577515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,32,balanced,0.5827786525090536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,32,power_law_1.01,0.6675648212432861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,1,power_law_1.2,0.08789759874343872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,1,power_law_1.01,0.12108160257339477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,1,power_law_1.2,0.13307520151138305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,32,power_law_1.01,0.7888959884643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,1,balanced,3.125125249226888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,1,power_law_1.01,0.128601598739624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,1,power_law_1.2,0.13768320083618163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,32,balanced,0.7713599999745687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,32,power_law_1.01,1.0664511680603028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,1,power_law_1.01,0.13173120021820067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,1,power_law_1.2,0.046982398629188536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,1,power_law_1.2,0.14300160408020018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,32,power_law_1.01,1.2865535736083984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,1,power_law_1.01,0.13640960454940795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,1,power_law_1.2,0.050335997343063356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,1,balanced,5.011119842529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,1,power_law_1.01,0.14755840301513673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,1,power_law_1.2,0.1471295952796936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,32,power_law_1.01,1.6725631713867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,1,power_law_1.2,0.05937920212745666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,1,power_law_1.01,0.15418879985809325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,32,balanced,0.9895839691162109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,1,power_law_1.2,0.15251840353012086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,1,power_law_1.2,0.07066879868507385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,32,power_law_1.01,2.4593599319458006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,1,power_law_1.01,0.16081279516220093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,1,balanced,8.642127990722656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,1,power_law_1.2,0.16120959520339967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,1,power_law_1.2,0.0837823987007141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,1,power_law_1.01,0.1733247995376587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,32,power_law_1.01,4.763353729248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,1,power_law_1.2,0.16442240476608277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,1,power_law_1.2,0.11386239528656006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,1,power_law_1.01,0.19371520280838012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,1,power_law_1.2,0.17290879487991334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,1,power_law_1.2,0.12282880544662475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,1,power_law_1.01,0.20497920513153076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,1,power_law_1.2,0.178220796585083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,1,power_law_1.2,0.12634880542755128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,1,power_law_1.2,0.19178240299224852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,1,power_law_1.2,0.1324288010597229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,1,power_law_1.01,0.20519680976867677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,1,power_law_1.2,0.1987264037132263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,1,power_law_1.2,0.1401919960975647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,1,power_law_1.01,0.2541248083114624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,1,power_law_1.2,0.2160128116607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,1,power_law_1.2,0.1476032018661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,1,power_law_1.01,0.2975167989730835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,1,power_law_1.2,0.2623039960861206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,1,power_law_1.2,0.15395840406417846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,1,power_law_1.01,0.3820159912109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,1,power_law_1.2,0.29983999729156496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,1,power_law_1.2,0.1604864001274109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,1,power_law_1.01,0.4575615882873535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,1,power_law_1.2,0.3847615957260132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,1,power_law_1.2,0.1740991950035095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,1,power_law_1.01,0.6209856033325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,1,power_law_1.2,0.19263999462127684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,1,power_law_1.2,0.4706560134887695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,1,power_law_1.01,0.7998528003692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,1,power_law_1.2,0.2046720027923584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,1,power_law_1.2,0.6197120189666748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,1,power_law_1.2,0.20556159019470216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,1,power_law_1.01,1.0986495971679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,1,power_law_1.2,0.7696640014648437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,1,power_law_1.2,0.256659197807312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,1,power_law_1.2,1.0308159828186034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,1,power_law_1.01,1.4656895637512206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,1,power_law_1.2,0.3001919984817505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,32,balanced,1.4927253723144531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,1,power_law_1.2,1.3955967903137207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,1,power_law_1.01,2.030624008178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,1,power_law_1.2,0.38894081115722656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,1,power_law_1.2,1.896076774597168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,1,power_law_1.01,2.6221439361572267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,1,power_law_1.2,0.46405758857727053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,1,power_law_1.2,0.6257408142089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,1,power_law_1.2,2.315494346618652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,1,power_law_1.01,3.215251159667969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,1,power_law_1.2,0.806272029876709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,1,power_law_1.2,2.850092887878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,1,power_law_1.01,4.950271987915039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,1,power_law_1.2,1.102412796020508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,1,power_law_1.2,1.4873279571533202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,1,power_law_1.2,4.414067077636719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,1,power_law_1.01,9.78136978149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,1,power_law_1.2,2.061471939086914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,1,power_law_1.2,8.644563293457031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,1,power_law_1.2,2.5928064346313477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,1,power_law_1.2,3.205344009399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,1,balanced,9.77351442972819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,1,power_law_1.2,4.967372894287109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,1,power_law_1.2,9.760684967041016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,32,power_law_1.01,0.1283903956413269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,32,power_law_1.01,0.15996160507202148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,32,power_law_1.01,0.15312000513076782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,32,power_law_1.01,0.0893119990825653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,32,balanced,2.7713918685913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,32,power_law_1.01,0.10976639986038209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,32,power_law_1.01,0.12330880165100097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,32,power_law_1.01,0.11769599914550781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,32,power_law_1.01,0.1333567976951599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,32,power_law_1.01,0.14135040044784547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,32,power_law_1.01,0.14896639585494995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,32,power_law_1.01,0.14036480188369752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,32,power_law_1.01,0.15018880367279053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,32,power_law_1.01,0.16012799739837646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,32,power_law_1.01,0.14762239456176757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,32,power_law_1.01,0.15834879875183105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,32,power_law_1.01,0.1738495945930481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,32,power_law_1.01,0.16215039491653443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,32,power_law_1.01,0.18332799673080444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,32,power_law_1.01,0.20567679405212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,32,power_law_1.01,0.23580799102783204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,32,power_law_1.01,0.25816960334777833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,32,power_law_1.01,0.3412224054336548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,32,power_law_1.01,0.3989759922027588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,32,power_law_1.01,0.5385407924652099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,32,power_law_1.01,0.6751103878021241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,32,power_law_1.01,0.9040703773498535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,32,power_law_1.01,1.2604415893554688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,32,power_law_1.01,1.447532844543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,32,power_law_1.01,2.178144073486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,32,power_law_1.01,4.3724609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,32,power_law_1.01,0.1404863953590393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,32,power_law_1.2,0.19317760467529296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,32,power_law_1.01,0.1781440019607544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,32,power_law_1.01,0.17359360456466674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,32,power_law_1.2,0.2606015920639038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,32,power_law_1.01,0.09317759871482849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,32,power_law_1.2,0.23260159492492677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,32,power_law_1.01,0.10848000049591064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,32,power_law_1.2,0.12150399684906006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,32,power_law_1.2,0.1264896035194397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,32,power_law_1.01,0.12967679500579835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,32,power_law_1.2,0.1499008059501648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,32,power_law_1.01,0.13340799808502196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,32,power_law_1.2,0.16444159746170045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,32,power_law_1.01,0.14219520092010499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,32,power_law_1.2,0.1578752040863037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,32,power_law_1.01,0.14320640563964843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,32,power_law_1.2,0.17096320390701295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,32,power_law_1.2,0.14064639806747437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,32,power_law_1.01,0.1410815954208374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,32,power_law_1.2,0.1954751968383789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,32,power_law_1.2,0.09152640104293823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,32,power_law_1.01,0.15434240102767943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,32,power_law_1.2,0.23335680961608887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,32,power_law_1.01,0.16187520027160646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,32,power_law_1.2,0.10103679895401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,32,power_law_1.2,0.2283776044845581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,32,power_law_1.01,0.14677120447158815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,32,power_law_1.2,0.11657600402832032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,32,power_law_1.2,0.22108159065246583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,32,power_law_1.01,0.1475775957107544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,32,power_law_1.2,0.11735039949417114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,32,power_law_1.01,0.14969600439071656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,32,power_law_1.2,0.21728639602661132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,32,power_law_1.01,0.15332479476928712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,32,power_law_1.2,0.1321727991104126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,32,power_law_1.2,0.2515968084335327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,32,power_law_1.01,0.1660863995552063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,32,power_law_1.2,0.25064959526062014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,32,power_law_1.2,0.14195200204849243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,32,power_law_1.01,0.1710911989212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,32,power_law_1.2,0.23300480842590332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,32,power_law_1.2,0.14816000461578369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,32,power_law_1.01,0.1927616000175476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,32,power_law_1.2,0.24584319591522216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,32,power_law_1.2,0.14644479751586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,32,power_law_1.01,0.2198335886001587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,32,power_law_1.2,0.24911999702453613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,32,power_law_1.2,0.14754559993743896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,32,power_law_1.01,0.2500672101974487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,32,power_law_1.2,0.139520001411438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,32,power_law_1.2,0.2659264087677002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,32,power_law_1.2,0.1520959973335266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,32,power_law_1.01,0.3036223888397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,32,power_law_1.2,0.29062399864196775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,32,power_law_1.2,0.17667200565338134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,32,power_law_1.01,0.35608320236206054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,32,power_law_1.2,0.15083520412445067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,32,power_law_1.01,0.47422080039978026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,32,power_law_1.2,0.14117759466171265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,32,power_law_1.2,0.3431936025619507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,32,power_law_1.01,0.5806911945343017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,32,power_law_1.2,0.16353919506072997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,32,power_law_1.2,0.09031040072441102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,32,power_law_1.2,0.3623487949371338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,32,power_law_1.01,0.8125184059143067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,32,power_law_1.2,0.10506880283355713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,32,power_law_1.2,0.1644927978515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,32,power_law_1.2,0.46985602378845215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,32,power_law_1.01,1.0145471572875977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,32,power_law_1.2,0.1179967999458313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,32,power_law_1.2,0.5688320159912109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,32,power_law_1.2,0.16678400039672853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,32,power_law_1.2,0.11917439699172974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,32,power_law_1.01,1.247270393371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,32,power_law_1.2,0.7179008007049561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,32,power_law_1.2,0.19011199474334717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,32,power_law_1.2,0.14274560213088988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,32,power_law_1.2,1.0358271598815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,32,power_law_1.01,1.8612735748291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,32,power_law_1.2,0.20897281169891357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,32,power_law_1.2,0.14778879880905152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,32,power_law_1.2,0.2512063980102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,32,power_law_1.2,0.1443071961402893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,32,power_law_1.2,1.2138367652893067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,32,power_law_1.2,0.2922048091888428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,32,power_law_1.2,0.14972800016403198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,32,power_law_1.01,3.7017406463623046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,32,power_law_1.2,1.7564352035522461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,32,power_law_1.2,0.38428800106048583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,32,power_law_1.2,0.1495743989944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,32,power_law_1.2,0.1422271966934204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,32,power_law_1.2,2.1044864654541016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,32,power_law_1.2,0.4472320079803467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,32,power_law_1.2,0.15841280221939086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,32,power_law_1.2,0.5750656127929688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,32,power_law_1.2,2.9991680145263673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,32,power_law_1.2,0.7825151920318604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,32,power_law_1.2,0.1577855944633484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,32,power_law_1.2,0.15511679649353027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,32,power_law_1.2,1.045798397064209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,32,power_law_1.2,6.939993286132813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,32,power_law_1.2,0.1726207971572876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,32,power_law_1.2,1.427187156677246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,32,power_law_1.2,0.1862912058830261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,32,power_law_1.2,0.1966976046562195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,32,power_law_1.2,1.6925695419311524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,32,power_law_1.2,0.2344127893447876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,32,power_law_1.2,2.7824127197265627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,32,power_law_1.2,0.2671168088912964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,32,power_law_1.2,0.3359616041183472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,32,power_law_1.2,6.235270309448242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,32,power_law_1.2,0.39750399589538576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,32,power_law_1.2,0.5139071941375732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,32,power_law_1.2,0.6337344169616699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,32,power_law_1.2,0.9175423622131348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,32,power_law_1.2,1.2597951889038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,32,power_law_1.2,1.5424063682556153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,32,power_law_1.2,2.1921407699584963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,32,power_law_1.2,4.680352020263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,16,balanced,0.05418666700522105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,16,balanced,0.058042665322621666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,16,balanced,0.054048001766204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,16,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,16,balanced,0.06685866912206014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,16,balanced,0.08853866656621297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,16,balanced,0.10745599865913391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,16,balanced,0.10754666725794475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,16,balanced,0.10922666390736897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,16,balanced,0.10795733332633972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,16,balanced,0.11006399989128113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,16,balanced,0.11157332857449849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,16,balanced,0.11220266421635945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,16,balanced,0.11460799972216289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,16,balanced,0.12059733271598816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,16,balanced,0.12156800429026286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,16,balanced,0.1285706659158071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,16,balanced,0.1394613285859426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,16,balanced,0.1453279952208201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,16,balanced,0.16830933094024658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,16,balanced,0.1850186586380005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,16,balanced,0.22313600778579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,16,balanced,0.2618933320045471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,16,balanced,0.3307093381881714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,16,balanced,0.4258933464686076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,16,balanced,0.5631466706593832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,16,balanced,0.7232960065205892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,16,balanced,0.8875093460083008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,16,power_law_1.01,0.06063359975814819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,16,power_law_1.01,0.09015039801597595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,16,power_law_1.01,0.09710720181465149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,16,power_law_1.01,0.06111360192298889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,16,power_law_1.01,0.06764159798622131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,16,power_law_1.01,0.08034560084342957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,16,power_law_1.01,0.08305919766426087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,16,power_law_1.01,0.09342079758644103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,16,power_law_1.01,0.09169279932975768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,16,power_law_1.01,0.09664000272750854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,16,power_law_1.01,0.09543039798736572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,16,power_law_1.01,0.0970624029636383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,16,power_law_1.01,0.10499199628829955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,16,power_law_1.01,0.10361599922180176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,16,power_law_1.01,0.11080960035324097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,16,power_law_1.01,0.11237119436264038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,16,power_law_1.01,0.12031359672546386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,16,power_law_1.01,0.13755520582199096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,16,power_law_1.01,0.15093120336532592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,16,power_law_1.01,0.17607040405273439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,16,balanced,1.347040017445882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,16,power_law_1.01,0.20578560829162598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,16,power_law_1.01,0.264684796333313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,16,power_law_1.01,0.29462399482727053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,16,power_law_1.01,0.39429759979248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,16,power_law_1.01,0.49793281555175783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,16,power_law_1.01,0.6853312015533447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,16,power_law_1.01,0.874675178527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,16,power_law_1.01,1.0865280151367187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,16,power_law_1.01,1.7596288681030274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,16,power_law_1.01,3.355014419555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,16,balanced,0.047781333327293396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,16,balanced,0.048063998421033226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,16,balanced,2.614400068918864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,16,balanced,0.04870399832725525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,16,balanced,0.053802669048309326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,16,balanced,0.07177066802978516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,16,balanced,0.10411733388900757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,16,balanced,0.1288640002409617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,16,balanced,0.12985600034395853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,16,balanced,0.12974933783213297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,16,balanced,0.1309653321901957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,16,balanced,0.13301333785057068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,16,balanced,0.134442667166392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,16,balanced,0.13590400417645773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,16,balanced,0.04562133550643921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,16,balanced,0.13831999897956848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,16,balanced,0.14460800091425577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,16,balanced,0.04791999856630961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,16,balanced,0.14729600151379904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,16,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,16,balanced,0.15388799707094827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,16,balanced,0.07075733443101247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,16,balanced,0.09676266709963481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,16,balanced,0.166101336479187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,16,balanced,0.14662933349609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,16,balanced,0.1986773411432902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,16,balanced,0.20042665799458823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,16,balanced,0.17524800697962442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,16,balanced,0.19877866903940836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,16,balanced,0.19924267133076987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,16,balanced,0.20075200001398721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,16,balanced,0.1978986660639445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,16,balanced,0.20258132616678873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,16,balanced,0.20331199963887533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,16,balanced,0.22378667195638022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,16,balanced,0.20492267608642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,16,balanced,0.21353065967559814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,16,balanced,0.21106133858362833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,16,balanced,0.2701599995295207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,16,balanced,0.218831996122996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,16,balanced,0.22578666607538858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,16,balanced,0.23397332429885864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,16,balanced,0.32817065715789795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,16,balanced,0.2520800034205119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,16,balanced,0.26790932814280194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,16,balanced,0.4209173520406087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,16,balanced,0.30164267619450885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,16,balanced,0.3344159921010335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,16,balanced,0.5109599828720093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,16,balanced,0.3920746644337972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,16,balanced,0.6950506369272867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,16,balanced,0.47763200600941974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,16,balanced,0.5926560163497925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,16,balanced,0.9007573127746582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,16,balanced,0.7628746827443441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,16,balanced,1.1143946647644043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,16,power_law_1.01,0.09678720235824585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,16,power_law_1.01,0.149452805519104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,16,power_law_1.01,0.19882240295410156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,16,power_law_1.01,0.08201599717140198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,16,power_law_1.2,0.06062080264091492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,16,power_law_1.01,0.10180480480194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,16,power_law_1.01,0.1253119945526123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,16,balanced,0.9570559660593668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,16,power_law_1.2,0.09009280204772949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,16,power_law_1.01,0.13664640188217164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,16,power_law_1.2,0.06531199812889099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,16,power_law_1.01,0.15368959903717042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,16,power_law_1.2,0.06120319962501526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,16,power_law_1.01,0.16777600049972535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,16,power_law_1.2,0.06644480228424073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,16,power_law_1.01,0.16386560201644898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,16,power_law_1.2,0.077183997631073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,16,power_law_1.01,0.1660416007041931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,16,power_law_1.2,0.081740802526474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,16,power_law_1.01,0.16799999475479127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,16,power_law_1.2,0.09200000166893005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,16,power_law_1.01,0.16759040355682372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,16,power_law_1.2,0.09301120042800903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,16,power_law_1.01,0.18350720405578613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,16,power_law_1.2,0.09279999732971192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,16,power_law_1.01,0.1852288007736206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,16,power_law_1.2,0.09144960045814514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,16,balanced,1.6711360613505046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,16,power_law_1.01,0.19958399534225463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,16,power_law_1.2,0.10016000270843506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,16,power_law_1.01,0.20857601165771483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,16,power_law_1.2,0.09813119769096375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,16,power_law_1.01,0.2195136070251465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,16,power_law_1.2,0.10425599813461303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,16,power_law_1.01,0.23252480030059813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,16,power_law_1.2,0.11319680213928222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,16,power_law_1.01,0.28402559757232665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,16,power_law_1.2,0.11646720170974731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,16,power_law_1.01,0.29392640590667723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,16,power_law_1.2,0.12423039674758911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,16,power_law_1.01,0.3875839948654175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,16,power_law_1.2,0.14057600498199463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,16,power_law_1.01,0.38279678821563723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,16,power_law_1.2,0.15285760164260864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,16,power_law_1.01,0.5215680122375488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,16,balanced,1.3296373685201008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,16,power_law_1.2,0.18045439720153808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,16,power_law_1.01,0.6210752010345459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,16,power_law_1.2,0.22120320796966553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,16,power_law_1.01,0.7518464088439941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,16,power_law_1.2,0.2671231985092163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,16,power_law_1.2,0.3203711986541748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,16,power_law_1.01,1.0607999801635741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,16,power_law_1.2,0.4273856163024902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,16,power_law_1.01,1.2410816192626952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,16,power_law_1.2,0.5236800193786622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,16,power_law_1.01,1.819046401977539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,16,power_law_1.2,0.7937535762786865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,16,power_law_1.2,1.00381441116333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,16,power_law_1.01,3.650892639160156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,16,power_law_1.2,1.1718079566955566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,16,power_law_1.2,1.9302207946777343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,16,power_law_1.2,3.8951999664306642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,16,power_law_1.01,0.07515519857406616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,16,power_law_1.01,0.1027135968208313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,16,power_law_1.01,0.12410880327224731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,16,balanced,3.266357421875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,16,power_law_1.01,0.06257920265197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,16,power_law_1.01,0.07478399872779846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,16,power_law_1.01,0.08976640105247498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,16,power_law_1.01,0.09403520226478576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,16,power_law_1.01,0.10863360166549682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,16,power_law_1.01,0.108787202835083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,16,power_law_1.01,0.11354240179061889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,16,power_law_1.01,0.11399680376052856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,16,balanced,2.571034590403239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,16,power_law_1.01,0.11916799545288086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,16,power_law_1.01,0.11806720495223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,16,power_law_1.01,0.12563199996948243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,16,power_law_1.01,0.1321023941040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,16,power_law_1.01,0.1375167965888977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,16,power_law_1.01,0.1451647996902466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,16,power_law_1.01,0.16085760593414306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,16,power_law_1.2,0.07503359913825988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,16,power_law_1.01,0.1812991976737976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,16,power_law_1.2,0.10506880283355713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,16,power_law_1.01,0.2178368091583252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,16,power_law_1.2,0.09177600145339966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,16,power_law_1.01,0.23939199447631837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,16,power_law_1.2,0.06124160289764404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,16,power_law_1.01,0.3086656093597412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,16,power_law_1.2,0.06938239932060242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,16,power_law_1.01,0.3604543924331665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,16,power_law_1.2,0.08680319786071777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,16,power_law_1.01,0.4759359836578369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,16,power_law_1.2,0.09409919977188111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,16,power_law_1.2,0.10252159833908081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,16,power_law_1.01,0.6033664226531983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,16,power_law_1.2,0.10311679840087891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,16,power_law_1.01,0.8309696197509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,16,power_law_1.2,0.10359679460525513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,16,power_law_1.2,0.14869760274887084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,16,power_law_1.01,1.1044032096862793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,16,power_law_1.2,0.11028480529785156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,16,power_law_1.2,0.15071359872817994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,16,power_law_1.2,0.10524159669876099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,16,power_law_1.01,1.3443391799926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,16,power_law_1.2,0.08164479732513427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,16,power_law_1.2,0.11017600297927857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,16,power_law_1.2,0.09441919922828675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,16,power_law_1.01,2.1212543487548827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,16,power_law_1.2,0.11404800415039062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,16,power_law_1.2,0.10209920406341552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,16,power_law_1.2,0.12079360485076904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,16,power_law_1.01,4.066976165771484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,16,power_law_1.2,0.126310396194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,16,power_law_1.2,0.13066879510879517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,16,power_law_1.2,0.1405184030532837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,16,power_law_1.2,0.14996479749679564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,16,power_law_1.2,0.15422719717025757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,16,power_law_1.2,0.16108160018920897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,16,power_law_1.2,0.1707584023475647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,16,power_law_1.2,0.1673151969909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,16,power_law_1.2,0.18414080142974854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,16,power_law_1.2,0.1635967969894409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,16,power_law_1.2,0.22999041080474852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,16,power_law_1.2,0.16525440216064452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,16,power_law_1.2,0.24844160079956054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,16,power_law_1.2,0.16810879707336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,16,power_law_1.2,0.33900160789489747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,16,power_law_1.2,0.1759104013442993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,16,power_law_1.2,0.37842559814453125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,16,power_law_1.2,0.1920575976371765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,16,power_law_1.2,0.5229695796966553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,16,power_law_1.2,0.19822720289230347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,16,power_law_1.2,0.643884801864624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,16,power_law_1.2,0.20613760948181153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,16,power_law_1.2,0.8917440414428711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,16,power_law_1.2,0.22001919746398926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,16,power_law_1.2,0.24371840953826904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,16,power_law_1.2,1.17457275390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,16,power_law_1.2,0.30100479125976565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,16,power_law_1.2,1.4435199737548827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,16,power_law_1.2,0.317305588722229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,16,power_law_1.2,2.3090944290161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,16,power_law_1.2,0.4213376045227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,16,power_law_1.2,0.45435519218444825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,16,power_law_1.2,4.5365760803222654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,16,power_law_1.2,0.5555391788482666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,16,power_law_1.2,0.6568831920623779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,16,power_law_1.2,0.9345855712890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,16,power_law_1.2,1.1849472045898437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,16,power_law_1.2,1.5898367881774902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,16,power_law_1.2,2.0556928634643556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,16,power_law_1.2,4.374003219604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,8,balanced,0.054144000013669334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,8,power_law_1.01,0.055308800935745236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,8,balanced,0.05592533449331919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,8,power_law_1.01,0.06501759886741638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,8,balanced,0.0582826683918635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,8,power_law_1.01,0.08856959939002991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,8,balanced,0.0621066689491272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,8,balanced,0.07239999870459239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,8,power_law_1.01,0.06295679807662964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,8,balanced,0.10057600339253743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,8,power_law_1.01,0.07263360023498536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,8,balanced,0.12187199791272481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,8,balanced,0.12331199645996094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,8,power_law_1.01,0.08881919980049133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,8,balanced,0.12345600128173828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,8,power_law_1.01,0.09125760197639465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,8,balanced,0.12610133488972983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,8,power_law_1.01,0.09920639991760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,8,balanced,0.1262079974015554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,8,power_law_1.01,0.106278395652771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,8,balanced,0.12854933738708496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,8,power_law_1.01,0.10774400234222412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,8,balanced,0.12965333461761475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,8,power_law_1.01,0.1081279993057251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,8,balanced,0.13127467036247253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,8,power_law_1.01,0.11201920509338378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,8,balanced,0.13797866304715475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,8,power_law_1.01,0.11605759859085082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,8,balanced,0.14166933298110962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,8,power_law_1.01,0.11587200164794922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,8,balanced,0.14412267009417215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,8,power_law_1.01,0.12686079740524292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,8,balanced,0.1572533349196116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,8,power_law_1.01,0.1343999981880188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,8,balanced,0.16716265678405762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,8,power_law_1.01,0.1396607995033264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,8,power_law_1.01,0.1545151948928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,8,balanced,0.18959999084472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,8,power_law_1.01,0.16432640552520753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,8,power_law_1.01,0.1955839991569519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,8,balanced,0.20932799577713013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,8,power_law_1.01,0.22442879676818847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,8,power_law_1.01,0.283788800239563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,8,power_law_1.01,0.32946560382843015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,8,balanced,0.2547786633173625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,8,power_law_1.01,0.4472383975982666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,8,power_law_1.01,0.5805888175964355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,8,power_law_1.01,0.8278464317321778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,8,balanced,0.30084800720214844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,8,power_law_1.01,1.066975975036621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,8,power_law_1.01,1.28023042678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,8,power_law_1.01,2.0159231185913087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,8,balanced,0.39179734388987225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,8,power_law_1.01,3.931321716308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,8,balanced,0.5168586572011312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,8,balanced,0.6881813208262125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,8,balanced,0.049829334020614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,8,balanced,0.05201066533724467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,8,balanced,0.05397866666316986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,8,balanced,0.07416533430417378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,8,balanced,0.09988799691200256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,8,balanced,0.15160533785820007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,8,balanced,0.20083200931549072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,8,balanced,0.8984053134918213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,8,balanced,0.20363734165827432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,8,balanced,0.20326934258143106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,8,balanced,0.2037013371785482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,8,balanced,0.20473599433898926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,8,balanced,0.21116799116134644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,8,balanced,0.208079993724823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,8,balanced,0.21218132972717285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,8,balanced,0.21592533588409424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,8,balanced,0.21978133916854858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,8,balanced,0.2251466711362203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,8,balanced,0.23776533206303915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,8,balanced,0.2442506750424703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,8,balanced,0.2675573428471883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,8,balanced,1.1108427047729492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,8,balanced,0.28752533594767254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,8,balanced,0.3327786723772685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,8,balanced,0.3803466558456421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,8,balanced,0.4540213346481323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,8,balanced,0.5527679920196533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,8,balanced,1.7048427263895671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,8,balanced,0.7152960300445557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,8,balanced,0.9145973523457845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,8,balanced,1.1503199736277263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,8,balanced,3.3178345362345376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,8,balanced,1.6301546096801758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,8,power_law_1.2,0.055212801694869994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,8,power_law_1.2,0.06500480175018311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,8,power_law_1.2,0.06262400150299072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,8,balanced,0.04574400186538696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,8,power_law_1.2,0.06350719928741455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,8,power_law_1.2,0.06822400093078614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,8,balanced,0.047770669062932335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,8,balanced,0.04969066878159841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,8,power_law_1.2,0.08289920091629029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,8,balanced,0.05420266588528951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,8,power_law_1.2,0.08489599823951721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,8,balanced,0.07712000111738841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,8,balanced,0.1063253382841746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,8,power_law_1.2,0.09880959987640381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,8,balanced,0.13224533200263977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,8,balanced,0.13544533650080362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,8,power_law_1.2,0.10142079591751099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,8,balanced,0.13499200344085693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,8,power_law_1.2,0.1051967978477478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,8,balanced,0.13524267077445984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,8,balanced,0.13707199692726135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,8,power_law_1.2,0.1078976035118103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,8,balanced,0.14124799768129984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,8,power_law_1.2,0.11141120195388794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,8,balanced,0.14282666643460593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,8,balanced,0.14460800091425577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,8,power_law_1.2,0.10899840593338013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,8,balanced,0.15276267131169638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,8,power_law_1.2,0.11952639818191528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,8,balanced,0.1557919979095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,8,power_law_1.2,0.12995840311050416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,8,balanced,0.16410666704177856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,8,power_law_1.2,0.1346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,8,balanced,0.1767680048942566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,8,power_law_1.2,0.1423359990119934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,8,balanced,0.18998932838439941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,8,power_law_1.2,0.15253119468688964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,8,power_law_1.2,0.16722559928894043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,8,balanced,0.21966399749120077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,8,power_law_1.2,0.19749759435653685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,8,power_law_1.2,0.22951679229736327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,8,balanced,0.24874132871627808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,8,power_law_1.2,0.29601919651031494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,8,power_law_1.2,0.34767999649047854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,8,balanced,0.2993173400561015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,8,power_law_1.2,0.4855487823486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,8,power_law_1.01,0.051648002862930295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,8,power_law_1.01,0.07398399710655212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,8,power_law_1.2,0.6470719814300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,8,power_law_1.01,0.07338240146636962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,8,balanced,0.36777599652608234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,8,power_law_1.2,0.875551986694336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,8,power_law_1.01,0.10153599977493286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,8,power_law_1.01,0.09847040176391601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,8,power_law_1.01,0.06101120114326477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,8,power_law_1.01,0.1447808027267456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,8,power_law_1.2,1.1659071922302247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,8,power_law_1.01,0.07014399766921997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,8,power_law_1.01,0.07264639735221863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,8,power_law_1.2,1.3686528205871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,8,power_law_1.01,0.08680959939956664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,8,power_law_1.01,0.09356799721717834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,8,power_law_1.01,0.09359999895095825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,8,power_law_1.01,0.12147200107574463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,8,balanced,0.4762880007425944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,8,balanced,3.1536852518717446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,8,power_law_1.2,2.148896026611328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,8,power_law_1.01,0.1065600037574768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,8,power_law_1.01,0.12472959756851196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,8,power_law_1.01,0.11080319881439209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,8,power_law_1.01,0.1454848051071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,8,power_law_1.2,4.478182220458985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,8,power_law_1.01,0.15942399501800536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,8,power_law_1.01,0.117740797996521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,8,power_law_1.01,0.1591423988342285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,8,power_law_1.01,0.11601279973983765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,8,power_law_1.01,0.16543359756469728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,8,balanced,0.5992000102996826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,8,power_law_1.01,0.12202880382537842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,8,power_law_1.01,0.17585279941558837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,8,power_law_1.01,0.12398079633712769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,8,power_law_1.01,0.17221759557723998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,8,power_law_1.01,0.13464319705963135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,8,power_law_1.01,0.18272000551223755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,8,power_law_1.01,0.14071040153503417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,8,power_law_1.01,0.1881600022315979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,8,power_law_1.01,0.14498560428619384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,8,power_law_1.01,0.19381760358810424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,8,power_law_1.01,0.15529600381851197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,8,balanced,0.8240746657053629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,8,power_law_1.01,0.1729599952697754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,8,power_law_1.01,0.20666239261627198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,8,power_law_1.01,0.22142720222473145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,8,power_law_1.01,0.19252480268478395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,8,power_law_1.01,0.23992319107055665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,8,power_law_1.01,0.22836480140686036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,8,power_law_1.01,0.2831615924835205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,8,power_law_1.01,0.2531775951385498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,8,power_law_1.01,0.3036992073059082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,8,power_law_1.01,0.32188799381256106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,8,power_law_1.01,0.3770240068435669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,8,power_law_1.01,0.3908544063568115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,8,power_law_1.01,0.41185917854309084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,8,power_law_1.01,0.5354944229125976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,8,power_law_1.01,0.5070528030395508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,8,power_law_1.01,0.6825215816497803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,8,power_law_1.01,0.6290815830230713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,8,power_law_1.01,0.9883071899414062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,8,power_law_1.01,0.8844096183776855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,8,balanced,1.0555360317230225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,8,power_law_1.01,1.2576064109802245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,8,power_law_1.01,1.0741312026977539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,8,power_law_1.01,1.5004799842834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,8,power_law_1.01,1.2842623710632324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,8,power_law_1.01,2.2576255798339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,8,power_law_1.01,2.0981632232666017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,8,power_law_1.01,4.6460929870605465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,8,power_law_1.01,3.91673583984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,8,balanced,1.3002399603525798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,8,power_law_1.2,0.06848639845848084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,8,power_law_1.2,0.09770240187644959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,8,power_law_1.2,0.05170559883117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,8,balanced,1.9837172826131184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,8,power_law_1.2,0.10212479829788208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,8,power_law_1.2,0.07448959946632386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,8,power_law_1.2,0.06851840019226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,8,power_law_1.2,0.08227840065956116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,8,power_law_1.2,0.05692160129547119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,8,power_law_1.2,0.10846079587936401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,8,power_law_1.2,0.0605184018611908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,8,power_law_1.2,0.11779199838638306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,8,power_law_1.2,0.0723904013633728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,8,power_law_1.2,0.145196795463562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,8,power_law_1.2,0.0821120023727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,8,power_law_1.2,0.1491520047187805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,8,power_law_1.2,0.09482880234718323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,8,power_law_1.2,0.15682560205459595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,8,power_law_1.2,0.10595840215682983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,8,power_law_1.2,0.16197119951248168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,8,power_law_1.2,0.10826239585876465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,8,power_law_1.2,0.1647871971130371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,8,power_law_1.2,0.11085439920425415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,8,power_law_1.2,0.1686527967453003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,8,power_law_1.2,0.11616640090942383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,8,power_law_1.2,0.17526400089263916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,8,power_law_1.2,0.12229119539260865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,8,power_law_1.2,0.18457599878311157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,8,power_law_1.2,0.1262336015701294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,8,power_law_1.2,0.1938752055168152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,8,power_law_1.2,0.20565760135650635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,8,power_law_1.2,0.13401600122451782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,8,power_law_1.2,0.14043519496917725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,8,power_law_1.2,0.2202752113342285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,8,power_law_1.2,0.15050239562988282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,8,power_law_1.2,0.24496641159057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,8,power_law_1.2,0.1547968029975891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,8,power_law_1.2,0.30323200225830077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,8,power_law_1.2,0.17730560302734374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,8,power_law_1.2,0.32101759910583494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,8,power_law_1.2,0.19526400566101074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,8,power_law_1.2,0.40473599433898927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,8,power_law_1.2,0.23837440013885497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,8,power_law_1.2,0.4615935802459717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,8,power_law_1.2,0.2717184066772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,8,power_law_1.2,0.5721407890319824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,8,power_law_1.2,0.3501823902130127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,8,power_law_1.2,0.681171178817749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,8,power_law_1.2,0.4161983966827393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,8,power_law_1.2,0.9369279861450195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,8,power_law_1.2,0.608896017074585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,8,power_law_1.2,1.2829824447631837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,8,power_law_1.2,0.7410880088806152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,8,power_law_1.2,1.0243519783020019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,8,power_law_1.2,1.4754560470581055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,8,power_law_1.2,1.2811455726623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,8,power_law_1.2,2.2232255935668945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,8,power_law_1.2,1.6430976867675782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,8,power_law_1.2,4.630867385864258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,8,power_law_1.2,2.6232128143310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,8,power_law_1.2,4.803942489624023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,8,balanced,3.870512008666992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,4,balanced,0.05589866638183594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,4,balanced,0.060229331254959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,4,balanced,0.0621973325808843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,4,balanced,0.06818133095900218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,4,balanced,0.08613333106040955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,4,balanced,0.12401599685351054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,4,balanced,0.1564853290716807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,4,balanced,0.158842662970225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,4,balanced,0.15958399573961893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,4,balanced,0.1607146660486857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,4,balanced,0.15548800428708395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,4,balanced,0.1597653329372406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,4,balanced,0.16059199968973795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,4,balanced,0.16224533319473267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,4,balanced,0.1694506605466207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,4,balanced,0.17246933778127035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,4,balanced,0.18245333433151245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,4,balanced,0.19707733392715454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,4,balanced,0.21017066637674967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,4,balanced,0.23828800519307455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,4,balanced,0.26269867022832233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,4,balanced,0.32468267281850177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,4,balanced,0.3894346555074056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,4,power_law_1.01,0.05731840133666992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,4,balanced,0.5176479816436768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,4,power_law_1.01,0.06267520189285278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,4,power_law_1.01,0.07688959836959838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,4,power_law_1.01,0.07031040191650391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,4,power_law_1.01,0.08190720081329346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,4,power_law_1.01,0.09786239862442017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,4,power_law_1.01,0.1115007996559143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,4,balanced,0.6861173311869303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,4,power_law_1.01,0.12360320091247559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,4,power_law_1.01,0.12866560220718384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,4,power_law_1.01,0.12899199724197388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,4,power_law_1.01,0.13198720216751098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,4,power_law_1.01,0.14165760278701783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,4,power_law_1.01,0.14090880155563354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,4,power_law_1.01,0.14417920112609864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,4,power_law_1.01,0.15926400423049927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,4,power_law_1.01,0.16097919940948485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,4,balanced,0.934063990910848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,4,power_law_1.01,0.16990079879760742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,4,power_law_1.01,0.18581119775772095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,4,power_law_1.01,0.20378239154815675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,4,power_law_1.01,0.2445120096206665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,4,power_law_1.01,0.2749631881713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,4,power_law_1.01,0.346943998336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,4,power_law_1.01,0.428985595703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,4,power_law_1.01,0.5892032146453857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,4,power_law_1.01,0.7827839851379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,4,power_law_1.01,1.0560704231262208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,4,balanced,1.2422613302866619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,4,power_law_1.01,1.3251392364501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,4,power_law_1.01,1.7457279205322265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,4,power_law_1.01,2.6413248062133787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,4,power_law_1.01,4.998892974853516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,4,balanced,1.5361067454020183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,4,balanced,2.36025603612264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,4,balanced,4.529269218444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,4,balanced,0.04975466430187225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,4,balanced,0.0518453319867452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,4,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,4,balanced,0.07811200122038524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,4,balanced,0.10430399576822917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,4,balanced,0.15665066242218018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,4,balanced,0.2093600034713745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,4,balanced,0.20971733331680298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,4,balanced,0.2076586683591207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,4,balanced,0.21176532904307047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,4,balanced,0.214026669661204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,4,balanced,0.21684267123540243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,4,balanced,0.21930134296417236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,4,balanced,0.22579733530680338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,4,balanced,0.2288960019747416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,4,balanced,0.23152534166971842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,4,balanced,0.24267733097076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,4,balanced,0.2582133412361145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,4,balanced,0.2722559968630473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,4,balanced,0.30156266689300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,4,balanced,0.3288053274154663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,4,balanced,0.38393068313598633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,4,balanced,0.45684266090393066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,4,balanced,0.5697386662165324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,4,balanced,0.71233598391215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,4,power_law_1.2,0.057158398628234866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,4,power_law_1.2,0.06282240152359009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,4,power_law_1.2,0.06357759833335877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,4,power_law_1.2,0.0705344021320343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,4,balanced,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,4,balanced,0.9667253494262695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,4,power_law_1.2,0.07918720245361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,4,balanced,0.050010666251182556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,4,balanced,0.05374933282534281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,4,power_law_1.2,0.09303680062294006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,4,balanced,0.058965335289637245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,4,power_law_1.2,0.10462080240249634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,4,balanced,0.08409600456555684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,4,balanced,0.11551466584205627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,4,power_law_1.2,0.12000000476837158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,4,balanced,0.14633066455523172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,4,power_law_1.2,0.12537599802017213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,4,balanced,0.14734400312105814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,4,power_law_1.2,0.12735359668731688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,4,balanced,0.15033599734306335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,4,power_law_1.2,0.13069440126419068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,4,balanced,0.1492853363355001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,4,balanced,0.14654399951299033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,4,power_law_1.2,0.13719680309295654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,4,balanced,0.15106667081514993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,4,power_law_1.2,0.1400256037712097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,4,balanced,0.1518186628818512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,4,power_law_1.2,0.13954559564590455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,4,balanced,0.1560373306274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,4,power_law_1.2,0.15152000188827514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,4,balanced,0.16471466422080994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,4,balanced,1.2061973412831624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,4,power_law_1.2,0.16100480556488037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,4,balanced,0.16741865873336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,4,balanced,0.1797599991162618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,4,power_law_1.2,0.1726591944694519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,4,power_law_1.2,0.18971519470214843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,4,balanced,0.1994826594988505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,4,power_law_1.2,0.20817279815673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,4,balanced,0.22759467363357544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,4,power_law_1.2,0.25100159645080566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,4,power_law_1.2,0.28645761013031007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,4,balanced,0.2630666693051656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,4,power_law_1.2,0.3807935953140259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,4,balanced,0.2914186716079712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,4,power_law_1.2,0.43999361991882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,4,power_law_1.2,0.5989823818206788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,4,balanced,0.36302932103474933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,4,power_law_1.2,0.7935488224029541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,4,power_law_1.2,1.1761919975280761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,4,power_law_1.2,1.4598591804504395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,4,balanced,0.45869867006937665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,4,balanced,1.529088020324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,4,power_law_1.2,1.7091072082519532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,4,power_law_1.2,3.014796829223633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,4,balanced,0.5994826555252075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,4,power_law_1.2,5.43884162902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,4,balanced,0.7767786979675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,4,power_law_1.01,0.048742398619651794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,4,power_law_1.01,0.05553920269012451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,4,power_law_1.01,0.05430399775505066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,4,power_law_1.01,0.07957760095596314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,4,power_law_1.01,0.07116159796714783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,4,power_law_1.01,0.05896959900856018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,4,power_law_1.01,0.09950720071792603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,4,power_law_1.01,0.07341439723968506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,4,power_law_1.01,0.06798719763755798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,4,balanced,1.0804853439331055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,4,power_law_1.01,0.09073280096054077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,4,power_law_1.01,0.09057919979095459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,4,power_law_1.01,0.09831680059432983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,4,power_law_1.01,0.11504640579223632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,4,power_law_1.01,0.11442559957504272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,4,power_law_1.01,0.12749439477920532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,4,balanced,2.245541254679362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,4,power_law_1.01,0.11847039461135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,4,power_law_1.01,0.15383039712905883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,4,power_law_1.01,0.121343994140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,4,power_law_1.01,0.15631359815597534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,4,power_law_1.01,0.12655999660491943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,4,power_law_1.01,0.16316800117492675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,4,power_law_1.01,0.1306048035621643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,4,power_law_1.01,0.16516480445861817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,4,power_law_1.01,0.13690880537033082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,4,power_law_1.01,0.16970239877700805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,4,power_law_1.01,0.1471935987472534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,4,power_law_1.01,0.1774656057357788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,4,balanced,1.3749386469523113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,4,power_law_1.01,0.1578495979309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,4,power_law_1.01,0.18353279829025268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,4,power_law_1.01,0.16232320070266723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,4,power_law_1.01,0.20235519409179686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,4,power_law_1.01,0.1784000039100647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,4,power_law_1.01,0.20613760948181153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,4,power_law_1.01,0.21477119922637938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,4,power_law_1.01,0.21290879249572753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,4,power_law_1.01,0.219321608543396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,4,power_law_1.01,0.24311039447784424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,4,power_law_1.01,0.26382079124450686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,4,power_law_1.01,0.2629823923110962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,4,power_law_1.01,0.3017600059509277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,4,power_law_1.01,0.3004224061965942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,4,power_law_1.01,0.3859647989273071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,4,power_law_1.01,0.33844480514526365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,4,power_law_1.01,0.5087999820709228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,4,power_law_1.01,0.4130239963531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,4,power_law_1.01,0.6771584033966065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,4,power_law_1.01,0.49755520820617677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,4,power_law_1.01,0.8193344116210938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,4,power_law_1.01,0.630847978591919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,4,balanced,1.7240427335103352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,4,power_law_1.01,1.209823989868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,4,power_law_1.01,0.7782976150512695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,4,power_law_1.01,1.5198847770690918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,4,power_law_1.01,1.0929471969604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,4,power_law_1.01,1.8927423477172851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,4,power_law_1.01,1.4176511764526367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,4,power_law_1.01,2.9672191619873045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,4,power_law_1.01,1.687424087524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,4,power_law_1.01,2.548384094238281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,4,power_law_1.01,5.605548858642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,4,power_law_1.01,4.8280384063720705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,4,balanced,2.5979785919189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,4,balanced,4.330405235290527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,4,power_law_1.2,0.048556798696517946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,4,power_law_1.2,0.05358080267906189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,4,power_law_1.2,0.05551999807357788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,4,power_law_1.2,0.07118719816207886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,4,power_law_1.2,0.06279039978981019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,4,power_law_1.2,0.07518720030784606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,4,power_law_1.2,0.05906559824943543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,4,power_law_1.2,0.06789759993553161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,4,power_law_1.2,0.07074559926986694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,4,power_law_1.2,0.08777599930763244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,4,power_law_1.2,0.085452800989151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,4,power_law_1.2,0.1028223991394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,4,power_law_1.2,0.0930624008178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,4,power_law_1.2,0.12138880491256714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,4,power_law_1.2,0.10960639715194702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,4,power_law_1.2,0.1552575945854187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,4,power_law_1.2,0.11639039516448975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,4,power_law_1.2,0.1519744038581848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,4,power_law_1.2,0.11719039678573609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,4,power_law_1.2,0.15701760053634645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,4,power_law_1.2,0.16456320285797119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,4,power_law_1.2,0.12413439750671387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,4,power_law_1.2,0.17288960218429567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,4,power_law_1.2,0.1336192011833191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,4,power_law_1.2,0.17863039970397948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,4,power_law_1.2,0.13776639699935914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,4,power_law_1.2,0.18227200508117675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,4,power_law_1.2,0.14357759952545165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,4,power_law_1.2,0.19818880558013915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,4,power_law_1.2,0.15779839754104613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,4,power_law_1.2,0.20797441005706788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,4,power_law_1.2,0.16588799953460692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,4,power_law_1.2,0.21502718925476075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,4,power_law_1.2,0.1823232054710388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,4,power_law_1.2,0.24585599899291993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,4,power_law_1.2,0.21365759372711182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,4,power_law_1.2,0.26530559062957765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,4,power_law_1.2,0.23352320194244386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,4,power_law_1.2,0.3084800004959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,4,power_law_1.2,0.28365440368652345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,4,power_law_1.2,0.36085119247436526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,4,balanced,5.055055936177571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,4,power_law_1.2,0.32495999336242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,4,power_law_1.2,0.4372608184814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,4,power_law_1.2,0.415558385848999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,4,power_law_1.2,0.5483903884887695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,4,power_law_1.2,0.5387519836425781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,4,power_law_1.2,0.6821568012237549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,4,power_law_1.2,0.6886655807495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,4,power_law_1.2,0.8454463958740235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,4,power_law_1.2,0.8722432136535645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,4,power_law_1.2,1.1496000289916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,4,power_law_1.2,1.339622402191162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,4,power_law_1.2,1.4971072196960449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,4,power_law_1.2,1.6259136199951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,4,power_law_1.2,1.806559944152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,4,power_law_1.2,2.0008256912231444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,4,power_law_1.2,2.7808767318725587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,4,power_law_1.2,2.985670471191406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,4,power_law_1.2,5.750080108642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,4,power_law_1.2,5.708870315551758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,2,balanced,0.07439466814200084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,2,balanced,0.07673066854476929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,2,balanced,0.08253333469231923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,2,balanced,0.09308800101280212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,2,balanced,0.12361066540082295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,2,balanced,0.1832746664683024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,2,balanced,0.23467199007670084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,2,balanced,0.22554665803909302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,2,balanced,0.22644799947738647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,2,balanced,0.22713067134221396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,2,balanced,0.2282080054283142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,2,balanced,0.23406400283177695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,2,balanced,0.23305066426595053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,2,balanced,0.23843199014663696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,2,balanced,0.2485706607500712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,2,balanced,0.25110934178034466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,2,balanced,0.2640906572341919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,2,balanced,0.28892799218495685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,2,balanced,0.3073493242263794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,2,balanced,0.3577813307444255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,2,balanced,0.39875201384226483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,2,balanced,0.4995466470718384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,2,balanced,0.6057440042495728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,2,balanced,0.8138186931610107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,2,balanced,1.105141321818034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,2,balanced,1.5271306037902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,2,power_law_1.01,0.07517439723014832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,2,power_law_1.01,0.08045439720153809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,2,power_law_1.01,0.09120000004768372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,2,balanced,2.026591936747233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,2,power_law_1.01,0.09230719804763794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,2,power_law_1.01,0.11085439920425415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,2,power_law_1.01,0.13726719617843627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,2,power_law_1.01,0.15240319967269897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,2,power_law_1.01,0.17264000177383423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,2,power_law_1.01,0.17783679962158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,2,power_law_1.01,0.1865664005279541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,2,power_law_1.01,0.1922047972679138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,2,power_law_1.01,0.19884159564971923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,2,power_law_1.01,0.2033600091934204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,2,power_law_1.01,0.21187200546264648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,2,power_law_1.01,0.22188160419464112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,2,power_law_1.01,0.2316864013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,2,power_law_1.01,0.2462847948074341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,2,power_law_1.01,0.2689471960067749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,2,balanced,2.5052053133646646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,2,power_law_1.01,0.29511680603027346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,2,power_law_1.01,0.3622335910797119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,2,power_law_1.01,0.41483521461486816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,2,power_law_1.01,0.5312704086303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,2,power_law_1.01,0.651423978805542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,2,power_law_1.01,0.8822336196899414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,2,power_law_1.01,1.1299903869628907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,2,power_law_1.01,1.6182527542114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,2,power_law_1.01,2.0480255126953124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,2,power_law_1.01,2.556096076965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,2,power_law_1.01,4.012985610961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,2,power_law_1.01,7.620038604736328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,2,balanced,3.863663991292318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,2,balanced,0.05209066470464071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,2,balanced,0.05201066533724467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,2,balanced,0.059845333298047386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,2,balanced,0.07863466441631317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,2,balanced,0.11285866300264995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,2,balanced,0.1727893352508545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,2,balanced,0.2343519926071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,2,balanced,0.2397813399632772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,2,balanced,0.2437653342882792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,2,balanced,0.24617600440979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,2,balanced,0.24212799469629923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,2,balanced,0.24487467606862387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,2,balanced,0.24699199199676514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,2,balanced,0.24859732389450073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,2,balanced,0.25917333364486694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,2,balanced,0.26129066944122314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,2,balanced,0.27481067180633545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,2,balanced,0.30135999123255414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,2,balanced,0.3198773264884949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,2,balanced,7.420410792032878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,2,balanced,0.37107733885447186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,2,balanced,0.41025598843892414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,2,balanced,0.495904008547465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,2,balanced,0.6100053389867147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,2,balanced,0.7862719694773356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,2,balanced,1.0272746880849202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,2,balanced,1.3750240008036296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,2,balanced,1.8299627304077148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,2,balanced,2.308821360270182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,2,balanced,0.04975466430187225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,2,balanced,0.05202666421731313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,2,balanced,0.05622399846712748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,2,balanced,0.06596266726652782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,2,balanced,0.09504533807436626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,2,balanced,0.13870400190353394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,2,balanced,0.17468800147374472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,2,balanced,0.17598400513331094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,2,balanced,0.17840532461802164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,2,balanced,0.18039999405543009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,2,power_law_1.2,0.07119359970092773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,2,balanced,0.1811093290646871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,2,balanced,0.18781334161758423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,2,balanced,0.18781334161758423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,2,power_law_1.2,0.07715200185775757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,2,balanced,0.19368000825246176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,2,power_law_1.2,0.08140159845352173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,2,balanced,0.20458666483561197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,2,balanced,3.367055892944336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,2,power_law_1.2,0.09277439713478089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,2,balanced,0.20759467283884683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,2,power_law_1.2,0.10675840377807617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,2,balanced,0.22430932521820068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,2,power_law_1.2,0.12883199453353883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,2,balanced,0.25220266977945965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,2,power_law_1.2,0.14480639696121217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,2,balanced,0.27180800835291546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,2,power_law_1.2,0.1692352056503296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,2,power_law_1.2,0.1784767985343933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,2,balanced,0.32707200447718304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,2,power_law_1.2,0.18358399868011474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,2,power_law_1.2,0.1860479950904846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,2,balanced,0.36739198366800946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,2,power_law_1.2,0.19296000003814698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,2,power_law_1.2,0.19986560344696044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,2,balanced,0.47306132316589355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,2,power_law_1.2,0.2070336103439331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,2,power_law_1.2,0.22179839611053467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,2,power_law_1.2,0.23045120239257813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,2,balanced,0.6020319859186808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,2,power_law_1.2,0.2452159881591797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,2,power_law_1.2,0.27523839473724365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,2,power_law_1.2,0.3049855947494507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,2,power_law_1.2,0.3666176080703735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,2,power_law_1.2,0.42727041244506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,2,balanced,0.7961546579996744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,2,power_law_1.2,0.5472447872161865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,2,power_law_1.01,0.04971520006656647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,2,power_law_1.2,0.6769408226013184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,2,power_law_1.2,0.8969535827636719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,2,power_law_1.01,0.05244799852371216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,2,power_law_1.01,0.07667840123176575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,2,power_law_1.2,1.1641663551330566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,2,power_law_1.01,0.06857600212097167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,2,balanced,1.0660053094228108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,2,power_law_1.2,1.614303970336914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,2,power_law_1.01,0.09708799719810486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,2,power_law_1.2,2.25230712890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,2,power_law_1.01,0.1261504054069519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,2,power_law_1.01,0.13966720104217528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,2,power_law_1.2,2.5681856155395506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,2,power_law_1.01,0.17072639465332032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,2,power_law_1.01,0.18096640110015869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,2,power_law_1.2,4.136262512207031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,2,power_law_1.01,0.181004798412323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,2,balanced,1.5055146217346191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,2,power_law_1.2,8.101971435546876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,2,power_law_1.01,0.049491199851036075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,2,power_law_1.01,0.18792959451675414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,2,power_law_1.01,0.052697598934173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,2,power_law_1.01,0.062431997060775755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,2,power_law_1.01,0.199072003364563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,2,balanced,6.49020258585612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,2,power_law_1.01,0.0640447974205017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,2,power_law_1.01,0.20602240562438964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,2,power_law_1.01,0.08417279720306396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,2,power_law_1.01,0.21377921104431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,2,power_law_1.01,0.10505599975585937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,2,power_law_1.01,0.22558081150054932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,2,power_law_1.01,0.115283203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,2,power_law_1.01,0.22895359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,2,power_law_1.01,0.24805760383605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,2,power_law_1.01,0.13546240329742432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,2,balanced,1.9572854042053223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,2,power_law_1.01,0.2789504051208496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,2,power_law_1.01,0.14220800399780273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,2,power_law_1.01,0.14686720371246337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,2,power_law_1.01,0.30576000213623045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,2,power_law_1.01,0.15077120065689087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,2,power_law_1.01,0.36764800548553467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,2,power_law_1.01,0.4427072048187256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,2,power_law_1.01,0.16083199977874757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,2,power_law_1.01,0.5524928092956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,2,power_law_1.01,0.1692031979560852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,2,power_law_1.01,0.6607232093811035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,2,power_law_1.01,0.17626880407333373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,2,power_law_1.01,0.8717439651489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,2,power_law_1.01,0.1867967963218689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,2,power_law_1.01,1.1430399894714356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,2,power_law_1.01,0.191046404838562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,2,power_law_1.01,1.5062784194946288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,2,power_law_1.01,0.20823678970336915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,2,power_law_1.01,1.8708927154541015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,2,power_law_1.01,0.23997440338134765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,2,power_law_1.01,2.333523178100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,2,power_law_1.01,0.27621119022369384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,2,balanced,2.4529172579447427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,2,power_law_1.01,0.3382143974304199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,2,power_law_1.01,3.539654541015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,2,power_law_1.01,0.39278080463409426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,2,power_law_1.01,0.5161664009094238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,2,power_law_1.01,6.853330993652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,2,power_law_1.01,0.6426559925079346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,2,power_law_1.01,0.8551679611206054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,2,power_law_1.01,1.0829631805419921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,2,power_law_1.01,1.5401984214782716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,2,power_law_1.01,2.0520191192626953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,2,power_law_1.01,2.453843116760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,2,power_law_1.01,3.9220672607421876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,2,power_law_1.01,7.513977813720703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,2,balanced,3.7262452443440757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,2,power_law_1.2,0.0527616024017334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,2,power_law_1.2,0.056480002403259275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,2,power_law_1.2,0.0648576021194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,2,power_law_1.2,0.07109760046005249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,2,power_law_1.2,0.0925055980682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,2,power_law_1.2,0.1119488000869751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,2,power_law_1.2,0.12854399681091308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,2,power_law_1.2,0.16168320178985596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,2,power_law_1.2,0.17845120429992675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,2,power_law_1.2,0.04801279902458191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,2,power_law_1.2,0.18376319408416747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,2,power_law_1.2,0.053523200750350955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,2,power_law_1.2,0.1893887996673584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,2,power_law_1.2,0.057548797130584715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,2,power_law_1.2,0.19809919595718384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,2,power_law_1.2,0.06380159854888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,2,power_law_1.2,0.20252161026000975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,2,power_law_1.2,0.07952640056610108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,2,power_law_1.2,0.22053759098052977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,2,power_law_1.2,0.09875839948654175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,2,power_law_1.2,0.22766079902648925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,2,power_law_1.2,0.11213439702987671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,2,power_law_1.2,0.23534080982208253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,2,power_law_1.2,0.13254400491714477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,2,power_law_1.2,0.1422144055366516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,2,power_law_1.2,0.25791358947753906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,2,power_law_1.2,0.14165120124816893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,2,power_law_1.2,0.28996479511260986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,2,power_law_1.2,0.1513152003288269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,2,power_law_1.2,0.3160576105117798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,2,power_law_1.2,0.15865600109100342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,2,power_law_1.2,0.3777791976928711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,2,power_law_1.2,0.16935039758682252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,2,power_law_1.2,0.4387904167175293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,2,power_law_1.2,0.17594239711761475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,2,power_law_1.2,0.5449344158172608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,2,power_law_1.2,0.18619519472122192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,2,power_law_1.2,0.6828864097595215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,2,power_law_1.2,0.19352960586547852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,2,power_law_1.2,0.8728256225585938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,2,power_law_1.2,0.21289598941802979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,2,power_law_1.2,1.1372032165527344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,2,power_law_1.2,0.24540159702301026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,2,power_law_1.2,1.5445759773254395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,2,power_law_1.2,0.27885439395904543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,2,power_law_1.2,1.9520320892333984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,2,power_law_1.2,0.34634239673614503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,2,power_law_1.2,0.39828479290008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,2,power_law_1.2,2.3512191772460938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,2,power_law_1.2,0.52773118019104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,2,power_law_1.2,3.790265655517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,2,power_law_1.2,0.659935998916626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,2,balanced,7.262389500935872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,2,power_law_1.2,0.890015983581543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,2,power_law_1.2,6.960307312011719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,2,power_law_1.2,1.1195136070251466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,2,power_law_1.2,1.5773247718811034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,2,power_law_1.2,2.052396774291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,2,power_law_1.2,2.459129524230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,2,power_law_1.2,4.073600006103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,2,power_law_1.2,8.058009338378906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,64,balanced,0.035536001125971474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,64,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,64,balanced,0.03124266614516576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,64,balanced,0.033413333197434746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,64,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,64,balanced,0.0356480007370313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,64,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,64,balanced,0.03365333378314972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,64,balanced,0.036464000741640724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,64,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,64,balanced,0.033615998923778534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,64,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,64,balanced,0.03150933235883713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,64,balanced,0.037445334096749626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,64,balanced,0.03128000100453695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,64,balanced,0.03608533243338267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,64,balanced,0.033701332906881966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,64,balanced,0.037690666814645134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,64,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,64,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,64,balanced,0.035690667728583016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,64,balanced,0.03974399964014689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,64,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,64,balanced,0.04180799921353658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,64,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,64,balanced,0.03754133234421412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,64,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,64,balanced,0.037045332292715706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,64,balanced,0.04347200194994608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,64,balanced,0.03751999884843826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,64,balanced,0.03738666574160258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,64,balanced,0.046256000796953835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,64,balanced,0.038831998904546104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,64,balanced,0.04351999859015147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,64,balanced,0.05202133456865946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,64,balanced,0.04146666576464971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,64,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,64,balanced,0.05821333328882853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,64,balanced,0.04457066456476847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,64,balanced,0.04765866696834564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,64,balanced,0.06617600222428639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,64,balanced,0.05197333296140035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,64,balanced,0.0639466643333435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,64,balanced,0.06780266761779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,64,balanced,0.08479467034339905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,64,balanced,0.0804906686147054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,64,balanced,0.09841066598892212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,64,balanced,0.12571733196576437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,64,balanced,0.09481066465377808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,64,balanced,0.15307733416557312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,64,balanced,0.1104159951210022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,64,balanced,0.20523732900619507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,64,balanced,0.13806399703025818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,64,balanced,0.2585013310114543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,64,balanced,0.16366933782895407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,64,balanced,0.31522132953008014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,1,balanced,0.06029866635799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,1,balanced,0.0622026671965917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,1,balanced,0.07182399928569794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,1,balanced,0.09619200229644775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,1,balanced,0.14366400241851807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,1,balanced,0.22002132733662924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,1,balanced,0.29817599058151245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,1,balanced,0.2988319993019104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,1,balanced,0.30161599318186444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,1,balanced,0.3060213327407837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,1,balanced,0.30742400884628296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,64,balanced,0.213536004225413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,1,balanced,0.3128160039583842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,1,balanced,0.3142186601956685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,1,balanced,0.31964800755182904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,1,balanced,0.3264906605084737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,1,balanced,0.33291733264923096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,64,balanced,0.45092801253000897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,1,balanced,0.34869333108266193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,1,balanced,0.3943733374277751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,1,balanced,0.4233493407567342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,1,balanced,0.5004053513209025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,1,balanced,0.5608319838841757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,1,balanced,0.7027680079142252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,1,balanced,0.9091839790344238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,1,balanced,1.200229326883952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,1,balanced,1.596362590789795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,1,balanced,2.2681973775227866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,64,balanced,0.3647093375523885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,64,balanced,0.8546773592631022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,1,balanced,3.0262934366861978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,1,balanced,3.7122666041056314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,1,balanced,5.685210545857747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,1,power_law_1.01,0.05190399885177612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,1,power_law_1.01,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,1,power_law_1.01,0.06509439945220948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,1,power_law_1.01,0.08123520016670227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,1,power_law_1.01,0.11743359565734864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,1,power_law_1.01,0.14937599897384643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,1,power_law_1.01,0.17015039920806885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,1,power_law_1.01,0.2108544111251831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,1,power_law_1.01,0.22042241096496581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,1,power_law_1.01,0.22488958835601808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,1,power_law_1.01,0.2374783992767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,1,power_law_1.01,0.2448256015777588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,1,power_law_1.01,0.2575167894363403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,1,power_law_1.01,0.26791040897369384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,1,power_law_1.01,0.27779200077056887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,1,power_law_1.01,0.29280641078948977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,1,power_law_1.01,0.3093631982803345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,1,power_law_1.01,0.36700799465179446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,1,power_law_1.01,0.40252161026000977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,1,power_law_1.01,0.503878402709961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,1,power_law_1.01,0.606060791015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,1,power_law_1.01,0.7864384174346923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,1,power_law_1.01,0.948031997680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,1,power_law_1.01,1.2695296287536622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,1,balanced,11.094069163004557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,1,power_law_1.01,1.594048023223877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,1,power_law_1.2,0.05507199764251709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,1,power_law_1.01,2.2747776031494142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,1,power_law_1.2,0.05797119736671448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,1,power_law_1.01,2.9636159896850587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,1,power_law_1.2,0.06934400200843811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,1,balanced,0.060693333546320595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,1,power_law_1.01,3.645555114746094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,1,power_law_1.2,0.08388479948043823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,1,balanced,0.06406933565934499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,1,power_law_1.2,0.10885119438171387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,1,balanced,0.0692799985408783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,1,power_law_1.01,5.743795013427734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,1,balanced,0.08668266733487447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,1,power_law_1.2,0.140774405002594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,1,balanced,0.12867732842763266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,1,power_law_1.2,0.15934720039367675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,1,balanced,0.19162132342656454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,1,balanced,0.25376532475153607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,1,power_law_1.2,0.19774080514907838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,1,balanced,0.2542453408241272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,1,power_law_1.01,10.951046752929688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,1,power_law_1.2,0.21278080940246583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,1,balanced,0.25835732618967694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,1,balanced,0.25887999931971234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,1,power_law_1.2,0.22257280349731445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,1,balanced,0.2641599973042806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,1,power_law_1.2,0.22872319221496581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,1,balanced,0.2725226680437724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,1,power_law_1.2,0.24474880695343018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,1,balanced,0.2750506599744161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,1,power_law_1.2,0.2475584030151367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,1,balanced,0.281546672185262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,1,power_law_1.2,0.26806399822235105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,1,balanced,0.29072000583012897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,1,power_law_1.2,0.2806848049163818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,1,balanced,0.29952534039815265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,1,power_law_1.2,0.2853503942489624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,1,balanced,0.3174186746279399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,1,power_law_1.2,0.30652799606323244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,1,balanced,0.37112001578013104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,1,power_law_1.01,0.050783997774124144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,1,power_law_1.2,0.3776128053665161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,1,power_law_1.2,0.4150847911834717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,1,balanced,0.40861864884694415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,1,power_law_1.01,0.05602560043334961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,1,power_law_1.2,0.5140607833862305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,1,power_law_1.01,0.06174719929695129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,1,balanced,0.493008017539978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,1,power_law_1.2,0.6173632144927979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,1,power_law_1.01,0.07399680018424988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,1,balanced,0.5578186511993408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,1,power_law_1.01,0.10776319503784179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,1,power_law_1.2,0.7930111885070801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,1,power_law_1.01,0.13252480030059816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,1,power_law_1.2,0.9621120452880859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,1,balanced,0.7136800289154053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,1,power_law_1.01,0.14648319482803346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,1,power_law_1.2,1.2836352348327638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,1,power_law_1.01,0.1775936007499695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,1,power_law_1.01,0.18659199476242067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,1,power_law_1.2,1.6147327423095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,1,balanced,0.9273973306020101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,1,power_law_1.01,0.19344639778137207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,1,power_law_1.2,2.2934656143188477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,1,power_law_1.01,0.20448639392852783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,1,power_law_1.2,2.9717504501342775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,1,power_law_1.01,0.2173759937286377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,1,power_law_1.01,0.2315903902053833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,1,balanced,1.239519993464152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,1,power_law_1.2,3.658528137207031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,1,power_law_1.01,0.25099520683288573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,1,power_law_1.01,0.27111039161682127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,1,power_law_1.2,5.805184173583984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,1,power_law_1.01,0.28995840549468993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,1,power_law_1.01,0.32103040218353274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,1,power_law_1.01,0.4034304141998291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,1,power_law_1.2,10.990656280517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,1,balanced,1.7000373204549153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,1,power_law_1.01,0.39017601013183595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,1,power_law_1.01,0.49582719802856445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,1,power_law_1.01,0.5777408123016358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,1,power_law_1.01,0.7736512184143066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,1,power_law_1.01,0.971008014678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,64,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,64,balanced,0.023530667026837666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,1,power_law_1.01,1.325267219543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,1,balanced,2.4275946617126465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,64,balanced,0.03586133321126302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,1,power_law_1.01,1.6790143966674804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,64,balanced,0.03549333413441976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,64,balanced,0.01940800001223882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,64,balanced,0.03369600077470144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,64,balanced,0.022831998765468597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,1,power_law_1.01,2.3987968444824217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,64,balanced,0.03777066618204117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,64,balanced,0.021712000171343487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,64,balanced,0.037130666275819145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,64,balanced,0.021514666577180225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,1,power_law_1.01,3.098579216003418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,64,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,64,balanced,0.025536000728607178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,64,balanced,0.03741333385308584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,1,power_law_1.01,3.8295871734619142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,64,balanced,0.02532800038655599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,64,balanced,0.037418665985266365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,64,balanced,0.03965333352486292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,64,balanced,0.037818667789300285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,64,balanced,0.03945599993069967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,1,power_law_1.01,6.116966247558594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,64,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,64,balanced,0.039162665605545044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,64,balanced,0.03958933303753535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,1,balanced,3.1496267318725586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,64,balanced,0.03879466652870178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,64,balanced,0.03952533255020777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,64,balanced,0.036389333506425224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,1,power_law_1.01,11.884690856933593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,64,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,64,balanced,0.037274666130542755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,64,balanced,0.043893332282702126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,64,balanced,0.03554133325815201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,64,balanced,0.043712000052134194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,64,balanced,0.04277333120505015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,64,balanced,0.046015997727712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,64,balanced,0.044922664761543274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,64,balanced,0.03821333249409994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,64,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,64,balanced,0.04378133515516917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,64,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,64,balanced,0.05243200063705444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,64,balanced,0.05049600203831991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,64,balanced,0.06273599962393443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,64,balanced,0.05598933498064677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,64,balanced,0.07664533456166585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,1,balanced,3.955557187398275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,64,balanced,0.06583466629187266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,64,balanced,0.10097600022951762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,64,balanced,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,1,power_law_1.2,0.054502397775650024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,64,balanced,0.1257866621017456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,1,power_law_1.2,0.05927680134773254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,1,power_law_1.2,0.06641280055046081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,1,power_law_1.2,0.0781440019607544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,64,balanced,0.09364266196886699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,1,power_law_1.2,0.10323200225830079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,64,balanced,0.17126933733622232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,1,power_law_1.2,0.12611199617385865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,1,power_law_1.2,0.13834880590438842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,1,power_law_1.2,0.17493120431900025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,1,power_law_1.2,0.18668160438537598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,1,power_law_1.2,0.1932096004486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,64,balanced,0.1071519951025645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,1,power_law_1.2,0.20395519733428955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,64,balanced,0.21893866856892905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,1,power_law_1.2,0.22221438884735106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,1,power_law_1.2,0.22946560382843018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,1,power_law_1.2,0.25137920379638673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,1,power_law_1.2,0.27044479846954345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,1,power_law_1.2,0.2908416032791138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,1,balanced,6.081786473592122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,1,power_law_1.2,0.3242815971374512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,1,power_law_1.2,0.4092864036560059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,1,power_law_1.2,0.40026240348815917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,64,balanced,0.1350879967212677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,1,power_law_1.2,0.5055808067321778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,64,balanced,0.3133653402328491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,1,power_law_1.2,0.5905151844024659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,1,power_law_1.2,0.7812352180480957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,1,power_law_1.2,0.9951871871948242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,1,power_law_1.2,1.3483519554138184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,1,power_law_1.2,1.6875072479248048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,1,power_law_1.2,2.405401611328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,1,power_law_1.2,3.101945686340332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,64,balanced,0.1711519956588745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,1,power_law_1.2,3.813663864135742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,64,power_law_1.01,0.0384768009185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,64,balanced,0.4085226853688558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,1,power_law_1.2,6.142169570922851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,64,power_law_1.01,0.042284798622131345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,64,power_law_1.01,0.030943998694419862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,64,power_law_1.01,0.034944000840187076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,64,power_law_1.01,0.0342848002910614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,1,power_law_1.2,11.962489318847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,64,power_law_1.01,0.035385599732398985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,64,power_law_1.01,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,64,power_law_1.01,0.03544960021972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,64,power_law_1.01,0.035622400045394895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,64,power_law_1.01,0.03575679957866669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,64,power_law_1.01,0.03761279881000519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,64,power_law_1.01,0.04359039962291718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,64,power_law_1.01,0.03315199911594391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,64,power_law_1.01,0.04471679925918579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,64,power_law_1.01,0.04767360091209412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,64,balanced,0.2035199999809265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,64,power_law_1.01,0.03057279884815216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,64,power_law_1.01,0.05379840135574341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,64,power_law_1.01,0.029760000109672547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,64,power_law_1.01,0.05882880091667175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,64,balanced,0.5034826596577963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,64,power_law_1.01,0.032927998900413515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,64,power_law_1.01,0.05656319856643677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,64,power_law_1.01,0.03261440098285675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,64,power_law_1.01,0.06433280110359192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,64,power_law_1.01,0.03370879888534546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,64,power_law_1.01,0.06846079826354981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,64,power_law_1.01,0.034201601147651674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,64,power_law_1.01,0.08133119940757752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,64,power_law_1.01,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,64,power_law_1.01,0.09519360065460206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,64,power_law_1.01,0.034995201230049136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,64,power_law_1.01,0.13493759632110597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,64,power_law_1.01,0.035846400260925296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,64,power_law_1.01,0.1389631986618042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,64,power_law_1.01,0.038726401329040525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,64,power_law_1.01,0.179257595539093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,64,power_law_1.01,0.03868800103664398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,64,power_law_1.01,0.2257472038269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,64,power_law_1.01,0.04161919951438904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,64,power_law_1.01,0.2954495906829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,64,power_law_1.01,0.04160000085830688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,64,power_law_1.01,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,64,power_law_1.01,0.3684544086456299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,64,power_law_1.01,0.05050240159034729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,1,balanced,11.833562215169271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,64,power_law_1.01,0.43167362213134763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,64,power_law_1.01,0.047167998552322385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,64,power_law_1.01,0.053939199447631835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,64,power_law_1.01,0.6693888187408448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,64,power_law_1.01,0.0652671992778778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,64,power_law_1.01,0.08355839848518372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,64,power_law_1.01,1.448153591156006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,64,power_law_1.01,0.09059839844703674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,64,power_law_1.01,0.12231040000915527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,64,power_law_1.01,0.14559359550476075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,64,balanced,0.27987200021743774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,64,power_law_1.01,0.21025280952453612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,64,power_law_1.01,0.2553855895996094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,64,balanced,0.7899893124898275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,64,power_law_1.01,0.37832961082458494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,64,power_law_1.01,0.44343042373657227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,64,power_law_1.01,0.5961215972900391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,64,power_law_1.01,0.8623231887817383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,64,power_law_1.01,1.8628799438476562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,64,power_law_1.01,0.0438400000333786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,64,power_law_1.01,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,64,power_law_1.01,0.04082559943199158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,64,power_law_1.01,0.04395520091056824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,64,power_law_1.01,0.043942400813102724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,64,power_law_1.01,0.04573439955711365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,64,power_law_1.01,0.046060800552368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,64,power_law_1.01,0.04655359983444214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,64,power_law_1.01,0.04565120041370392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,64,power_law_1.01,0.04511359930038452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,64,power_law_1.01,0.046137601137161255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,64,power_law_1.01,0.04565759897232056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,64,power_law_1.01,0.04647040069103241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,64,power_law_1.01,0.04619520008563995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,64,power_law_1.01,0.05012480020523071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,64,power_law_1.01,0.05262719988822937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,64,power_law_1.01,0.05616000294685364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,64,power_law_1.01,0.06170880198478699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.01,0.06789759993553161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.01,0.07634559869766236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.01,0.08714240193367004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,64,power_law_1.01,0.021190400421619415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.01,0.11340160369873047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.01,0.1337407946586609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,64,power_law_1.01,0.01881600022315979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,64,power_law_1.01,0.020768000185489653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.01,0.18403199911117554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,64,power_law_1.01,0.020428800582885744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.01,0.22382719516754152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.01,0.32238719463348386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,64,power_law_1.01,0.020659199357032774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.01,0.4340223789215088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,64,power_law_1.01,0.02563199996948242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,64,power_law_1.01,0.024377599358558655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,64,power_law_1.01,0.4895936012268066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,64,power_law_1.01,0.037862399220466615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,64,balanced,0.5108853181203207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,64,power_law_1.01,0.8725184440612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,64,power_law_1.01,0.03803519904613495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,64,power_law_1.01,0.037747201323509214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,64,power_law_1.01,0.03771519958972931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,64,power_law_1.01,1.6281343460083009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,64,balanced,1.5476800600687664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,64,power_law_1.01,0.03845120072364807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,64,power_law_1.01,0.0364544004201889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,64,power_law_1.01,0.03304960131645203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,64,power_law_1.01,0.034272000193595886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,64,power_law_1.01,0.03697279989719391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,64,power_law_1.01,0.040089601278305055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,64,power_law_1.01,0.042559999227523806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.01,0.047302401065826415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.01,0.0580735981464386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.01,0.06520959734916687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.01,0.0919871985912323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.01,0.11050879955291748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.01,0.14999680519104003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.01,0.2006848096847534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.01,0.2801408052444458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.01,0.35571839809417727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,64,power_law_1.01,0.4318399906158447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,64,power_law_1.01,0.6670591831207275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,64,power_law_1.01,1.38603515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,64,power_law_1.2,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,64,power_law_1.2,0.039801600575447085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,64,power_law_1.2,0.03175680041313171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,64,power_law_1.2,0.03454720079898834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,64,power_law_1.2,0.035257598757743834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,64,power_law_1.2,0.0353408008813858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,64,power_law_1.2,0.035545599460601804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,64,power_law_1.2,0.035897600650787356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,64,power_law_1.2,0.03627519905567169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,64,power_law_1.2,0.03697279989719391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,64,power_law_1.2,0.039750400185585025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,64,power_law_1.2,0.04220159947872162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,64,power_law_1.2,0.04668799936771393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,64,power_law_1.2,0.05139200091361999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,64,power_law_1.2,0.0566976010799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,64,power_law_1.2,0.06094719767570496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,64,power_law_1.2,0.06506879925727845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,64,power_law_1.2,0.0728767991065979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,64,power_law_1.2,0.0756608009338379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,64,power_law_1.2,0.0927295982837677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,64,power_law_1.2,0.12245759963989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,64,power_law_1.2,0.15927679538726808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,64,power_law_1.2,0.18997119665145873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,64,power_law_1.2,0.24599039554595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,64,power_law_1.2,0.3335616111755371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,64,power_law_1.2,0.4870016098022461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,64,power_law_1.2,0.7402815818786621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,64,power_law_1.2,0.8493120193481445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,64,power_law_1.2,0.03400320112705231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,64,power_law_1.2,1.3092927932739258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,64,power_law_1.2,0.03190400004386902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,64,power_law_1.2,2.934118461608887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,64,power_law_1.2,0.030905601382255555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,64,power_law_1.2,0.04316799938678741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,64,power_law_1.2,0.03426559865474701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,64,power_law_1.2,0.033542400598526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,64,power_law_1.2,0.040479999780654904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,64,power_law_1.2,0.03503359854221344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,64,power_law_1.2,0.04209280014038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,64,power_law_1.2,0.03524479866027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,64,power_law_1.2,0.03558399975299835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,64,power_law_1.2,0.043808001279830935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,64,power_law_1.2,0.03596799969673157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,64,power_law_1.2,0.04384639859199524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,64,power_law_1.2,0.03768320083618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,64,power_law_1.2,0.0458624005317688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,64,power_law_1.2,0.03968639969825745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,64,power_law_1.2,0.045747199654579164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,64,power_law_1.2,0.04028800129890442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,64,power_law_1.2,0.04609279930591583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,64,power_law_1.2,0.0430976003408432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,64,power_law_1.2,0.04421760141849518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,64,power_law_1.2,0.04584319889545441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,64,power_law_1.2,0.04805119931697845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,64,power_law_1.2,0.021663999557495116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,64,power_law_1.2,0.05028480291366577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,64,power_law_1.2,0.04481280148029328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,64,power_law_1.2,0.05300480127334595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,64,power_law_1.2,0.04645119905471802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,64,power_law_1.2,0.01907840073108673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,64,power_law_1.2,0.059987199306488034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,64,power_law_1.2,0.04567039906978607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,64,power_law_1.2,0.07207040190696716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,64,power_law_1.2,0.0208064004778862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,64,power_law_1.2,0.04613119959831238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,64,power_law_1.2,0.09529600143432618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,64,power_law_1.2,0.04873600006103516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,64,power_law_1.2,0.020640000700950623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,64,power_law_1.2,0.10750720500946045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,64,power_law_1.2,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,64,power_law_1.2,0.020838400721549986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,64,power_law_1.2,0.15303679704666137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,64,power_law_1.2,0.05432320237159729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,64,power_law_1.2,0.025971201062202454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,64,power_law_1.2,0.206496000289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,64,power_law_1.2,0.02399359941482544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,64,power_law_1.2,0.057868802547454835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,64,power_law_1.2,0.03763839900493622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,64,power_law_1.2,0.0627776026725769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,64,power_law_1.2,0.29216001033782957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,64,power_law_1.2,0.03768959939479828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.2,0.06604160070419311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,64,power_law_1.2,0.3576256036758423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,64,power_law_1.2,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.2,0.08884479999542236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,64,power_law_1.2,0.5654911994934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,64,power_law_1.2,0.0379584014415741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.2,0.09678080081939697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,64,power_law_1.2,0.03680639863014221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,64,power_law_1.2,0.6436992168426514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.2,0.12826240062713623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,64,power_law_1.2,0.036339199542999266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.2,0.1707584023475647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,64,power_law_1.2,0.8169919967651367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,64,power_law_1.2,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.2,0.2300800085067749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,64,power_law_1.2,0.03591679930686951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,64,power_law_1.2,1.5513279914855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.2,0.3021696090698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,64,power_law_1.2,0.03651840090751648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.2,0.5211135864257812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,64,power_law_1.2,0.0383296012878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,64,power_law_1.2,3.175859260559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,64,power_law_1.2,0.04737280011177063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.2,0.6774911880493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.2,0.05093119740486145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,64,power_law_1.2,0.8494463920593261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.2,0.058950400352478026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,64,power_law_1.2,1.6031808853149414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.2,0.07125759720802308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.2,0.0936959981918335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.2,0.1165503978729248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,64,power_law_1.2,3.163167953491211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.2,0.15671039819717408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.2,0.20955519676208495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.2,0.30188798904418945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.2,0.38791680335998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,64,power_law_1.2,0.4961088180541992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,64,power_law_1.2,0.7797247886657714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,64,power_law_1.2,1.5438528060913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,32,balanced,0.03532800078392029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,32,balanced,0.035589332381884255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,32,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,32,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,32,balanced,0.03734400123357773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,32,balanced,0.037605332831541695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,32,balanced,0.03935466706752777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,32,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,32,balanced,0.03761066744724909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,32,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,32,balanced,0.03939199944337209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,32,balanced,0.04041066765785217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,32,balanced,0.039642666776975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,32,balanced,0.018986667195955913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,32,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,32,balanced,0.02145066608985265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,32,balanced,0.04560000201066335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,32,balanced,0.01931200052301089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,32,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,32,balanced,0.020351999749739964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,32,balanced,0.04587733248869578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,32,balanced,0.019487999379634857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,32,balanced,0.0207893339296182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,32,balanced,0.04798933366934458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,32,balanced,0.023445333043734234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,32,balanced,0.023525332411130268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,32,balanced,0.050000001986821495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,32,balanced,0.029504001140594482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,32,balanced,0.05384533107280731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,32,balanced,0.028101332485675812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,32,balanced,0.03842666745185852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,32,balanced,0.056090667843818665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,32,balanced,0.03774400055408478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,32,balanced,0.03555200000603994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,32,balanced,0.03372266640265783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,32,balanced,0.07046933472156525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,32,balanced,0.04554666578769684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,32,balanced,0.044954667488733925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,32,balanced,0.04011200120051702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,32,balanced,0.08037333190441132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,32,balanced,0.04470933477083842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,32,balanced,0.05715733269850413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,32,balanced,0.06809600194295247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,32,balanced,0.10161599516868591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,32,balanced,0.08248533308506012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,32,balanced,0.10547733306884766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,32,balanced,0.12168000141779582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,32,balanced,0.1353546679019928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,32,balanced,0.18210667371749878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,32,balanced,0.1550933321317037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,32,balanced,0.23433067401250204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,32,balanced,0.19798400004704794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,32,balanced,0.33355732758839923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,32,balanced,0.43570133050282794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,32,balanced,0.23467733462651572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,32,balanced,0.5408000151316324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,32,power_law_1.01,0.03487359881401062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,32,balanced,0.03549333413441976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,32,power_law_1.01,0.03377920091152191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,32,balanced,0.03333866596221924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,32,power_law_1.01,0.032915198802948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,32,balanced,0.029232000311215717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,32,power_law_1.01,0.03511680066585541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,32,balanced,0.03297599901755651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,32,power_law_1.01,0.03527039885520935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,32,balanced,0.03549333413441976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,32,power_law_1.01,0.03700479865074158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,32,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,32,balanced,0.03156800071398417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,32,power_law_1.01,0.03738240003585815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,32,balanced,0.03536533315976461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,32,power_law_1.01,0.037376001477241516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,32,balanced,0.03579733272393545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,32,balanced,0.03327466547489166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,32,balanced,0.3329919974009196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,32,power_law_1.01,0.03743999898433685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,32,balanced,0.03533333291610082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,32,balanced,0.03175999969244003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,32,power_law_1.01,0.03797760009765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,32,balanced,0.03723733375469843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,32,balanced,0.03327466547489166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,32,power_law_1.01,0.03864319920539856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,32,balanced,0.037151999771595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,32,power_law_1.01,0.018329599499702455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,32,balanced,0.03536533315976461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,32,power_law_1.01,0.039129599928855896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,32,balanced,0.03642133375008901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,32,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,32,power_law_1.01,0.040582400560379026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,32,balanced,0.0377866675456365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,32,power_law_1.01,0.01870719939470291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,32,balanced,0.037477334340413414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,32,power_law_1.01,0.042483198642730716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,32,balanced,0.038736000657081604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,32,power_law_1.01,0.018624000251293182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,32,balanced,0.035429333647092186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,32,power_law_1.01,0.04607360064983368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,32,balanced,0.04058133314053217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,32,power_law_1.01,0.01857919991016388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,32,balanced,0.0356480007370313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,32,power_law_1.01,0.04796159863471985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,32,power_law_1.01,0.021408000588417055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,32,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,32,balanced,0.037402667105197906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,32,power_law_1.01,0.05017600059509277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,32,power_law_1.01,0.02573440074920654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,32,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,32,balanced,0.0418453315893809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,32,power_law_1.01,0.055961602926254274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,32,power_law_1.01,0.040761598944664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,32,balanced,0.03770666569471359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.01,0.05863680243492127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,32,balanced,0.04383466641108195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,32,power_law_1.01,0.040243199467658995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,32,balanced,0.03766933331886927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.01,0.06760960221290588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,32,power_law_1.01,0.037868800759315493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,32,balanced,0.039488000174363456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,32,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,32,power_law_1.01,0.03903999924659729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.01,0.07794560194015503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,32,balanced,0.04359466830889384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,32,power_law_1.01,0.038176000118255615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.01,0.09752960205078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,32,balanced,0.041797334949175514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,32,power_law_1.01,0.03939839899539947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,32,balanced,0.05593599875768026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.01,0.11445759534835816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,32,balanced,0.04553600152333578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,32,balanced,0.8580426375071207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,32,power_law_1.01,0.03857280015945434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.01,0.17447680234909058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,32,balanced,0.04606399933497111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,32,power_law_1.01,0.03668479919433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,32,balanced,0.06161599854628245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.01,0.20650880336761473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,32,power_law_1.01,0.03612160086631775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,32,balanced,0.05043200155099233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,32,power_law_1.01,0.03700479865074158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.01,0.24757759571075438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,32,power_law_1.01,0.03868159949779511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,32,balanced,0.05435200035572052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.01,0.39426560401916505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,32,balanced,0.06781333188215892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,32,power_law_1.01,0.04190079867839813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,32,power_law_1.01,0.44493441581726073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.01,0.04447360038757324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,32,balanced,0.06711466610431671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.01,0.05920000076293945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,32,power_law_1.01,0.7374335765838623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.01,0.06662399768829345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,32,balanced,0.07306666672229767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.01,0.08569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,32,balanced,0.08846400181452434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,32,power_law_1.01,1.2340352058410644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.01,0.10263680219650269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.01,0.13371520042419432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.01,0.1717120051383972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,32,balanced,0.1053706705570221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.01,0.23086080551147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,32,balanced,0.0885653297106425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.01,0.31033599376678467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,32,power_law_1.01,0.3753983974456787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,32,power_law_1.01,0.5928959846496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,32,balanced,0.1360586682955424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,32,power_law_1.01,1.1634495735168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,32,balanced,0.10837866862614949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,32,balanced,0.1709280014038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,32,balanced,0.6122293472290039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,32,balanced,0.12920000155766806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,32,balanced,0.2264159917831421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,32,balanced,0.1600053310394287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,32,balanced,0.2829119960467021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,32,balanced,1.6871253649393718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,32,power_law_1.01,0.03754239976406097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,32,power_law_1.01,0.037862399220466615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,32,power_law_1.01,0.031999999284744264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,32,power_law_1.01,0.03284479975700379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,32,power_law_1.01,0.03277440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,32,power_law_1.01,0.03535360097885132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,32,power_law_1.01,0.03543039858341217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,32,power_law_1.01,0.035180801153182985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,32,balanced,0.18976000944773355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,32,power_law_1.01,0.03511680066585541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,32,power_law_1.01,0.035846400260925296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,32,balanced,0.34330666065216064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,32,power_law_1.01,0.03932160139083862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,32,power_law_1.01,0.042105600237846375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,32,power_law_1.01,0.041075199842453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,32,power_law_1.01,0.04627839922904968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,32,power_law_1.01,0.050892800092697144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,32,power_law_1.01,0.0546239972114563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,32,power_law_1.01,0.054073601961135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,32,power_law_1.01,0.06117759943008423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,32,power_law_1.01,0.06484479904174804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,32,power_law_1.01,0.0762880027294159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,32,power_law_1.01,0.09392639994621277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,32,power_law_1.01,0.10350719690322877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,32,power_law_1.01,0.11881599426269532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,32,power_law_1.01,0.17943040132522584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,32,power_law_1.01,0.2209791898727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,32,power_law_1.01,0.24355840682983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,32,power_law_1.01,0.36266241073608396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,32,power_law_1.01,0.4520895957946777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,32,balanced,0.2511626680692037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,32,power_law_1.01,0.726796817779541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,32,balanced,0.5055359999338785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,32,power_law_1.01,1.3951744079589843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,32,power_law_1.01,0.03933440148830414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,32,power_law_1.01,0.03573760092258453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,32,power_law_1.01,0.031699201464653014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,32,power_law_1.01,0.032543998956680295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,32,power_law_1.01,0.03342719972133636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,32,power_law_1.01,0.03463039994239807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,32,power_law_1.01,0.03459199965000152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,32,power_law_1.01,0.035846400260925296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,32,power_law_1.01,0.03620480000972748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,32,power_law_1.01,0.03656319975852966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,32,power_law_1.01,0.038073599338531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,32,power_law_1.01,0.04056319892406464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,32,power_law_1.01,0.04141440093517303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,32,power_law_1.01,0.04279040098190308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,32,power_law_1.01,0.046374401450157164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,32,power_law_1.01,0.0494271993637085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,32,power_law_1.01,0.05129600167274475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,32,power_law_1.01,0.05678079724311828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,32,power_law_1.01,0.06169599890708923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,32,power_law_1.01,0.07923200130462646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,32,power_law_1.01,0.0938431978225708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,32,power_law_1.01,0.12916480302810668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,32,power_law_1.01,0.15175679922103882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,32,power_law_1.01,0.22181758880615235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,32,power_law_1.01,0.25816318988800047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,32,power_law_1.01,0.40108799934387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,32,power_law_1.01,0.49884800910949706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,32,power_law_1.01,0.5594304084777832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,32,power_law_1.01,0.9362815856933594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,32,power_law_1.01,1.8129663467407227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,32,balanced,0.44870932896931964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,32,balanced,0.9649120171864828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,32,power_law_1.2,0.04529280066490173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,32,power_law_1.2,0.0366784006357193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,32,power_law_1.2,0.030028799176216127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,32,power_law_1.2,0.033036801218986514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,32,power_law_1.2,0.03249920010566711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,32,power_law_1.2,0.034534400701522826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,32,power_law_1.2,0.03505280017852783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,32,power_law_1.2,0.03498240113258362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,32,power_law_1.2,0.03547520041465759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,32,power_law_1.2,0.03635840117931366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,32,power_law_1.2,0.03940480053424835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,32,power_law_1.2,0.04012799859046936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,32,power_law_1.2,0.043705600500106814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,32,power_law_1.2,0.046323201060295104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,32,power_law_1.2,0.05242879986763001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,32,power_law_1.2,0.03157120048999786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,32,power_law_1.2,0.053376001119613645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,32,power_law_1.2,0.055769598484039305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,32,power_law_1.2,0.03038719892501831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,32,power_law_1.2,0.07155839800834655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,32,power_law_1.2,0.02991360127925873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,32,power_law_1.2,0.06844800114631652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,32,power_law_1.2,0.030643200874328612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,32,power_law_1.2,0.08501759767532349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,32,power_law_1.2,0.03150720000267029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,32,power_law_1.2,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,32,power_law_1.2,0.10111360549926758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,32,power_law_1.2,0.03390080034732819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,32,power_law_1.2,0.13076479434967042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,32,power_law_1.2,0.03497599959373474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,32,power_law_1.2,0.03472639918327332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,32,power_law_1.2,0.13768320083618163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,32,power_law_1.2,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,32,power_law_1.2,0.23393280506134034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,32,power_law_1.2,0.03294720053672791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,32,power_law_1.2,0.03646079897880554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,32,power_law_1.2,0.2358975887298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,32,power_law_1.2,0.03828479945659637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,32,power_law_1.2,0.032716798782348636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,32,power_law_1.2,0.0392767995595932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,32,power_law_1.2,0.35713920593261717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,32,power_law_1.2,0.034822401404380796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,32,power_law_1.2,0.04123519957065582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,32,power_law_1.2,0.03514240086078644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,32,power_law_1.2,0.5794047832489013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,32,power_law_1.2,0.042233601212501526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,32,power_law_1.2,0.03682560026645661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,32,power_law_1.2,0.6886528015136719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,32,power_law_1.2,0.03701759874820709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,32,power_law_1.2,0.04708479940891266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,32,power_law_1.2,0.04922879934310913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,32,power_law_1.2,0.03687680065631867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,32,power_law_1.2,0.018540799617767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,32,power_law_1.2,1.0297792434692383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,32,power_law_1.2,0.052537602186203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,32,power_law_1.2,0.037625598907470706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,32,power_law_1.2,0.0587007999420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,32,power_law_1.2,0.01850239932537079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,32,power_law_1.2,1.8800960540771485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,32,power_law_1.2,0.038073599338531494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,32,power_law_1.2,0.06446719765663148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,32,power_law_1.2,0.018822400271892546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,32,power_law_1.2,0.03866240084171295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,32,power_law_1.2,0.07602559924125671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,32,power_law_1.2,0.11479040384292602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,32,power_law_1.2,0.03984639942646027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,32,power_law_1.2,0.018675200641155243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,32,power_law_1.2,0.139302396774292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,32,power_law_1.2,0.04133760035037994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,32,power_law_1.2,0.020703999698162077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,32,power_law_1.2,0.17642879486083984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,32,power_law_1.2,0.04335359930992126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,32,power_law_1.2,0.02600319981575012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,32,power_law_1.2,0.24764800071716309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,32,power_law_1.2,0.046489599347114566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,32,power_law_1.2,0.04048640131950378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,32,power_law_1.2,0.04878720045089722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,32,power_law_1.2,0.31797759532928466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,32,power_law_1.2,0.05060480237007141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,32,power_law_1.2,0.040479999780654904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,32,power_law_1.2,0.42451839447021483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,32,power_law_1.2,0.059385597705841064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,32,power_law_1.2,0.038252800703048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.2,0.06551679968833923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,32,power_law_1.2,0.670252799987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,32,power_law_1.2,0.03873279988765717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.2,0.07447680234909057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,32,power_law_1.2,0.037939199805259706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.2,0.08068479895591736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,32,power_law_1.2,1.039027214050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,32,power_law_1.2,0.038764798641204835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.2,0.10795520544052124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,32,power_law_1.2,0.037567999958992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,32,power_law_1.2,1.2100799560546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.2,0.1378368020057678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,32,power_law_1.2,0.03553279936313629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.2,0.19761279821395875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,32,power_law_1.2,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,32,power_law_1.2,3.0751232147216796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.2,0.2538048028945923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,32,power_law_1.2,0.035872000455856326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.2,0.3967871904373169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,32,power_law_1.2,0.03898879885673523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.2,0.6374783992767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,32,power_law_1.2,0.04059520065784454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.2,0.04791040122509003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,32,power_law_1.2,0.6039807796478271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.2,0.06100479960441589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,32,power_law_1.2,1.009984016418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.2,0.06810879707336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.2,0.0902079999446869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,32,power_law_1.2,2.491859245300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.2,0.10899840593338013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.2,0.15083520412445067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.2,0.1915711998939514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.2,0.2797696113586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.2,0.35474560260772703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,32,power_law_1.2,0.43605761528015136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,32,power_law_1.2,0.7039040088653564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,32,power_law_1.2,1.408409595489502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,16,balanced,0.021295999487241108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,16,balanced,0.021146667500336964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,16,balanced,0.019237333287795384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,16,balanced,0.020288000504175823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,16,balanced,0.03743999948104223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,16,balanced,0.021253332495689392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,16,balanced,0.021354667842388153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,16,balanced,0.03869866579771042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,16,balanced,0.023397333920001984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,16,balanced,0.03765333443880081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,16,balanced,0.023120000958442688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,16,balanced,0.03734933336575826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,16,balanced,0.02518400053183238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,16,balanced,0.03974399964014689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,16,balanced,0.023317334552605946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,16,balanced,0.0414986660083135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,16,balanced,0.027461332579453785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,16,balanced,0.042170668641726174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,16,balanced,0.0301706666747729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,16,balanced,0.04154133299986521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,16,balanced,0.02916266769170761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,16,balanced,0.041493333876132965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,16,balanced,0.040149333576361336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,16,balanced,0.046053335070610046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,16,balanced,0.04351999859015147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,16,balanced,0.042581334710121155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,16,balanced,0.04604266583919525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,16,balanced,0.04385599990685781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,16,balanced,0.044218664367993675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,16,balanced,0.04354666670163473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,16,balanced,0.04394666850566864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,16,balanced,0.04795733094215393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,16,balanced,0.04863466819127401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,16,balanced,0.05486399928728739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,16,balanced,0.047877331574757896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,16,balanced,0.05004799862702688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,16,balanced,0.07158933579921722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,16,balanced,0.05227200190226237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,16,balanced,0.056101332108179726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,16,balanced,0.08017600079377492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,16,balanced,0.06021333237489065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,16,balanced,0.10285866260528564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,16,balanced,0.06654400130112965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,16,balanced,0.08277333279450734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,16,balanced,0.11748266220092773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,16,balanced,0.09681066870689392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,16,balanced,0.16844266653060913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,16,balanced,0.12729066610336304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,16,power_law_1.01,0.04971520006656647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,16,power_law_1.01,0.03806079924106598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,16,power_law_1.01,0.040345600247383116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,16,power_law_1.01,0.04440320134162903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,16,balanced,0.20681599775950113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,16,power_law_1.01,0.043219199776649474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,16,power_law_1.01,0.04717440009117126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,16,power_law_1.01,0.048089599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,16,balanced,0.15742400288581848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,16,power_law_1.01,0.047679999470710756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,16,power_law_1.01,0.04769920110702515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,16,power_law_1.01,0.04935039877891541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,16,power_law_1.01,0.02022400051355362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,16,power_law_1.01,0.04854399859905243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,16,power_law_1.01,0.05100160241127014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,16,power_law_1.01,0.018675200641155243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,16,power_law_1.01,0.050271999835968015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,16,power_law_1.01,0.01860480010509491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,16,power_law_1.01,0.05127679705619812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,16,power_law_1.01,0.01825920045375824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,16,power_law_1.01,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,16,balanced,0.2940746744473775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,16,power_law_1.01,0.055366402864456175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,16,power_law_1.01,0.019046400487422944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,16,power_law_1.01,0.059359997510910034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,16,power_law_1.01,0.01976960003376007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,16,balanced,0.20605866114298502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,16,power_law_1.01,0.06739839911460876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,16,power_law_1.01,0.02069759964942932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.01,0.07047680020332336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,16,power_law_1.01,0.02030719965696335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.01,0.08213760256767273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,16,power_law_1.01,0.020960000157356263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.01,0.09835519790649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,16,power_law_1.01,0.0215488001704216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.01,0.13159040212631226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,16,power_law_1.01,0.025279998779296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,16,power_law_1.01,0.025171199440956117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.01,0.14737919569015503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,16,power_law_1.01,0.025856000185012818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.01,0.2020479917526245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,16,balanced,0.03350399931271871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,16,power_law_1.01,0.041222399473190306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.01,0.25489919185638427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,16,power_law_1.01,0.041119998693466185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.01,0.32648959159851076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,16,balanced,0.03428266694148382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,16,power_law_1.01,0.03941119909286499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,16,balanced,0.03161066770553589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.01,0.45637760162353513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,16,balanced,0.033258666594823204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,16,power_law_1.01,0.04319359958171844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,16,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,16,balanced,0.3821599880854289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,16,power_law_1.01,0.508076810836792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,16,power_law_1.01,0.04225279986858368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,16,balanced,0.25421865781148273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,16,balanced,0.0354666660229365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.01,0.047193598747253415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,16,balanced,0.03573866685231527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,16,power_law_1.01,0.8830143928527832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.01,0.05427839756011963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,16,balanced,0.03677333394686381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,16,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.01,0.061536002159118655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,16,power_law_1.01,1.761894416809082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,16,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.01,0.08124160170555114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,16,balanced,0.0377866675456365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.01,0.09192320108413696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,16,balanced,0.037717332442601524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,16,balanced,0.039493332306543984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.01,0.12319999933242798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,16,balanced,0.039749334255854286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.01,0.14599679708480834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,16,balanced,0.04189866781234741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,16,balanced,0.04364266494909922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.01,0.21345279216766358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,16,balanced,0.04384533564249674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.01,0.2726272106170654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,16,balanced,0.04755199948946635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,16,power_law_1.01,0.36440320014953614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,16,balanced,0.050661335388819374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,16,power_law_1.01,0.531001615524292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,16,balanced,0.06242666641871134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,16,balanced,0.3058240016301473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,16,balanced,0.4697493314743042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,16,power_law_1.01,1.0566271781921386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,16,balanced,0.06810666620731354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,16,balanced,0.07803733150164287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,16,balanced,0.08521067102750142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,16,balanced,0.10734933614730835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,16,balanced,0.13193066914876303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,16,balanced,0.033402666449546814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,16,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,16,balanced,0.45609064896901447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,16,balanced,0.033173332611719765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,16,balanced,0.0332640012105306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,16,balanced,0.7486026287078857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,16,balanced,0.03603733330965042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,16,balanced,0.03742400060097376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,16,balanced,0.03833599885304769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,16,balanced,0.037621334195137024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,16,balanced,0.1627893348534902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,16,balanced,0.037685332198937736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,16,balanced,0.03973866750796636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,16,balanced,0.039359999199708305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,16,balanced,0.03984000037113825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,16,balanced,0.03951466580231985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,16,balanced,0.041850666205088295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,16,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,16,balanced,0.04383466641108195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,16,balanced,0.045968001087506614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,16,balanced,0.049653331438700356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,16,balanced,0.0558186670144399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,16,balanced,0.20339200894037882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,16,balanced,0.05978666742642721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,16,balanced,0.07438399891058604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,16,balanced,0.09900800387064616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,16,balanced,0.11758400003115337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,16,balanced,0.24651734034220377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,16,balanced,0.15381866693496704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,16,balanced,0.197434663772583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,16,balanced,0.86899201075236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,16,power_law_1.01,0.03996799886226654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,16,power_law_1.01,0.031148800253868104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,16,balanced,0.2652906576792399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,16,power_law_1.01,0.03256320059299469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,16,balanced,1.4714345932006836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,16,power_law_1.01,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,16,power_law_1.01,0.03404799997806549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,16,power_law_1.01,0.03480319976806641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,16,power_law_1.01,0.03616639971733093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,16,power_law_1.01,0.03716480135917664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,16,power_law_1.01,0.035872000455856326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,16,balanced,0.34459733963012695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,16,power_law_1.01,0.03736959993839264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,16,power_law_1.01,0.03831680119037628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,16,power_law_1.01,0.03958399891853333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,16,power_law_1.01,0.04087679982185364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,16,power_law_1.01,0.04545280039310455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,16,power_law_1.01,0.052051198482513425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,16,balanced,0.33460267384847003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,16,power_law_1.01,0.053504002094268796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,16,power_law_1.01,0.05399680137634277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,16,power_law_1.01,0.06253439784049988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,16,power_law_1.01,0.07006080150604248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,16,power_law_1.01,0.07838079929351807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,16,power_law_1.01,0.034835198521614076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,16,power_law_1.01,0.08992000222206116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,16,power_law_1.01,0.11854720115661621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,16,power_law_1.01,0.03447679877281189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,16,power_law_1.01,0.12231680154800414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,16,power_law_1.01,0.030950400233268737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,16,power_law_1.01,0.03136000037193298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,16,power_law_1.01,0.16915199756622315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,16,power_law_1.01,0.03255679905414581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,16,power_law_1.01,0.19398399591445922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,16,power_law_1.01,0.03546879887580871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,16,power_law_1.01,0.288319993019104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,16,power_law_1.01,0.035820800065994265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,16,power_law_1.01,0.3532864093780518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,16,power_law_1.01,0.03697920143604279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,16,balanced,0.4097013473510742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,16,power_law_1.01,0.44268159866333007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,16,power_law_1.01,0.03711999952793121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,16,power_law_1.01,0.6853888034820557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,16,power_law_1.01,0.03761920034885406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,16,power_law_1.01,0.03869439959526062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,16,power_law_1.01,1.1484607696533202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,16,power_law_1.01,0.04037759900093078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,16,power_law_1.01,0.04149760007858276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,16,power_law_1.01,0.04355199933052063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,16,power_law_1.01,0.04814079999923706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,16,power_law_1.01,0.052147197723388675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,16,power_law_1.01,0.051583999395370485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,16,power_law_1.01,0.05731199979782105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,16,power_law_1.01,0.06510720252990723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,16,power_law_1.01,0.08177279829978942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,16,power_law_1.01,0.09733120203018189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,16,power_law_1.01,0.12407040596008301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,16,power_law_1.2,0.03544319868087768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,16,power_law_1.01,0.1530303955078125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,16,power_law_1.2,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,16,power_law_1.01,0.21571199893951415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,16,balanced,0.6159306764602661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,16,power_law_1.2,0.030297601222991945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,16,power_law_1.01,0.25479679107666015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,16,balanced,0.6347413460413615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,16,power_law_1.2,0.032204800844192506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,16,power_law_1.01,0.4081855773925781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,16,power_law_1.2,0.03356159925460815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,16,power_law_1.01,0.5539455890655518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,16,power_law_1.2,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,16,power_law_1.2,0.03619840145111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,16,power_law_1.01,0.6314239978790284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,16,power_law_1.2,0.036697599291801455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,16,power_law_1.2,0.03682560026645661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,16,power_law_1.01,0.9252863883972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,16,power_law_1.2,0.03018240034580231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,16,power_law_1.2,0.03811840116977692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,16,power_law_1.2,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,16,power_law_1.2,0.029311999678611755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,16,power_law_1.2,0.0387584000825882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,16,power_law_1.01,1.9698623657226562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,16,power_law_1.2,0.040915200114250184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,16,power_law_1.2,0.030559998750686646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,16,power_law_1.2,0.04286719858646393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,16,power_law_1.2,0.03235200047492981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,16,power_law_1.2,0.04730879962444305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,16,power_law_1.2,0.03468160033226013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,16,power_law_1.2,0.05322880148887634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,16,power_law_1.2,0.03578239977359772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,16,power_law_1.2,0.054713600873947145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,16,power_law_1.2,0.03594880104064942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,16,power_law_1.2,0.05883520245552063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,16,power_law_1.2,0.0662015974521637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,16,power_law_1.2,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,16,power_law_1.2,0.06935039758682252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,16,power_law_1.2,0.036697599291801455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,16,power_law_1.2,0.08273919820785522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,16,power_law_1.2,0.04022400081157684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,16,power_law_1.2,0.04267520010471344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,16,power_law_1.2,0.0984063982963562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,16,power_law_1.2,0.044409599900245664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,16,power_law_1.2,0.12267520427703857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,16,power_law_1.2,0.044761601090431216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,16,power_law_1.2,0.14199039936065674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,16,power_law_1.2,0.04997119903564453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,16,power_law_1.2,0.23660800457000733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,16,power_law_1.2,0.05145599842071533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,16,power_law_1.2,0.275980806350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,16,power_law_1.2,0.05232639908790589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,16,power_law_1.2,0.37575039863586424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,16,power_law_1.2,0.06074879765510559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,16,power_law_1.2,0.6273087978363037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,16,power_law_1.2,0.07303040027618408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,16,power_law_1.2,0.09317759871482849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,16,power_law_1.2,0.6137728214263916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,16,power_law_1.2,0.10380799770355224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,16,power_law_1.2,1.0650303840637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,16,power_law_1.2,0.1520959973335266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,16,power_law_1.2,0.17459839582443237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,16,power_law_1.2,0.23277440071105956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,16,power_law_1.2,2.582534408569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,16,power_law_1.2,0.3463680028915405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,16,power_law_1.2,0.49447040557861327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,16,power_law_1.2,0.7343999862670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,16,power_law_1.2,0.8968704223632813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,16,power_law_1.2,1.3916799545288085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,16,balanced,1.176911989847819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,16,power_law_1.2,3.1679935455322266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,16,power_law_1.2,0.049491199851036075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,16,power_law_1.2,0.037401598691940305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,16,power_law_1.2,0.040633600950241086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,16,power_law_1.2,0.043987199664115906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,16,power_law_1.2,0.045772799849510194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,16,power_law_1.2,0.04885759949684143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,16,power_law_1.2,0.048819199204444885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,16,power_law_1.2,0.04973439872264862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,16,power_law_1.2,0.04902400076389313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,16,power_law_1.2,0.04880639910697937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,16,power_law_1.2,0.04927999973297119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,16,power_law_1.2,0.05055999755859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,16,power_law_1.2,0.050995200872421265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,16,power_law_1.2,0.0516543984413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,16,power_law_1.2,0.05623679757118225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,16,power_law_1.2,0.06096640229225159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,16,power_law_1.2,0.02019200026988983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,16,power_law_1.2,0.0623744010925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,16,power_law_1.2,0.06899200081825256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,16,power_law_1.2,0.019142399728298187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.2,0.07701759934425353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.2,0.08577920198440551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,16,power_law_1.2,0.018534399569034576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.2,0.10954240560531617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,16,power_law_1.2,0.018751999735832213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.2,0.13048959970474244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,16,power_law_1.2,0.01913599967956543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.2,0.1532863974571228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,16,power_law_1.2,0.01993599981069565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.2,0.22631680965423584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,16,power_law_1.2,0.02024960070848465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.2,0.3277440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,16,power_law_1.2,0.02072319984436035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.2,0.46404480934143066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,16,power_law_1.2,0.021132799983024596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.2,0.6901696205139161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,16,power_law_1.2,0.021830399334430695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,16,power_law_1.2,0.02497279942035675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,16,power_law_1.2,0.8505087852478027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,16,power_law_1.2,0.025216001272201537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,16,power_law_1.2,0.026073598861694337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,16,power_law_1.2,1.3008447647094727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,16,power_law_1.2,0.04082559943199158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,16,power_law_1.2,0.03952000141143799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,16,power_law_1.2,2.738323211669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,16,power_law_1.2,0.040012800693511964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,16,power_law_1.2,0.04097279906272888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,16,power_law_1.2,0.043724799156188966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.2,0.046086400747299194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.2,0.055289602279663085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.2,0.06293119788169861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.2,0.08277119994163513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.2,0.10227839946746826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.2,0.1385472059249878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.2,0.16640640497207643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.2,0.24925439357757567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.2,0.30224640369415284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,16,power_law_1.2,0.3994240045547485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,16,power_law_1.2,0.5728127956390381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,16,power_law_1.2,1.251091194152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,8,balanced,0.029258665939172108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,8,balanced,0.03133866687615713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,8,balanced,0.03164266546567281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,8,balanced,0.031583999594052635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,8,balanced,0.03738133360942205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,8,balanced,0.039781334499518074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,8,balanced,0.04155199974775314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,8,balanced,0.03942399968703588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,8,balanced,0.03945599993069967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,8,balanced,0.03945599993069967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,8,balanced,0.03968533376852671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,8,balanced,0.03976533313592275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,8,balanced,0.039887999494870506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,8,balanced,0.04160533348719279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,8,balanced,0.04585599899291992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,8,balanced,0.044821331898371376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,8,balanced,0.049125333627065025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,8,balanced,0.05366399884223938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,8,balanced,0.06794133285681407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,8,balanced,0.07523733377456665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,8,balanced,0.07972266773382823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,8,balanced,0.09696533282597859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,8,balanced,0.11705066760381062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,8,balanced,0.1456160048643748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,8,balanced,0.031258667508761086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,8,balanced,0.17669866482416788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,8,power_law_1.01,0.03819519877433777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,8,balanced,0.03159466634194056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,8,balanced,0.03357866654793421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,8,power_law_1.01,0.0316864013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,8,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,8,power_law_1.01,0.030348798632621764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,8,balanced,0.03979199876387914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,8,power_law_1.01,0.032076799869537355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,8,balanced,0.040949332217375435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,8,power_law_1.01,0.034534400701522826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,8,balanced,0.04016000032424927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,8,power_law_1.01,0.03771519958972931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,8,balanced,0.04176533222198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,8,power_law_1.01,0.03850240111351013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,8,balanced,0.041562666495641075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,8,power_law_1.01,0.040703999996185306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,8,balanced,0.04185600082079569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,8,power_law_1.01,0.03943679928779602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,8,balanced,0.04159999887148539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,8,power_law_1.01,0.0394239991903305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,8,balanced,0.23695999383926392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,8,balanced,0.04248533149560293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,8,power_law_1.01,0.041094401478767396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,8,balanced,0.044639999667803444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,8,power_law_1.01,0.04206719994544983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,8,balanced,0.04389866689840952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,8,power_law_1.01,0.04371840059757233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,8,balanced,0.048895999789237976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,8,power_law_1.01,0.04639999866485596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,8,balanced,0.04970133304595947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,8,power_law_1.01,0.053235197067260744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,8,power_law_1.01,0.05617280006408691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,8,balanced,0.0517546683549881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,8,power_law_1.01,0.05793920159339905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,8,balanced,0.05412800113360087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,8,power_law_1.01,0.06446719765663148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,8,power_law_1.01,0.08282880187034607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,8,balanced,0.06396799782911937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,8,power_law_1.01,0.09671040177345276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,8,power_law_1.01,0.10497920513153076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,8,balanced,0.07014933228492737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,8,power_law_1.01,0.13308800458908082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,8,power_law_1.01,0.14914560317993164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,8,power_law_1.01,0.22703359127044678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,8,balanced,0.08475733796755473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,8,balanced,0.297818660736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,8,power_law_1.01,0.24625279903411865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,8,power_law_1.01,0.38217599391937257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,8,balanced,0.11730133493741353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,8,power_law_1.01,0.48061442375183105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,8,power_law_1.2,0.03813759982585907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,8,power_law_1.01,0.5248767852783203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,8,power_law_1.2,0.03047040104866028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,8,power_law_1.01,0.8723263740539551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,8,power_law_1.2,0.03054719865322113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,8,balanced,0.14281599720319113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,8,power_law_1.2,0.031193599104881287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,8,power_law_1.01,1.9785472869873046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,8,power_law_1.2,0.03374080061912536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,8,power_law_1.2,0.0374208003282547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,8,power_law_1.2,0.03857280015945434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,8,power_law_1.2,0.03937920033931732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,8,balanced,0.19055465857187906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,8,power_law_1.2,0.03978880047798157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,8,power_law_1.2,0.04051840007305145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,8,power_law_1.01,0.035411199927330016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,8,power_law_1.2,0.040992000699043275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,8,balanced,0.3633439938227336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,8,power_law_1.2,0.04222719967365265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,8,power_law_1.01,0.033888000249862674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,8,power_law_1.2,0.04627200067043304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,8,power_law_1.01,0.03214080035686493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,8,power_law_1.2,0.04780159890651703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,8,balanced,0.24362132946650186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,8,power_law_1.01,0.03378559947013855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,8,power_law_1.2,0.05408639907836914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,8,power_law_1.01,0.03615359961986542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,8,power_law_1.2,0.05896959900856018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,8,power_law_1.01,0.038553598523139956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,8,power_law_1.2,0.061926400661468504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,8,power_law_1.2,0.06777600049972535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,8,power_law_1.01,0.039417600631713866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,8,power_law_1.01,0.03991039991378784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,8,power_law_1.2,0.08476160168647766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,8,power_law_1.01,0.040550398826599124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,8,power_law_1.2,0.10115200281143188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,8,power_law_1.01,0.04246399998664856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,8,power_law_1.2,0.10952320098876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,8,power_law_1.01,0.04344959855079651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,8,power_law_1.2,0.1327679991722107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,8,power_law_1.01,0.04664320051670075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,8,balanced,0.34942932923634845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,8,power_law_1.2,0.1820736050605774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,8,power_law_1.01,0.04850560128688812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,8,power_law_1.2,0.2554431915283203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,8,power_law_1.01,0.04639999866485596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,8,power_law_1.2,0.03496319949626923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,8,power_law_1.2,0.29689600467681887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,8,power_law_1.01,0.051046401262283325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,8,power_law_1.01,0.05309439897537231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,8,power_law_1.2,0.031353598833084105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,8,power_law_1.2,0.3924288034439087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,8,power_law_1.01,0.05589759945869446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,8,power_law_1.2,0.030380800366401672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,8,power_law_1.2,0.5659071922302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,8,power_law_1.2,0.03239679932594299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,8,power_law_1.01,0.06500480175018311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,8,power_law_1.2,0.03330560028553009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,8,power_law_1.2,0.7520639896392822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,8,power_law_1.01,0.07618560194969178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,8,power_law_1.2,0.03684479892253876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,8,power_law_1.01,0.10567679405212402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,8,power_law_1.2,0.03826560080051422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,8,power_law_1.2,1.2245759963989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,8,power_law_1.01,0.116211199760437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,8,power_law_1.2,0.0387584000825882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,8,power_law_1.2,0.03911679983139038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,8,power_law_1.01,0.1524608016014099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,8,power_law_1.2,0.040064001083374025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,8,power_law_1.2,2.446553611755371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,8,power_law_1.01,0.17918720245361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,8,power_law_1.2,0.04259839951992035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,8,balanced,0.5455679893493652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,8,balanced,0.44185598691304523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,8,power_law_1.01,0.25691521167755127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,8,power_law_1.2,0.04357759952545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,8,power_law_1.01,0.3566528081893921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,8,power_law_1.2,0.04746240079402923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,8,power_law_1.01,0.46581759452819826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,8,power_law_1.2,0.04860160052776337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,8,power_law_1.2,0.053625601530075076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,8,power_law_1.01,0.6263360023498535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,8,power_law_1.2,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,8,power_law_1.2,0.05648639798164368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,8,power_law_1.01,0.8337216377258301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,8,power_law_1.2,0.06487039923667907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,8,power_law_1.01,1.153331184387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,8,power_law_1.2,0.07971839904785157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,8,power_law_1.2,0.10461440086364746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,8,power_law_1.2,0.11268479824066162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,8,power_law_1.01,2.5033344268798827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,8,power_law_1.2,0.16596479415893556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,8,power_law_1.2,0.19591679573059081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,8,power_law_1.2,0.283187198638916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,8,power_law_1.2,0.42565121650695803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,8,power_law_1.2,0.5737792015075683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,8,balanced,0.5407199859619141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,8,power_law_1.2,0.8616127967834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,8,power_law_1.2,0.9041791915893554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,8,power_law_1.2,1.5920512199401855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,8,power_law_1.2,3.2116416931152343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,4,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,4,balanced,0.033557333052158356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,4,balanced,0.03358400116364161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,4,balanced,0.03595199932654699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,4,balanced,0.044405331214269005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,4,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,4,balanced,0.04358399907747904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,4,balanced,0.04367466767628988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,8,balanced,0.8334720134735107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,4,balanced,0.04373333354791006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,4,balanced,0.04357333481311798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,4,balanced,0.04554666578769684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,4,balanced,0.04588800172011057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,4,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,4,balanced,0.04779199759165446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,4,balanced,0.0517439991235733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,4,balanced,0.056405335664749146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,8,balanced,1.01036270459493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,4,balanced,0.05784533421198527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,4,balanced,0.06427733103434245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,4,balanced,0.0748533308506012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,4,balanced,0.0923520028591156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,4,balanced,0.10356799761454265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,4,balanced,0.13012799620628357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,4,balanced,0.029264000554879505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,4,balanced,0.16269866625467935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,4,balanced,0.031290667752424874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,4,balanced,0.033589333295822144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,4,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,4,balanced,0.04151466737190882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,4,balanced,0.04570133487383524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,4,balanced,0.04394133388996124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,4,balanced,0.22022400299708048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,4,balanced,0.04562666515509287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,4,balanced,0.0439573327700297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,4,balanced,0.044906665881474815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,4,balanced,0.046309332052866616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,4,balanced,0.04769066472848257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,4,balanced,0.05005866785844167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,4,balanced,0.04940799872080485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,4,balanced,0.27089067300160724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,4,balanced,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,4,balanced,0.057962665955225624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,4,balanced,0.059903999169667564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,4,balanced,0.06638399759928386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,4,balanced,0.07060266534487407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,4,balanced,0.09959466258684795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,4,balanced,0.11528533697128296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,4,balanced,0.39101866881052655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,8,balanced,1.606661319732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,4,balanced,0.15793066223462424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,4,balanced,0.18859734137852988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,4,balanced,0.271071990331014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,4,balanced,0.49423468112945557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,4,power_law_1.01,0.03387520015239716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,4,power_law_1.01,0.0289792001247406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,4,balanced,0.34534935156504315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,4,power_law_1.01,0.029862400889396668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,4,power_law_1.01,0.03441919982433319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,4,power_law_1.01,0.03619199991226196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,4,power_law_1.01,0.04113920032978058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,4,power_law_1.01,0.0435263991355896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,4,power_law_1.01,0.04431360065937042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,4,power_law_1.01,0.04482559859752655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,4,power_law_1.01,0.045337599515914914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,4,balanced,0.6097013155619303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,4,balanced,0.5028479894002279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,4,power_law_1.01,0.04696959853172302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,4,power_law_1.01,0.048825600743293764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,4,power_law_1.01,0.049292799830436704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,4,power_law_1.01,0.054201602935791016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,4,power_law_1.01,0.05953279733657837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,4,power_law_1.01,0.06561279892921448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,4,power_law_1.01,0.06856319904327393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,4,power_law_1.01,0.07978879809379577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,4,power_law_1.01,0.08995199799537659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,4,power_law_1.01,0.1078336000442505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,4,power_law_1.01,0.13276159763336182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,4,power_law_1.01,0.15628160238265992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,4,balanced,0.6424586772918701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,4,power_law_1.01,0.031033599376678468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,4,power_law_1.01,0.20775039196014405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,4,power_law_1.01,0.2774208068847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,4,power_law_1.01,0.029791998863220214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,4,power_law_1.01,0.3418560028076172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,4,power_law_1.01,0.03139840066432953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,4,power_law_1.01,0.5168000221252441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,4,power_law_1.01,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,4,power_law_1.01,0.6823296070098877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,4,power_law_1.01,0.03635199964046478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,4,power_law_1.01,0.835308837890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,4,power_law_1.2,0.03619199991226196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,4,balanced,0.9276479880015055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,4,power_law_1.01,0.04111360013484955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,4,power_law_1.01,1.3369152069091796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,4,power_law_1.01,0.04216960072517395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,4,power_law_1.2,0.030380800366401672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,4,power_law_1.01,0.04323199987411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,4,power_law_1.2,0.03118079900741577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,4,power_law_1.01,2.552960014343262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,4,balanced,0.8086026509602865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,4,power_law_1.01,0.04466559886932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,4,power_law_1.2,0.03403519988059998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,4,power_law_1.01,0.04580479860305786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,4,power_law_1.2,0.037510401010513304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,4,power_law_1.01,0.04810880124568939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,4,power_law_1.2,0.04262399971485138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,4,power_law_1.2,0.04382080137729645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,4,power_law_1.01,0.05143679976463318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,4,power_law_1.2,0.02961919903755188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,4,power_law_1.2,0.04639999866485596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,4,power_law_1.01,0.05440000295639038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,4,power_law_1.2,0.04656000137329101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,4,power_law_1.2,0.0290367990732193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,4,power_law_1.01,0.05280640125274658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,4,power_law_1.2,0.04716159999370575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,4,power_law_1.2,0.030687999725341798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,4,power_law_1.01,0.05911679863929749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,4,power_law_1.2,0.04794879853725433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,4,power_law_1.01,0.06245120167732239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,4,power_law_1.2,0.049516800045967105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,4,power_law_1.2,0.03456639945507049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,4,power_law_1.2,0.052742397785186766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,4,power_law_1.01,0.06729599833488464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,4,power_law_1.2,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,4,power_law_1.2,0.055379199981689456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,4,power_law_1.2,0.041382399201393125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,4,power_law_1.01,0.0772607982158661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,4,power_law_1.2,0.06444799900054932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,4,power_law_1.2,0.04141440093517303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,4,power_law_1.01,0.09247360229492188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,4,power_law_1.2,0.04384639859199524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,4,power_law_1.01,0.11416319608688355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,4,power_law_1.2,0.06749439835548401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,4,power_law_1.2,0.04444159865379334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,4,power_law_1.2,0.07134720087051391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,4,power_law_1.01,0.13763840198516847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,4,power_law_1.2,0.04520959854125976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,4,power_law_1.2,0.07964159846305847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,4,power_law_1.01,0.18019200563430787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,4,power_law_1.2,0.04836480021476745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,4,power_law_1.2,0.09608960151672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,4,power_law_1.2,0.050886398553848265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,4,power_law_1.2,0.11360000371932984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,4,power_law_1.01,0.23442559242248534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,4,power_law_1.2,0.0548799991607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,4,power_law_1.2,0.13193600177764891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,4,power_law_1.01,0.31335039138793946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,4,power_law_1.2,0.1732800006866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,4,power_law_1.2,0.053260797262191774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,4,balanced,1.262063980102539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,4,power_law_1.2,0.23669118881225587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,4,power_law_1.2,0.059385597705841064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,4,power_law_1.01,0.4596223831176758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,4,power_law_1.2,0.33879680633544923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,4,power_law_1.2,0.0617792010307312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,4,power_law_1.01,0.6399424076080322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,4,power_law_1.2,0.07185279726982116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,4,power_law_1.2,0.3879551887512207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,4,power_law_1.01,0.8752448081970214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,4,power_law_1.2,0.08190079927444457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,4,power_law_1.2,0.5524479866027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,4,power_law_1.2,0.09260159730911255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,4,power_law_1.01,0.9657407760620117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,4,power_law_1.2,0.7486591815948487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,4,power_law_1.2,0.12789119482040406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,4,power_law_1.2,0.14741120338439942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,4,power_law_1.01,1.509727954864502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,4,power_law_1.2,0.9254528045654297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,4,power_law_1.2,0.1904255986213684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,4,power_law_1.2,1.3550016403198242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,4,power_law_1.2,0.23922560214996338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,4,power_law_1.01,2.958559989929199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,4,power_law_1.2,0.339468789100647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,4,power_law_1.2,2.943827247619629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,4,power_law_1.2,0.4734079837799072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,4,power_law_1.2,0.7048255920410156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,4,power_law_1.2,0.9140543937683105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,4,balanced,1.7617120742797852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,4,power_law_1.2,1.1409536361694337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,4,power_law_1.2,1.7468864440917968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,4,power_law_1.2,3.8217342376708983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,4,balanced,2.4060160319010415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,2,balanced,0.033546666304270424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,2,balanced,0.03530666728814443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,2,balanced,0.03665599972009659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,2,balanced,0.041008000572522484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,2,balanced,0.04781333108743032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,2,balanced,0.05003733436266581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,2,balanced,0.050474668542544045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,2,balanced,0.052058666944503784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,2,balanced,0.051957334081331887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,2,balanced,0.05195199946562449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,2,balanced,0.05402666827042898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,2,balanced,0.056458666920661926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,2,balanced,0.057861333092053734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,2,balanced,0.0602453351020813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,2,balanced,0.06439466774463654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,2,balanced,0.07062399884064992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,2,balanced,0.07779733339945476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,2,balanced,0.08945600191752116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,2,balanced,0.10320533315340678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,2,balanced,0.11993066469828288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,2,balanced,0.14630400141080221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,2,balanced,0.22343466679255167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,2,balanced,0.24543466170628866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,2,balanced,0.353274663289388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,2,balanced,0.45576000213623047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,2,balanced,0.6639093160629272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,2,balanced,0.8569493293762207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,2,balanced,0.03346133232116699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,2,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,2,balanced,0.037621334195137024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,2,balanced,0.04287466903527578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,2,balanced,0.05317866802215576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,2,balanced,0.057855998476346336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,2,balanced,0.05716800192991892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,2,balanced,0.05649599929650625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,2,balanced,0.058874666690826416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,2,balanced,0.058975999553998314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,2,balanced,0.05780800183614095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,2,balanced,0.060421332716941833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,2,balanced,0.060592000683148704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,2,balanced,0.06635733445485432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,2,balanced,0.07142933209737141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,2,balanced,0.07278400162855785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,2,balanced,1.0407360394795735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,2,balanced,0.07665066421031952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,2,balanced,0.09498666723569234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,2,balanced,0.11316266655921936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,2,power_law_1.01,0.03420799970626831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,2,balanced,0.13777066270510355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,2,power_law_1.01,0.031091201305389404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,2,power_law_1.01,0.033983999490737916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,2,power_law_1.01,0.03703039884567261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,2,balanced,0.16083733240763345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,2,power_law_1.01,0.040300801396369934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,2,power_law_1.01,0.04836480021476745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,2,power_law_1.01,0.05247359871864319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,2,balanced,0.23530133565266928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,2,power_law_1.01,0.05251839756965637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,2,power_law_1.01,0.05255680084228516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,2,power_law_1.01,0.053548800945281985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,2,power_law_1.01,0.058950400352478026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,2,balanced,0.29613866408665973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,2,power_law_1.01,0.06107519865036011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,2,power_law_1.01,0.06222720146179199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,2,power_law_1.01,0.0667967975139618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,2,power_law_1.01,0.0769536018371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,32,balanced,0.05996266504128774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,2,power_law_1.01,0.08108159899711609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,2,power_law_1.01,0.08529919981956482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,2,balanced,0.4280159870783488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,32,balanced,0.04799999793370565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,2,power_law_1.01,0.0989184021949768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,32,balanced,0.048010667165120445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,32,balanced,0.05172266562779745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,2,power_law_1.01,0.11239039897918701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,32,balanced,0.07820799946784973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,2,balanced,0.5423946777979533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,2,balanced,1.6137439409891765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,32,balanced,0.045738667249679565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,32,balanced,0.08165866633256276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,2,power_law_1.01,0.13688960075378417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,32,balanced,0.084389328956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,32,balanced,0.04580266773700714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,2,power_law_1.01,0.1818943977355957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,32,balanced,0.0846453309059143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,32,balanced,0.04438933233420054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,32,balanced,0.08370133241017659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,2,power_law_1.01,0.23840000629425048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,32,balanced,0.04589866598447164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,32,balanced,0.08473599950472514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,32,balanced,0.05074666440486908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,2,power_law_1.01,0.2857664108276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,32,balanced,0.08611733714739482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,32,balanced,0.06292800108591716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,2,power_law_1.01,0.408403205871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,32,balanced,0.06404266754786174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,32,balanced,0.08543466528256734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,32,balanced,0.06471999982992808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,32,balanced,0.08684266606966655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,2,power_law_1.01,0.5192831993103028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,32,balanced,0.06612266600131989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,32,balanced,0.08798399567604065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,2,power_law_1.01,0.7319615840911865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,32,balanced,0.06809600194295247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,32,balanced,0.09145599603652954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,2,balanced,0.801146666208903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,32,balanced,0.06822933256626129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,2,power_law_1.01,0.9511232376098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,32,balanced,0.06810133159160614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,32,balanced,0.09126399954160054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,2,power_law_1.01,1.211359977722168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,32,balanced,0.07130666573842366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,32,balanced,0.09470400214195251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,32,balanced,0.07402133444945018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,32,balanced,0.1021066705385844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,2,power_law_1.01,1.7308544158935546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,32,balanced,0.0783786674340566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,32,balanced,0.08179733157157898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,32,balanced,0.10114133358001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,2,power_law_1.01,3.43201904296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,32,balanced,0.08287466565767924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,32,balanced,0.1195199986298879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,32,balanced,0.09678399562835693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,32,balanced,0.09717333316802979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,32,balanced,0.11656000216801961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,32,balanced,0.11563199758529663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,32,balanced,0.1497119963169098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,2,balanced,1.0525493621826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,32,balanced,0.12801067034403482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,32,balanced,0.16275200247764587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,2,power_law_1.01,0.032262399792671204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,32,balanced,0.15472533305486044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,2,power_law_1.01,0.03222399950027466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,2,power_law_1.2,0.03383040130138397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,2,power_law_1.01,0.03493120074272156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,32,balanced,0.18926932414372763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,2,power_law_1.01,0.03759360015392303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,2,power_law_1.2,0.0313728004693985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,2,power_law_1.01,0.04213759899139404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,2,power_law_1.2,0.03338240087032318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,2,power_law_1.01,0.050835198163986205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,32,balanced,0.2241333325703939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,2,power_law_1.01,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,2,power_law_1.2,0.03650560081005096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,2,power_law_1.01,0.05427200198173523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,32,balanced,0.2625173330307007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,2,power_law_1.2,0.039129599928855896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,2,power_law_1.01,0.05626879930496216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,2,power_law_1.01,0.05731199979782105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,2,power_law_1.2,0.049472001194953916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,2,power_law_1.01,0.06098560094833374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,2,power_law_1.2,0.050732797384262084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,2,balanced,1.301909367243449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,2,power_law_1.01,0.06372479796409607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,2,power_law_1.2,0.05308160185813904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,2,power_law_1.01,0.06725119948387145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,32,balanced,0.2352799971898397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,2,power_law_1.2,0.053446400165557864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,2,power_law_1.01,0.06711040139198303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,2,power_law_1.2,0.0531328022480011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,32,balanced,0.3036853273709615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,2,power_law_1.01,0.07496320009231568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,2,power_law_1.2,0.05786240100860596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,2,power_law_1.01,0.08064640164375306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,2,power_law_1.2,0.06251519918441772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,2,power_law_1.01,0.08606079816818238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,2,power_law_1.2,0.06389759778976441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,2,power_law_1.01,0.10620800256729127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,2,power_law_1.2,0.06938239932060242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,2,power_law_1.01,0.12669440507888793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,2,power_law_1.2,0.077920001745224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,2,power_law_1.01,0.15497599840164183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,2,power_law_1.2,0.08083840012550354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,2,power_law_1.2,0.08544639945030212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,2,power_law_1.01,0.18895360231399536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,2,power_law_1.2,0.10069760084152221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,2,balanced,3.1394240061442056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,2,power_law_1.01,0.2553600072860718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,32,balanced,0.32050132751464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,2,power_law_1.2,0.11806080341339112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,32,balanced,0.4527466694513957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,2,power_law_1.01,0.31889920234680175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,2,power_law_1.2,0.14284160137176513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,2,power_law_1.01,0.4576576232910156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,2,power_law_1.2,0.18814079761505126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,2,power_law_1.2,0.033011201024055484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,2,power_law_1.2,0.242905592918396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,2,power_law_1.01,0.5828927993774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,2,power_law_1.2,0.29358079433441164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,2,power_law_1.2,0.0333759993314743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,2,power_law_1.01,0.8937151908874512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,2,power_law_1.2,0.39468159675598147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,2,power_law_1.2,0.03583999872207642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,2,power_law_1.01,1.1601792335510255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,2,power_law_1.2,0.5097216129302978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,2,power_law_1.2,0.037894400954246524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,2,power_law_1.01,1.3707839965820312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,2,power_law_1.2,0.04048640131950378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,2,power_law_1.2,0.7828095912933349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,2,power_law_1.2,0.05103359818458557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,2,power_law_1.2,1.013491153717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,2,power_law_1.01,2.133407974243164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,2,power_law_1.2,0.052934402227401735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,2,balanced,2.011722723642985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,2,power_law_1.2,1.2883584022521972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,2,power_law_1.2,0.0555840015411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,32,balanced,0.3826666673024495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,32,balanced,0.5368266503016154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,2,power_law_1.2,0.05767679810523987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,2,power_law_1.01,4.320844650268555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,2,power_law_1.2,1.898931121826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,2,power_law_1.2,0.05974400043487549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,2,power_law_1.2,3.698566436767578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,2,power_law_1.2,0.06257280111312866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,2,power_law_1.2,0.06547840237617493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,2,power_law_1.2,0.06752640008926392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,2,power_law_1.2,0.06835200190544129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,2,power_law_1.2,0.07601919770240784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,2,power_law_1.2,0.08458880186080933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,2,power_law_1.2,0.09111679792404175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,2,power_law_1.2,0.10851839780807496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,2,power_law_1.2,0.12651519775390624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,32,balanced,0.4469333489735921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,32,balanced,0.6409386793772379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,2,power_law_1.2,0.15800960063934327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,2,power_law_1.2,0.19864319562911986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,2,power_law_1.2,0.2698496103286743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,2,power_law_1.2,0.33274240493774415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,2,power_law_1.2,0.4498432159423828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,2,power_law_1.2,0.6185215950012207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,2,power_law_1.2,0.8855104446411133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,2,power_law_1.2,1.1922368049621581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,2,power_law_1.2,1.628927993774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,2,power_law_1.2,2.279225540161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,2,power_law_1.2,4.687628936767578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,32,balanced,0.9787147045135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,32,balanced,0.6493013302485148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,2,balanced,3.960298538208008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,32,balanced,1.8976853688557942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,32,balanced,1.2145013014475505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,32,balanced,0.04887466629346212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,32,balanced,0.049813335140546165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,32,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,32,balanced,0.051445335149765015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,32,balanced,0.053871999184290566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,32,balanced,0.055946667989095054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,32,balanced,0.057664001981417336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,32,balanced,0.057962665955225624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,32,balanced,0.05682133138179779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,32,balanced,0.05793066819508871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,32,balanced,0.058005332946777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,32,balanced,0.05795733133951823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,32,balanced,0.05997333427270254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,32,power_law_1.01,0.09427840113639832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,32,balanced,0.06211733321348826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,32,balanced,0.06617600222428639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,32,balanced,0.02658133457104365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,32,power_law_1.01,0.07552639842033386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,32,balanced,0.06483733157316844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,32,power_law_1.01,0.05733759999275208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,32,balanced,0.06820799907048543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,32,power_law_1.01,0.07029119729995728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,32,balanced,0.031712000568707786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,32,balanced,0.0286613330245018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,32,power_law_1.01,0.07330560088157653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,32,balanced,0.07381333410739899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,32,balanced,0.02514133354028066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,32,power_law_1.01,0.07915520071983337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,32,balanced,0.025466665625572205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,32,power_law_1.01,0.0791808009147644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,32,balanced,0.08054399987061818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,32,balanced,0.025583999852339428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,32,power_law_1.01,0.0790719985961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,32,balanced,0.0359199990828832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,32,power_law_1.01,0.07948160171508789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,32,balanced,0.03706666578849157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,32,power_law_1.01,0.08236160278320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,32,balanced,0.10302933057149251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,32,balanced,0.04955733319123586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,32,power_law_1.01,0.08076800107955932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,32,balanced,0.0476693312327067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,32,power_law_1.01,0.08321920037269592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,32,power_law_1.01,0.05960959792137146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,32,power_law_1.01,0.08394240140914917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,32,balanced,0.056549335519472756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,32,balanced,0.1132426659266154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,32,power_law_1.01,0.0889087975025177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,32,balanced,0.08005333443482716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,32,power_law_1.01,0.09607679843902588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,32,balanced,0.07946666578451793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,32,power_law_1.01,0.05606399774551392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,32,power_law_1.01,0.09841920137405395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,32,balanced,0.08099733293056488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,32,power_law_1.01,0.04437119960784912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,32,power_law_1.01,0.10355839729309083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,32,balanced,0.1392159958680471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,32,balanced,0.08474133412043254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,32,power_law_1.01,0.1217344045639038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,32,power_law_1.01,0.0505407989025116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,32,power_law_1.01,0.11498240232467652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,32,balanced,0.08531733353932698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,32,power_law_1.01,0.050988799333572386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,32,power_law_1.01,0.14194560050964355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,32,balanced,0.11326932907104492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,32,power_law_1.01,0.06097919940948486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,32,power_law_1.01,0.1651136040687561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,32,power_law_1.01,0.20154879093170167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,32,power_law_1.01,0.06117119789123535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,32,balanced,0.15612266461054483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,32,balanced,0.14485333363215128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,32,power_law_1.01,0.22712318897247313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,32,power_law_1.01,0.061875200271606444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,32,power_law_1.01,0.3048192024230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,32,balanced,0.09131733576456706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,32,power_law_1.01,0.06318079829216003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,32,power_law_1.01,0.3770560026168823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,32,power_law_1.01,0.06702079772949218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,32,power_law_1.01,0.5480703830718994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,32,balanced,0.1151626706123352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,32,power_law_1.01,0.06590719819068909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,32,power_law_1.01,0.6312255859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,32,power_law_1.01,0.07056639790534973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,32,balanced,0.2017013430595398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,32,power_law_1.01,0.8308032035827637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,32,balanced,0.13825600345929465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,32,power_law_1.01,0.07345280051231384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,32,power_law_1.01,0.07575039863586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,32,power_law_1.01,1.2714688301086425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,32,power_law_1.01,0.0812928020954132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,32,power_law_1.01,0.08757759928703308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,32,power_law_1.01,2.448281669616699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,32,balanced,0.18837867180506387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,32,power_law_1.01,0.08979200124740601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,32,power_law_1.01,0.10839680433273316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,32,balanced,0.2299999992052714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,32,power_law_1.01,0.10734080076217652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,32,power_law_1.01,0.13663359880447387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,32,power_law_1.01,0.05475199818611145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,32,balanced,0.2414720058441162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,32,power_law_1.01,0.16171519756317138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,32,power_law_1.01,0.20297598838806152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,32,power_law_1.01,0.050937598943710326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,32,power_law_1.01,0.2526911973953247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,32,power_law_1.01,0.048198398947715757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,32,power_law_1.01,0.35146241188049315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,32,power_law_1.01,0.05110399723052979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,32,power_law_1.01,0.4394495964050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,32,power_law_1.01,0.052019202709198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,32,balanced,0.3349706729253133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,32,power_law_1.01,0.05511680245399475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,32,power_law_1.01,0.6869376182556153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,32,power_law_1.01,0.05544319748878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,32,balanced,0.3197919925053914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,32,power_law_1.01,0.7827648162841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,32,power_law_1.01,0.05663359761238098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,32,power_law_1.01,0.05697280168533325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,32,power_law_1.01,0.9594944000244141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,32,power_law_1.01,0.05678079724311828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,32,power_law_1.01,0.05802879929542541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,32,power_law_1.01,1.5464320182800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,32,power_law_1.01,0.059308797121047974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,32,balanced,0.43596800168355304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,32,power_law_1.01,0.0243136003613472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,32,power_law_1.01,0.058739197254180905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,32,power_law_1.01,3.000716781616211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,32,power_law_1.01,0.06294400095939637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,32,power_law_1.01,0.02417919933795929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,32,power_law_1.01,0.06863359808921814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,32,power_law_1.01,0.07182080149650574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,32,power_law_1.01,0.023155200481414794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,32,power_law_1.01,0.07767680287361145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,32,power_law_1.01,0.024294400215148927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,32,power_law_1.01,0.09031680226325989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,32,power_law_1.01,0.027507200837135315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.01,0.09952639937400817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,32,power_law_1.01,0.03552640080451965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.01,0.12507519721984864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,32,power_law_1.01,0.05394560098648071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.01,0.1335744023323059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.01,0.17912319898605347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,32,balanced,0.39111467202504474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,32,power_law_1.01,0.05399680137634277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.01,0.2029439926147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,32,balanced,0.6400426626205444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.01,0.26406400203704833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,32,power_law_1.01,0.053324800729751584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.01,0.34828801155090333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,32,power_law_1.01,0.059545600414276124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.01,0.48055038452148435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,32,power_law_1.01,0.060108798742294314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,32,power_law_1.01,0.05925120115280151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.01,0.6048768043518067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,32,power_law_1.01,0.0574783980846405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,32,power_law_1.01,0.7342080116271973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,32,power_law_1.01,0.056505602598190305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,32,power_law_1.01,0.054016000032424925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,32,power_law_1.01,1.1942399978637694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,32,power_law_1.01,0.054688000679016115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,32,power_law_1.01,0.052121597528457644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,32,power_law_1.01,2.400044822692871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,32,power_law_1.01,0.06763520240783691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.01,0.06895999908447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.01,0.08266879916191101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,32,balanced,0.8485120137532552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.01,0.10008319616317748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.01,0.12192000150680542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,32,balanced,0.46744000911712646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.01,0.14647680521011353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.01,0.2128063917160034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.01,0.2866879940032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.01,0.4019455909729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.01,0.5168255805969239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,32,power_law_1.01,0.5933887958526611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,32,power_law_1.01,1.004800033569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,32,power_law_1.01,2.003071975708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,32,balanced,1.048906644185384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,2,32,power_law_1.2,0.09520639777183533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,2,32,power_law_1.2,0.059628802537918094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,2,32,power_law_1.2,0.06453760266304016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,2,32,power_law_1.2,0.05115519762039185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,2,32,power_law_1.2,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,2,32,power_law_1.2,0.043968001008033754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,2,32,power_law_1.2,0.06325119733810425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,2,32,power_law_1.2,0.0685375988483429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,2,32,power_law_1.2,0.04839679896831513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,2,32,power_law_1.2,0.07955840229988098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,2,32,power_law_1.2,0.05223039984703064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,32,balanced,0.7022026379903158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,2,32,power_law_1.2,0.07988479733467102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,2,32,power_law_1.2,0.06277120113372803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,2,32,power_law_1.2,0.08250240087509156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,2,32,power_law_1.2,0.06028159856796265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,2,32,power_law_1.2,0.08124160170555114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,2,32,power_law_1.2,0.06372479796409607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,2,32,power_law_1.2,0.08144000172615051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,2,32,power_law_1.2,0.06675840020179749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,2,32,power_law_1.2,0.08202880024909973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,2,32,power_law_1.2,0.06436480283737182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,2,32,power_law_1.2,0.08794879913330078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,2,32,power_law_1.2,0.070387202501297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,2,32,power_law_1.2,0.08640639781951905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,2,32,power_law_1.2,0.07255679965019227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,2,32,power_law_1.2,0.09477120041847228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,2,32,power_law_1.2,0.07392640113830566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,2,32,power_law_1.2,0.0992959976196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,2,32,power_law_1.2,0.07900159955024719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,2,32,power_law_1.2,0.09850239753723145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,2,32,power_law_1.2,0.08538879752159119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,2,32,power_law_1.2,0.11310080289840699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,2,32,power_law_1.2,0.08776320219039917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,2,32,power_law_1.2,0.13703680038452148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,2,32,power_law_1.2,0.09602559804916382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,2,32,power_law_1.2,0.13898240327835082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,2,32,power_law_1.2,0.11281919479370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,32,balanced,1.6672852834065754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,2,32,power_law_1.2,0.1696447968482971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,2,32,power_law_1.2,0.12126079797744752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,2,32,power_law_1.2,0.18229759931564332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,2,32,power_law_1.2,0.14529919624328613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,2,32,power_law_1.2,0.24792959690093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,2,32,power_law_1.2,0.1638527989387512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,2,32,power_law_1.2,0.2645440101623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,2,32,power_law_1.2,0.22583041191101075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,2,32,power_law_1.2,0.37154560089111327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,2,32,power_law_1.2,0.269215989112854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,2,32,power_law_1.2,0.5243711948394776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,2,32,power_law_1.2,0.374617600440979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,2,32,power_law_1.2,0.812992000579834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,2,32,power_law_1.2,0.5481599807739258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,2,32,power_law_1.2,1.0919872283935548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,2,32,power_law_1.2,0.8098560333251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,2,32,power_law_1.2,0.997152042388916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,2,32,power_law_1.2,1.335923194885254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,2,32,power_law_1.2,1.2442048072814942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,2,32,power_law_1.2,2.1780607223510744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,2,32,power_law_1.2,1.8151679992675782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,2,32,power_law_1.2,3.5019840240478515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,2,32,power_law_1.2,4.342700958251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,32,balanced,1.3539892832438152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,32,balanced,3.2944374084472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,2,32,power_law_1.2,0.05438719987869263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,2,32,power_law_1.2,0.04848639965057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,2,32,power_law_1.2,0.0474368005990982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,2,32,power_law_1.2,0.05088000297546387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,2,32,power_law_1.2,0.052070397138595584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,2,32,power_law_1.2,0.055078399181365964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,2,32,power_law_1.2,0.05509120225906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,2,32,power_law_1.2,0.05605760216712952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,2,32,power_law_1.2,0.055955201387405396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,2,32,power_law_1.2,0.05614719986915588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,2,32,power_law_1.2,0.057055997848510745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,2,32,power_law_1.2,0.05864319801330566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,2,32,power_law_1.2,0.059699201583862306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,2,32,power_law_1.2,0.06385279893875122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,2,32,power_law_1.2,0.07000319957733155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,2,32,power_law_1.2,0.07635200023651123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,2,32,power_law_1.2,0.024217599630355836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,2,32,power_law_1.2,0.08353919982910156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,2,32,power_law_1.2,0.09420160055160523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,2,32,power_law_1.2,0.022201600670814513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.2,0.10652799606323242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,2,32,power_law_1.2,0.023481599986553192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.2,0.1296447992324829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,2,32,power_law_1.2,0.024953599274158477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.2,0.1447551965713501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,2,32,power_law_1.2,0.027385601401329042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.2,0.1855679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,2,32,power_law_1.2,0.03594239950180054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.2,0.22576000690460205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,2,32,power_law_1.2,0.052774399518966675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.2,0.31347839832305907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,2,32,power_law_1.2,0.053452801704406736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.2,0.381606388092041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,2,32,power_law_1.2,0.0536575973033905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.2,0.7040128231048584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,2,32,power_law_1.2,0.05794559717178345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.2,0.7907072067260742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,2,32,power_law_1.2,0.05857279896736145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,2,32,power_law_1.2,1.04716796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,2,32,power_law_1.2,0.05886719822883606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,2,32,power_law_1.2,0.05759360194206238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,2,32,power_law_1.2,1.7439231872558594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,2,32,power_law_1.2,0.054745602607727054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,2,32,power_law_1.2,0.055174398422241214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,2,32,power_law_1.2,3.5002815246582033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,2,32,power_law_1.2,0.05707520246505737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,2,32,power_law_1.2,0.053932797908782956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,2,32,power_law_1.2,0.06279039978981019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.2,0.0680191993713379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.2,0.08893439769744874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.2,0.10101120471954346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.2,0.14385279417037963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.2,0.1588544011116028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.2,0.23398399353027344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.2,0.32561919689178465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.2,0.41457281112670896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.2,0.6139328002929687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,2,32,power_law_1.2,0.7108223915100098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,2,32,power_law_1.2,0.9872639656066895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,2,32,power_law_1.2,2.427788734436035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,16,balanced,0.023103999594847362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,16,balanced,0.02314666658639908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,16,balanced,0.02508266766866048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,16,balanced,0.023797333240509033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,16,balanced,0.02681066592534383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,16,balanced,0.029215998947620392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,16,balanced,0.03600533306598663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,16,balanced,0.03217600037654241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,16,balanced,0.03898133337497711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,16,balanced,0.03920533259709676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,16,balanced,0.04308266441027323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,16,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,16,balanced,0.05136000116666158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,16,balanced,0.06402666866779327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,16,balanced,0.04994666576385498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,16,balanced,0.0835093359152476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,16,balanced,0.051685333251953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,16,balanced,0.08444799979527791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,16,balanced,0.04976533353328705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,16,balanced,0.08872000376383464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,16,balanced,0.052416001756985985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,16,balanced,0.055770665407180786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,16,balanced,0.12040533622105916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,16,balanced,0.05804799993832906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,16,balanced,0.05807466804981232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,16,balanced,0.09259733557701111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,16,balanced,0.05955199897289276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,16,balanced,0.0598880002895991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,16,balanced,0.11271466811498006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,16,balanced,0.05982399980227152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,16,balanced,0.06215466558933258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,16,balanced,0.060266668597857155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,16,balanced,0.12517866492271423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,16,balanced,0.06221333146095276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,16,balanced,0.06422399977842967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,16,balanced,0.06805866460005443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,16,balanced,0.16022400061289468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,16,balanced,0.06832533578077953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,16,balanced,0.07045866549015045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,16,power_law_1.01,0.0785152018070221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,16,balanced,0.07970133423805237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,16,power_law_1.01,0.07548159956932068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,16,balanced,0.1930613319079081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,16,balanced,0.08710400263468425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,16,power_law_1.01,0.057555198669433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,16,power_law_1.01,0.06490880250930786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,16,balanced,0.1150986651579539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,16,power_law_1.01,0.06890239715576171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,16,power_law_1.01,0.0788096010684967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,16,balanced,0.12339733044306438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,16,power_law_1.01,0.08119040131568908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,16,power_law_1.01,0.07957760095596314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,16,balanced,0.28033600250879925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,16,power_law_1.01,0.08001279830932617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,16,balanced,0.15690132975578308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,16,power_law_1.01,0.07802879810333252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,16,balanced,0.34990398089090985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,16,balanced,0.1686506668726603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,16,power_law_1.01,0.07927039861679078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,16,power_law_1.01,0.08367999792098998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,16,power_law_1.01,0.08288639783859253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,16,power_law_1.01,0.08363519906997681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,16,balanced,0.22756266593933105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,16,power_law_1.01,0.08933119773864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,16,power_law_1.01,0.0926912009716034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,16,power_law_1.01,0.09634559750556945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,16,balanced,0.5036480029424032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,16,power_law_1.01,0.10546560287475586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,16,power_law_1.01,0.023795199394226075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.01,0.1202623963356018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.01,0.1388479948043823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,16,balanced,0.274944007396698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,16,power_law_1.01,0.02366719990968704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.01,0.16442879438400268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.01,0.20592639446258545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,16,power_law_1.01,0.02120320051908493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.01,0.23377280235290526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,16,power_law_1.01,0.022489599883556366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.01,0.3221888065338135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.01,0.40556797981262205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,16,power_law_1.01,0.02451840043067932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.01,0.6247871875762939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,16,power_law_1.01,0.025088000297546386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.01,0.7238207817077636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,16,power_law_1.01,0.02592639923095703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,16,power_law_1.01,0.02803199887275696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,16,power_law_1.01,0.9663935661315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,16,balanced,0.38399465878804523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,16,balanced,0.6654026508331299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,16,power_law_1.01,0.02885119915008545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,16,power_law_1.01,1.4098688125610352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,16,power_law_1.01,0.028723201155662535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,16,power_law_1.01,0.04088320136070252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,16,power_law_1.01,3.3166080474853517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,16,power_law_1.01,0.039955198764801025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,16,power_law_1.01,0.04178560078144074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,16,power_law_1.01,0.05764480233192444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,16,power_law_1.01,0.05850239992141724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,16,power_law_1.01,0.05969280004501343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,16,power_law_1.01,0.07676159739494323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,16,balanced,0.46727999051411945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,16,power_law_1.01,0.08802559971809387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.01,0.07518079876899719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.01,0.08195840120315552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.01,0.0930176019668579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,16,balanced,0.8111680348714193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.01,0.10788480043411255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.01,0.14133119583129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.01,0.18775039911270142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.01,0.23431680202484131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.01,0.3286976099014282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,16,balanced,0.04773333172003428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.01,0.4351039886474609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,16,balanced,0.04859733581542969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,16,power_law_1.01,0.5392896175384522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,16,balanced,0.04799999793370565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,16,balanced,0.050026665131251015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,16,power_law_1.01,0.8731136322021484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,16,balanced,0.07670400043328603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,16,balanced,0.5642879803975424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,16,balanced,0.08345066507657369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,16,power_law_1.01,1.7556224822998048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,16,balanced,0.08190933366616567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,16,balanced,0.04594666759173075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,16,balanced,0.08458667000134786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,16,balanced,0.08556800087292989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,16,balanced,0.08460799853006999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,16,balanced,0.04569066564242045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,16,balanced,0.08482133348782857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,16,balanced,0.04569066564242045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,16,balanced,0.04786133269468943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,16,balanced,0.08658132950464885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,16,balanced,0.05541333556175232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,16,balanced,0.06654933094978333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,16,balanced,0.08662399649620056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,16,balanced,0.0686773310105006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,16,balanced,0.06804800033569336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,16,balanced,0.08724266290664673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,16,balanced,0.06905599931875865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,16,balanced,0.09302399555842082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,16,balanced,1.2743199666341145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,16,balanced,0.07003733515739441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,16,balanced,0.09299733241399129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,16,balanced,0.07150400181611379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,16,balanced,0.07075733443101247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,16,balanced,0.09517332911491394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,16,balanced,0.07482133309046428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,16,balanced,0.10417067011197408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,16,balanced,0.07518933216730754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,16,balanced,0.08079466720422109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,16,balanced,0.1039626697699229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,16,balanced,0.0830026666323344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,16,balanced,0.12586667140324911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,16,balanced,0.08585066596666972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,16,balanced,0.8617066542307535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,16,balanced,0.09880533814430237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,16,balanced,0.12800000111262003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,16,balanced,0.10123200217882793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,16,balanced,0.159578671058019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,16,balanced,0.12821333607037863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,16,balanced,0.1402293344338735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,16,balanced,0.1698933243751526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,16,balanced,0.18587199846903482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,16,balanced,0.2378773291905721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,16,balanced,0.20742400487263998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,16,balanced,0.2688000003496806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,16,balanced,0.2871893246968587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,16,balanced,0.3469173510869344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,16,balanced,0.36903464794158936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,16,balanced,2.4933066368103027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,16,balanced,0.49578134218851727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,16,balanced,0.4377439816792806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,16,balanced,1.6697173118591309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,16,power_law_1.01,0.07713919878005981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,16,power_law_1.01,0.07414399981498718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,16,power_law_1.01,0.05380480289459229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,16,balanced,0.6028960148493449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,16,power_law_1.01,0.05761280059814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,16,power_law_1.01,0.05866879820823669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,16,power_law_1.01,0.06504960060119629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,16,power_law_1.01,0.0769216001033783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,16,power_law_1.01,0.05925760269165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,16,power_law_1.01,0.07801600098609925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,16,power_law_1.01,0.04899199903011322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,16,power_law_1.01,0.08012800216674805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,16,power_law_1.01,0.049798399209976196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,16,balanced,0.5050133466720581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,16,power_law_1.01,0.05223039984703064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,16,power_law_1.01,0.07923840284347534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,16,power_law_1.01,0.06323199868202209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,16,power_law_1.01,0.06220800280570984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,16,power_law_1.01,0.07474560141563416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,16,power_law_1.01,0.06537600159645081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,16,power_law_1.01,0.08172799944877625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,16,power_law_1.01,0.06538239717483521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,16,power_law_1.01,0.08358399868011475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,16,power_law_1.01,0.06691840291023254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,16,power_law_1.01,0.08453760147094727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,16,power_law_1.01,0.0664192020893097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,16,power_law_1.01,0.08740479946136474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,16,power_law_1.01,0.07473919987678528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,16,balanced,0.7249546845753988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,16,power_law_1.01,0.08961920142173767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,16,power_law_1.01,0.07479040026664734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,16,power_law_1.01,0.09475200176239014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,16,power_law_1.01,0.07481600046157837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,16,power_law_1.01,0.10236159563064576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,16,power_law_1.01,0.08085119724273682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,16,power_law_1.01,0.11827199459075928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,16,power_law_1.01,0.08549119830131531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,16,power_law_1.01,0.11866879463195801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,16,power_law_1.01,0.0933184027671814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,16,power_law_1.01,0.1414463996887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,16,power_law_1.01,0.10823680162429809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,16,power_law_1.01,0.15136640071868895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,16,power_law_1.01,0.1189120054244995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,16,power_law_1.01,0.18804479837417604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,16,power_law_1.01,0.13882240056991577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,16,power_law_1.01,0.23567359447479247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,16,power_law_1.01,0.1663424015045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,16,power_law_1.01,0.3219072103500366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,16,power_law_1.01,0.2159679889678955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,16,power_law_1.01,0.4055488109588623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,16,power_law_1.01,0.25267839431762695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,16,power_law_1.01,0.5407999992370606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,16,power_law_1.01,0.35810561180114747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,16,power_law_1.01,0.6792128086090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,16,power_law_1.01,0.43695998191833496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,16,balanced,0.7630720138549805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,16,power_law_1.01,0.8160639762878418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,16,power_law_1.01,0.6583295822143554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,16,power_law_1.01,1.1522496223449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,16,power_law_1.01,0.8224127769470215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,16,power_law_1.2,0.07219840288162231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,16,power_law_1.01,0.9528575897216797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,16,power_law_1.2,0.05138559937477112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,16,power_law_1.01,2.3874624252319334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,16,power_law_1.2,0.04708479940891266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,16,power_law_1.01,1.5877696037292481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,16,power_law_1.2,0.05542399883270264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,16,power_law_1.2,0.06493440270423889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,16,balanced,1.1141280333201091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,16,power_law_1.2,0.07296640276908875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,16,power_law_1.01,2.9685504913330076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,16,power_law_1.2,0.07762560248374939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,16,power_law_1.2,0.07939839959144593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,16,power_law_1.2,0.07916799783706666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,16,power_law_1.2,0.07918720245361328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,16,power_law_1.2,0.08138880133628845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,16,power_law_1.2,0.08326399922370911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,16,power_law_1.2,0.08316159844398499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,16,power_law_1.2,0.058745598793029784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,16,power_law_1.2,0.08281599879264831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,16,power_law_1.2,0.09271039962768554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,16,power_law_1.2,0.0436928004026413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,16,power_law_1.2,0.098854398727417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,16,power_law_1.2,0.10621440410614014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,16,power_law_1.2,0.04251520037651062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,16,power_law_1.2,0.1310271978378296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,16,power_law_1.2,0.04872959852218628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,16,power_law_1.2,0.05100160241127014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,16,power_law_1.2,0.12150399684906006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,16,power_law_1.2,0.05751039981842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,16,power_law_1.2,0.14846080541610718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,16,power_law_1.2,0.06000639796257019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,16,power_law_1.2,0.17103999853134155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,16,power_law_1.2,0.06119040250778198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,16,power_law_1.2,0.2157439947128296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,16,power_law_1.2,0.06378880143165588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,16,power_law_1.2,0.26788480281829835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,16,power_law_1.2,0.0660863995552063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,16,power_law_1.2,0.3753472089767456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,16,power_law_1.2,0.07089279890060425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,16,power_law_1.2,0.5653759956359863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,16,power_law_1.2,0.07420799732208253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,16,power_law_1.2,0.07470080256462097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,16,power_law_1.2,0.6362048149108886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,16,power_law_1.2,0.07464960217475891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,16,power_law_1.2,1.0023679733276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,16,power_law_1.2,0.08488320112228394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,16,power_law_1.2,0.08750079870223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,16,power_law_1.2,1.0225024223327637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,16,power_law_1.2,0.09373440146446228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,16,power_law_1.2,1.6177024841308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,16,power_law_1.2,0.11255680322647095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,16,power_law_1.2,0.1165120005607605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,16,power_law_1.2,0.15185279846191407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,16,power_law_1.2,3.740300750732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,16,power_law_1.2,0.1673151969909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,16,power_law_1.2,0.22740480899810792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,16,power_law_1.2,0.2919935941696167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,16,power_law_1.2,0.4086143970489502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,16,power_law_1.2,0.5250175952911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,16,balanced,1.4394240379333496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,16,power_law_1.2,0.8119104385375977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,16,power_law_1.2,0.8462080001831055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,16,power_law_1.2,1.1919296264648438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,16,power_law_1.2,2.089401626586914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,16,balanced,2.1702720324198403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,16,power_law_1.2,4.3810688018798825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,16,power_law_1.2,0.08119040131568908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,16,power_law_1.2,0.06698240041732788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,16,power_law_1.2,0.06099200248718262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,16,power_law_1.2,0.07028480172157288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,16,power_law_1.2,0.06955519914627076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,16,power_law_1.2,0.08505600094795226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,16,power_law_1.2,0.08134400248527526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,16,power_law_1.2,0.08130559921264649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,16,power_law_1.2,0.0824832022190094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,16,power_law_1.2,0.08312320113182067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,16,power_law_1.2,0.08353279829025269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,16,power_law_1.2,0.08721280097961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,16,power_law_1.2,0.023980799317359924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,16,power_law_1.2,0.08463360071182251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,16,power_law_1.2,0.02019200026988983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,16,power_law_1.2,0.08624640107154846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,16,power_law_1.2,0.021907199919223786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,16,power_law_1.2,0.0915008008480072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,16,power_law_1.2,0.022303999960422517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,16,power_law_1.2,0.09703680276870727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,16,power_law_1.2,0.02391040027141571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,16,power_law_1.2,0.10238720178604126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,16,power_law_1.2,0.024665600061416625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,16,power_law_1.2,0.11291520595550537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,16,power_law_1.2,0.027161601185798644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.2,0.12437119483947753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,16,power_law_1.2,0.028313601016998292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,16,power_law_1.2,0.02929919958114624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.2,0.14256000518798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,16,power_law_1.2,0.027775999903678895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.2,0.17103999853134155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,16,power_law_1.2,0.039929598569869995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.2,0.21602559089660645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,16,power_law_1.2,0.04071680009365082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.2,0.27136640548706054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,16,power_law_1.2,0.04188160002231598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.2,0.3730047941207886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,16,power_law_1.2,0.05623040199279785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.2,0.5188032150268554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,16,power_law_1.2,0.059673601388931276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.2,0.7205056190490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,16,power_law_1.2,0.06364160180091857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,16,power_law_1.2,0.08176640272140503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.2,0.8912575721740723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,16,power_law_1.2,0.08973439931869506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,16,power_law_1.2,1.2004287719726563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.2,0.06935679912567139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.2,0.08477439880371093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,16,power_law_1.2,2.1037439346313476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.2,0.09246079921722412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.2,0.11519999504089355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,16,power_law_1.2,3.8669246673583983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.2,0.1440384030342102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.2,0.2014336109161377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.2,0.2518336057662964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.2,0.3680959939956665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.2,0.4933631896972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,16,power_law_1.2,0.637011194229126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,16,power_law_1.2,0.9590208053588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,16,power_law_1.2,2.0705663681030275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,8,balanced,0.04458666841189066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,8,balanced,0.04500266909599304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,8,balanced,0.044031997521718345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,8,balanced,0.052330667773882546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,8,balanced,0.07691733539104462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,8,balanced,0.08221866687138875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,8,balanced,0.08373333017031352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,8,balanced,0.08547733227411906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,8,balanced,0.0844533344109853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,8,balanced,0.08589866757392883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,8,balanced,0.08657067020734151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,8,balanced,0.08798399567604065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,8,balanced,0.08649067083994548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,8,balanced,0.09244799613952637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,8,balanced,0.09443199634552002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,8,balanced,0.09748799602190654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,8,balanced,0.10114133358001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,8,balanced,0.10935999949773152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,8,balanced,0.11422399679819743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,8,balanced,0.1377226710319519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,8,balanced,0.1448853313922882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,8,balanced,0.1858773430188497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,8,balanced,0.0443146675825119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,8,balanced,0.20764267444610596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,8,balanced,0.04574400186538696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,8,balanced,0.04619200030962626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,8,balanced,0.050069332122802734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,8,balanced,0.05621333420276642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,8,balanced,0.06869866450627644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,8,balanced,0.07150933146476746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,8,balanced,0.2829386591911316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,8,balanced,0.07260266443093617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,8,balanced,0.07190399865309398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,8,balanced,0.07422400017579396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,8,balanced,0.07394133508205414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,8,balanced,0.07491733133792877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,8,balanced,0.07640533149242401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,8,balanced,0.08063466846942902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,8,balanced,0.08679466446240743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,8,balanced,0.33316266536712646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,8,balanced,0.09009066224098206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,8,balanced,0.09378133217493693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,8,balanced,0.10505066315333049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,8,balanced,0.11307733257611592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,8,power_law_1.01,0.05720319747924805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,8,balanced,0.14219199617703757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,8,power_law_1.01,0.06531199812889099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,8,power_law_1.01,0.04959999918937683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,8,power_law_1.01,0.05367680191993714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,8,balanced,0.16100266575813293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,8,balanced,0.4657013416290283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,8,power_law_1.01,0.06279039978981019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,8,power_law_1.01,0.07624319791793824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,8,power_law_1.01,0.07710719704627991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,8,balanced,0.2156426707903544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,8,power_law_1.01,0.0766592025756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,8,power_law_1.01,0.0803712010383606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,8,power_law_1.01,0.08092799782752991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,8,power_law_1.01,0.0822655975818634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,8,power_law_1.01,0.0840511977672577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,8,balanced,0.2413439949353536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,8,power_law_1.01,0.08525440096855164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,8,power_law_1.01,0.09020159840583801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,8,power_law_1.01,0.09719039797782898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,8,power_law_1.01,0.10532480478286743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,8,power_law_1.01,0.10900479555130005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,8,balanced,0.3429653247197469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,8,power_law_1.01,0.12177920341491699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,8,balanced,0.5813546578089396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,8,power_law_1.01,0.04859519898891449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,8,power_law_1.01,0.1276352047920227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,8,power_law_1.2,0.05379840135574341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,8,power_law_1.01,0.1468287944793701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,8,power_law_1.01,0.056704002618789676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,8,power_law_1.01,0.1784832000732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,8,power_law_1.01,0.04992640018463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,8,power_law_1.2,0.04926080107688904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,8,power_law_1.01,0.21094400882720948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,8,power_law_1.01,0.05084159970283508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,8,power_law_1.2,0.051737600564956666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,8,power_law_1.01,0.05448319911956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,8,power_law_1.01,0.27089920043945315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,8,power_law_1.01,0.06123520135879516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,8,power_law_1.2,0.055622398853302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,8,power_law_1.01,0.36738560199737547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,8,balanced,0.4172000090281169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,8,power_law_1.01,0.06364160180091857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,8,power_law_1.2,0.06053119897842407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,8,power_law_1.01,0.4455872058868408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,8,power_law_1.01,0.06984959840774536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,8,power_law_1.2,0.07697920203208923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,8,power_law_1.01,0.5925312042236328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,8,power_law_1.2,0.0775551974773407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,8,power_law_1.01,0.06698880195617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,8,power_law_1.01,0.7651264190673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,8,power_law_1.2,0.0800320029258728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,8,power_law_1.2,0.04800640046596527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,8,power_law_1.01,0.06974719762802124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,8,power_law_1.2,0.08090879917144775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,8,power_law_1.01,0.07334399819374085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,8,power_law_1.01,1.0005632400512696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,8,power_law_1.01,0.07626240253448487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,8,power_law_1.2,0.04536960124969482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,8,power_law_1.2,0.08216320276260376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,8,power_law_1.01,1.539840030670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,8,power_law_1.01,0.07761920094490052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,8,power_law_1.2,0.046009600162506104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,8,power_law_1.01,0.08256639838218689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,8,power_law_1.2,0.0853056013584137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,8,power_law_1.2,0.04952319860458374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,8,balanced,0.6898773511250814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,8,power_law_1.01,2.7204095840454103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,8,power_law_1.01,0.08967679738998413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,8,balanced,0.5794026851654053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,8,power_law_1.2,0.053452801704406736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,8,power_law_1.2,0.08944640159606934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,8,power_law_1.2,0.06136959791183472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,8,power_law_1.01,0.09548799991607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,8,power_law_1.2,0.06386560201644897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,8,power_law_1.2,0.08993279933929443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,8,power_law_1.01,0.10041600465774536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,8,power_law_1.2,0.06677119731903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,8,power_law_1.2,0.09541119933128357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,8,power_law_1.01,0.11114879846572875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,8,power_law_1.2,0.07031679749488831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,8,power_law_1.2,0.10524159669876099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,8,power_law_1.2,0.07069439888000488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,8,power_law_1.01,0.12627840042114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,8,power_law_1.2,0.0732096016407013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,8,power_law_1.2,0.11010559797286987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,8,power_law_1.2,0.0769599974155426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,8,power_law_1.01,0.15352319478988646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,8,power_law_1.2,0.10871679782867431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,8,power_law_1.2,0.08241919875144958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,8,power_law_1.01,0.18444160223007203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,8,power_law_1.2,0.08472319841384887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,8,power_law_1.2,0.12139519453048705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,8,power_law_1.01,0.23895039558410644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,8,power_law_1.2,0.09229440093040467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,8,power_law_1.2,0.13289599418640136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,8,power_law_1.2,0.09611520171165466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,8,power_law_1.01,0.2981055974960327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,8,power_law_1.2,0.16850559711456298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,8,power_law_1.2,0.0989247977733612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,8,power_law_1.01,0.40207362174987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,8,power_law_1.2,0.11367679834365844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,8,power_law_1.2,0.1946176052093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,8,balanced,0.7513066927591959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,8,power_law_1.01,0.5290048122406006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,8,power_law_1.2,0.1260800004005432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,8,power_law_1.2,0.2393728017807007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,8,power_law_1.2,0.15750399827957154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,8,power_law_1.2,0.3012671947479248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,8,power_law_1.01,0.7434944152832031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,8,power_law_1.2,0.1905087947845459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,8,power_law_1.2,0.4040063858032227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,8,power_law_1.2,0.24972159862518312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,8,power_law_1.01,0.9341055870056152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,8,power_law_1.2,0.5116096019744873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,8,power_law_1.2,0.34399359226226806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,8,power_law_1.01,1.144371223449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,8,power_law_1.2,0.8174271583557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,8,power_law_1.2,0.4593791961669922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,8,power_law_1.01,1.8459135055541993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,8,power_law_1.2,0.5948927879333497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,8,power_law_1.2,0.8600640296936035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,8,power_law_1.2,0.844217586517334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,8,power_law_1.2,1.3061375617980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,8,power_law_1.2,1.277984046936035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,8,power_law_1.01,3.5082046508789064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,8,balanced,1.0452640056610107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,8,power_law_1.2,2.0347776412963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,8,power_law_1.2,1.413036823272705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,8,power_law_1.2,2.299577522277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,8,power_law_1.2,3.8607486724853515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,8,power_law_1.2,4.474687957763672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,8,balanced,0.9169119993845621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,8,balanced,1.4017759958902996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,4,balanced,0.04576000074545542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,4,balanced,0.046015997727712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,4,balanced,0.05113600194454193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,4,balanced,0.06140799820423126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,8,balanced,1.991194725036621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,4,balanced,0.08461866776148479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,4,balanced,0.09104532996813457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,4,balanced,0.09173867106437683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,4,balanced,0.09408533573150635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,4,balanced,0.09489066402117412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,4,balanced,0.09308800101280212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,4,balanced,0.09673600395520528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,4,balanced,0.09778666496276855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,4,balanced,0.09741866588592529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,4,balanced,0.09893332918485005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,4,balanced,0.106495996316274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,4,balanced,0.10840533177057902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,4,balanced,0.11319999893506368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,4,balanced,0.12417599558830261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,4,balanced,0.13150399923324585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,4,balanced,0.17090133825937906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,4,balanced,0.18490133682886759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,4,balanced,0.2534240086873372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,4,balanced,0.041637333730856575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,4,balanced,0.2768266598383586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,4,balanced,0.04562666515509287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,4,balanced,0.04775999983151754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,8,balanced,2.73472531636556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,4,balanced,0.05184000233809153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,4,balanced,0.06449066599210103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,4,balanced,0.07714133461316426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,4,balanced,0.07642666498819987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,4,balanced,0.08045866588751475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,4,balanced,0.4012426535288493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,4,balanced,0.08072533210118611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,4,balanced,0.07959466675917308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,4,balanced,0.0820906658967336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,4,balanced,0.08486933509508769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,4,balanced,0.08669333656628926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,4,balanced,0.08925333619117737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,4,balanced,0.09718400239944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,4,balanced,0.4869226614634196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,4,balanced,0.10019733508427937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,4,balanced,0.10635733604431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,4,balanced,0.12069333593050639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,4,balanced,0.12949867049853006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,4,balanced,0.1704639991124471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,4,balanced,0.19504000743230185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,4,balanced,0.6581653356552124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,4,balanced,0.2702133258183797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,4,balanced,0.3111093242963155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,4,power_law_1.01,0.04947839975357056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,4,power_law_1.01,0.04999040067195892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,4,power_law_1.01,0.046854400634765626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,4,balanced,0.45204798380533856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,4,power_law_1.01,0.058847999572753905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,4,balanced,0.8537387053171793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,4,power_law_1.01,0.06515200138092041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,4,power_law_1.01,0.08054400086402894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,4,power_law_1.01,0.08376960158348083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,4,power_law_1.01,0.08542720079421998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,4,power_law_1.01,0.0859008014202118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,4,balanced,0.5678293307622274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,4,power_law_1.01,0.08925439715385437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,4,power_law_1.01,0.09028480052947999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,4,power_law_1.01,0.09228799939155578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,4,power_law_1.01,0.09410560131072998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,4,power_law_1.01,0.09621760249137878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,4,power_law_1.01,0.11252479553222657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,4,power_law_1.01,0.1158911943435669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,4,power_law_1.01,0.11649279594421387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,4,power_law_1.01,0.12995840311050416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,4,power_law_1.01,0.14739840030670165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,4,balanced,0.793333371480306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,4,balanced,1.0321813424428303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,4,power_law_1.01,0.18103040456771852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,4,power_law_1.01,0.21520640850067138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,4,power_law_1.01,0.27180159091949463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,4,power_law_1.01,0.347052788734436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,4,power_law_1.01,0.4688576221466064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,4,power_law_1.01,0.5697408199310303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,4,power_law_1.01,0.7990079879760742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,4,power_law_1.01,1.034284782409668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,4,power_law_1.2,0.04871039986610413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,4,power_law_1.01,0.04683519899845123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,4,power_law_1.01,1.3686207771301269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,4,power_law_1.2,0.04557439982891083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,4,power_law_1.01,0.04831359982490539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,4,power_law_1.01,2.159231948852539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,4,power_law_1.2,0.04670720100402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,4,power_law_1.2,0.04699519872665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,4,balanced,1.0383093357086182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,4,power_law_1.2,0.055264002084732054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,4,power_law_1.01,0.04596480131149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,4,power_law_1.2,0.04466559886932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,4,power_law_1.2,0.05989760160446167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,4,power_law_1.01,0.05681279897689819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,4,power_law_1.01,3.833657455444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,4,power_law_1.2,0.04644480049610138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,4,power_law_1.2,0.07826560139656066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,4,power_law_1.01,0.05945600271224975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,4,power_law_1.2,0.05655679702758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,4,power_law_1.2,0.08284159898757934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,4,power_law_1.01,0.06939520239830017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,4,power_law_1.2,0.05824000239372253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,4,power_law_1.2,0.08687360286712646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,4,power_law_1.01,0.07034879922866821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,4,power_law_1.2,0.06940159797668458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,4,power_law_1.2,0.07119359970092773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,4,power_law_1.2,0.08769279718399048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,4,power_law_1.01,0.07304959893226623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,4,power_law_1.2,0.07672960162162781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,4,power_law_1.01,0.07719039916992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,4,power_law_1.2,0.08845440149307252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,4,power_law_1.2,0.07608960270881653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,4,power_law_1.01,0.07824000120162963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,4,power_law_1.2,0.0911679983139038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,4,power_law_1.2,0.08091520071029663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,4,power_law_1.2,0.08633599877357483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,4,power_law_1.01,0.08338559865951538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,4,power_law_1.2,0.09453439712524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,4,power_law_1.2,0.09175040125846863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,4,balanced,1.5857866605122883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,4,power_law_1.01,0.088646399974823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,4,power_law_1.2,0.0965503990650177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,4,power_law_1.2,0.0948736011981964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,4,power_law_1.01,0.09560319781303406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,4,power_law_1.2,0.10107519626617431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,4,power_law_1.2,0.09413120150566101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,4,power_law_1.2,0.11484160423278808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,4,power_law_1.01,0.09071360230445862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,4,power_law_1.2,0.10117119550704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,4,power_law_1.2,0.10371199846267701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,4,balanced,1.263375997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,4,power_law_1.2,0.11793279647827148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,4,power_law_1.01,0.09957759976387023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,4,power_law_1.2,0.11919360160827637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,4,power_law_1.2,0.11457279920578003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,4,power_law_1.01,0.10526080131530761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,4,power_law_1.2,0.13273600339889527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,4,power_law_1.01,0.11190400123596192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,4,power_law_1.2,0.1336127996444702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,4,power_law_1.2,0.14924800395965576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,4,power_law_1.01,0.13084160089492797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,4,power_law_1.2,0.1551743984222412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,4,power_law_1.2,0.18808319568634033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,4,power_law_1.01,0.1479423999786377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,4,power_law_1.2,0.18889600038528442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,4,power_law_1.2,0.23639039993286132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,4,power_law_1.01,0.18254719972610473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,4,power_law_1.2,0.22787840366363527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,4,power_law_1.2,0.3087424039840698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,4,power_law_1.01,0.22773120403289795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,4,power_law_1.2,0.30700159072875977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,4,power_law_1.2,0.39836161136627196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,4,power_law_1.01,0.2933504104614258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,4,power_law_1.2,0.3859519958496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,4,power_law_1.2,0.520633602142334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,4,power_law_1.01,0.3687936067581177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,4,power_law_1.2,0.4827263832092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,4,power_law_1.2,0.7393983840942383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,4,power_law_1.01,0.5358272075653077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,4,power_law_1.2,1.093113613128662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,4,power_law_1.2,0.6528704166412354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,4,power_law_1.01,0.6685696125030518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,4,power_law_1.2,1.343942356109619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,4,power_law_1.2,0.9293951988220215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,4,power_law_1.01,1.0091008186340331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,4,power_law_1.2,1.7448448181152343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,4,power_law_1.2,1.4038144111633302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,4,power_law_1.01,1.1296192169189454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,4,power_law_1.2,3.066374397277832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,4,power_law_1.2,1.5567423820495605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,4,power_law_1.01,1.4706175804138184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,4,power_law_1.2,2.2098560333251953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,4,power_law_1.01,2.3341440200805663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,4,power_law_1.2,5.357183837890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,4,balanced,1.9655520121256511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,4,power_law_1.2,4.806739044189453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,4,power_law_1.01,4.862905502319336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,4,balanced,3.0677172342936196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,4,balanced,3.785146713256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,64,balanced,0.08438400427500407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,64,balanced,0.08458133538564046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,64,balanced,0.08423466483751933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,64,balanced,0.0848426620165507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,64,balanced,0.09917333722114563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,64,balanced,0.127018670241038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,64,balanced,0.17179733514785767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,64,balanced,0.17255467176437378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,64,balanced,0.17328532536824545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,64,balanced,0.17459734280904135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,64,balanced,0.17727466424306235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,64,balanced,0.17051732540130615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,64,balanced,0.1707786719004313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,64,balanced,0.17497066656748453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,64,balanced,0.18022932608922324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,64,balanced,0.17457065979639688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,64,balanced,0.17627733945846558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,64,balanced,0.18785599867502847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,64,balanced,0.1957226594289144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,64,balanced,0.21861867109934488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,64,balanced,0.06258666515350342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,64,balanced,0.06333866715431213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,64,balanced,0.21895466248194376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,64,balanced,0.06233066817124685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,64,balanced,0.06419200201829274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,64,balanced,0.07841066519419353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,64,balanced,0.09262933333714803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,64,balanced,0.24655999739964804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,64,balanced,0.11532266934712727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,64,balanced,0.11346667011578877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,64,balanced,0.11142933368682861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,64,balanced,0.11963199575742085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,64,balanced,0.11757866541544597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,64,balanced,0.27456533908843994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,64,balanced,0.11515200138092041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,64,balanced,0.11737599968910217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,64,balanced,0.11813333630561829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,64,balanced,0.1309653321901957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,64,balanced,0.12558933099110922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,64,balanced,0.3301333387692769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,64,balanced,0.13987200458844504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,64,balanced,0.14146666725476584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,64,balanced,0.15180266896883646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,64,balanced,0.17695466677347818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,64,balanced,0.4339040120442708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,64,balanced,0.1965013345082601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,64,balanced,0.23785599072774252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,64,balanced,0.2824160059293111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,64,balanced,0.5251359939575195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,64,balanced,0.3667413393656413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,64,balanced,0.47764265537261963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,64,balanced,0.6739947001139323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,64,balanced,0.6304906606674194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,64,balanced,0.8153599898020426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,64,balanced,0.8805867036183676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,64,balanced,1.033413330713908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,64,balanced,1.2916159629821777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,2,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,2,balanced,0.05017599960168203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,2,balanced,0.053823997577031456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,2,balanced,0.06610133250554402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,2,balanced,0.09477866689364116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,2,balanced,0.11422933141390483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,64,balanced,1.574618657430013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,2,balanced,0.11321066816647847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,2,balanced,0.10673066973686218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,2,balanced,0.10618133346239726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,2,balanced,0.10732266306877136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,2,balanced,0.10831466317176819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,2,balanced,0.11316800117492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,2,balanced,0.11359999577204387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,2,balanced,0.11619733770688374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,2,balanced,0.12308800220489502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,2,balanced,0.1264959971110026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,2,balanced,0.13199466466903687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,2,balanced,0.15825066963831583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,2,balanced,0.16851733128229776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,2,balanced,0.23341333866119385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,2,balanced,0.24453866481781006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,2,balanced,0.36533331871032715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,2,balanced,0.4082560141881307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,2,balanced,0.5928853352864584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,2,balanced,0.04589866598447164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,2,balanced,0.049733335773150124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,2,balanced,0.05101333558559418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,2,balanced,0.060602664947509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,2,balanced,0.7241386572519938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,2,balanced,0.08185600241025288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,2,balanced,0.09273067116737366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,64,balanced,2.389568010965983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,2,balanced,0.09476266304651897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,2,balanced,0.09546132882436116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,2,balanced,0.09526933232943217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,2,balanced,0.09713600079218547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,2,balanced,0.10045333703358968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,2,balanced,0.10353066523869832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,2,balanced,0.10429867108662923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,2,balanced,0.11243733763694763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,2,balanced,0.11962667107582092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,2,balanced,1.0849013328552246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,2,balanced,0.12615999579429626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,64,balanced,3.0358079274495444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,2,balanced,0.13051199913024902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,2,balanced,0.15620799859364828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,2,balanced,0.16873600085576376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,2,balanced,0.23947733640670776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,2,balanced,0.2658453385035197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,2,balanced,1.33681058883667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,2,balanced,0.40668801466623944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,2,balanced,0.44818135102589923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,2,power_law_1.01,0.04732159972190857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,2,balanced,0.6857439676920573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,2,power_law_1.01,0.05067520141601563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,2,power_law_1.01,0.05493760108947754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,2,power_law_1.01,0.0615231990814209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,2,power_law_1.01,0.06995840072631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,2,power_law_1.01,0.09152640104293823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,2,power_law_1.01,0.09587200284004212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,2,power_law_1.01,0.1013759970664978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,2,balanced,1.6291732788085938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,2,power_law_1.01,0.1027008056640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,2,balanced,0.8189333279927572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,2,power_law_1.01,0.10296319723129273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,2,power_law_1.01,0.10629760026931763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,2,power_law_1.01,0.1088703989982605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,2,power_law_1.01,0.11194239854812622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,2,power_law_1.01,0.11500799655914307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,2,power_law_1.01,0.12245759963989258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,2,power_law_1.01,0.13189120292663575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,2,power_law_1.01,0.13904639482498168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,2,power_law_1.01,0.16524159908294678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,2,balanced,1.2319893042246501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,2,power_law_1.01,0.19258240461349488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,2,power_law_1.01,0.24837119579315187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,2,power_law_1.01,0.2889408111572266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,2,power_law_1.01,0.3740607976913452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,2,power_law_1.01,0.4706687927246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,2,power_law_1.01,0.6472256183624268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,2,power_law_1.01,0.8431039810180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,2,power_law_1.01,1.1760319709777831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,2,power_law_1.01,1.395475196838379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,2,power_law_1.01,1.7865215301513673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,2,balanced,1.5464906692504883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,2,power_law_1.01,2.5900096893310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,2,balanced,2.4779253005981445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,2,power_law_1.01,5.1339263916015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,2,balanced,1.876746654510498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,2,power_law_1.2,0.0466048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,2,power_law_1.2,0.04690560102462769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,2,power_law_1.2,0.05256959795951843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,2,power_law_1.2,0.059411197900772095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,2,power_law_1.2,0.06484479904174804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,2,power_law_1.2,0.09011840224266052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,2,power_law_1.2,0.0969215989112854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,2,power_law_1.2,0.10062079429626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,2,power_law_1.2,0.10278400182723998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,2,power_law_1.2,0.10281599760055542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,2,balanced,2.929077466328939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,2,power_law_1.2,0.10823680162429809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,2,power_law_1.01,0.047712001204490664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,2,power_law_1.2,0.10958080291748047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,2,power_law_1.2,0.048179200291633605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,2,power_law_1.2,0.11208959817886352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,2,power_law_1.01,0.05002880096435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,2,power_law_1.2,0.11560959815979004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,2,power_law_1.2,0.04807040095329285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,2,power_law_1.01,0.05187839865684509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,2,power_law_1.2,0.05196160078048706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,2,power_law_1.2,0.12639360427856444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,2,power_law_1.01,0.059033602476119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,2,power_law_1.2,0.05930240154266357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,2,power_law_1.2,0.1340224027633667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,2,power_law_1.01,0.06650879979133606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,2,power_law_1.2,0.06603519916534424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,2,power_law_1.01,0.08289920091629029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,2,power_law_1.2,0.08227199912071229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,2,power_law_1.2,0.14069759845733643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,2,balanced,4.798768043518066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,2,power_law_1.01,0.08691200017929077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,2,power_law_1.2,0.08732159733772278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,2,power_law_1.01,0.08984320163726807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,2,power_law_1.2,0.09219200015068055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,2,power_law_1.2,0.16811519861221313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,2,power_law_1.2,0.09489920139312744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,2,power_law_1.01,0.09308800101280212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,2,power_law_1.2,0.09747200012207032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,2,power_law_1.2,0.1965376019477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,2,power_law_1.01,0.09657599925994872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,2,power_law_1.2,0.10204160213470459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,2,power_law_1.01,0.10152959823608398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,2,power_law_1.2,0.10723199844360351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,2,power_law_1.2,0.24231040477752686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,2,power_law_1.01,0.10610560178756714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,2,power_law_1.2,0.1117311954498291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,2,power_law_1.2,0.3075200080871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,2,power_law_1.2,0.11690880060195923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,2,power_law_1.01,0.10955519676208496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,2,power_law_1.2,0.12533119916915894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,2,power_law_1.2,0.3963903903961182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,2,power_law_1.01,0.11340800523757935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,2,power_law_1.2,0.1348736047744751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,2,power_law_1.2,0.14206720590591432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,2,power_law_1.01,0.1259071946144104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,2,power_law_1.2,0.4667776107788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,2,power_law_1.2,0.17356159687042236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,2,power_law_1.01,0.12908799648284913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,2,power_law_1.2,0.6527359962463379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,2,power_law_1.2,0.1962623953819275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,2,power_law_1.01,0.1374400019645691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,2,power_law_1.2,0.2546560049057007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,2,power_law_1.2,0.8867520332336426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,2,power_law_1.2,0.32327680587768554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,2,power_law_1.01,0.16762239933013917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,2,power_law_1.2,1.1588735580444336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,2,power_law_1.2,0.4268159866333008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,2,power_law_1.01,0.19525760412216187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,2,power_law_1.2,1.6324607849121093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,2,power_law_1.2,0.5290944099426269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,2,power_law_1.01,0.24106240272521973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,2,power_law_1.2,0.7214272022247314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,2,power_law_1.2,1.924051284790039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,2,power_law_1.01,0.3149888038635254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,2,power_law_1.2,0.9133760452270507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,2,power_law_1.01,0.40692481994628904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,2,power_law_1.2,1.3714240074157715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,2,power_law_1.2,2.720102310180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,2,power_law_1.01,0.5234752178192139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,2,power_law_1.2,1.7218368530273438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,2,power_law_1.01,0.7279295921325684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,2,power_law_1.2,2.119398307800293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,2,power_law_1.2,6.291219329833984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,2,power_law_1.01,0.9060864448547363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,2,power_law_1.2,3.344473648071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,2,power_law_1.01,1.2825663566589356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,2,power_law_1.01,1.6166015625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,2,power_law_1.2,6.303539276123047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,2,power_law_1.01,2.1209280014038088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,2,power_law_1.01,3.0901439666748045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,64,balanced,0.0642986645301183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,64,balanced,0.0621973325808843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,2,power_law_1.01,6.170675277709961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,64,balanced,0.06499733527501424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,64,balanced,0.06391466657320659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,64,balanced,0.07876266539096832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,64,power_law_1.01,0.16536320447921754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,64,balanced,0.09596799810727437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,64,balanced,0.12560000022252402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,2,balanced,5.767957051595052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,64,power_law_1.01,0.16272640228271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,64,balanced,0.11601066589355469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,64,balanced,0.11633599797884624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,64,power_law_1.01,0.1599295973777771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,64,balanced,0.11482133467992146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,64,power_law_1.01,0.10595200061798096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,64,balanced,0.11091199517250061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,64,power_law_1.01,0.1299008011817932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,64,balanced,0.11741333206494649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,64,power_law_1.01,0.13366400003433226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,64,balanced,0.1183519959449768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,64,power_law_1.01,0.14778879880905152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,64,balanced,0.11838933825492859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,64,power_law_1.01,0.16531200408935548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,64,balanced,0.12342400352160136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,64,power_law_1.01,0.15639040470123292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,64,balanced,0.12237333257993062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,64,power_law_1.01,0.15991040468215942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,64,power_law_1.01,0.14097280502319337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,64,balanced,0.13272533814112344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,64,power_law_1.01,0.16780799627304077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,64,power_law_1.01,0.16454399824142457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,64,balanced,0.135754664738973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,64,power_law_1.01,0.15919359922409057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,64,balanced,0.1474026640256246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,64,power_law_1.01,0.16992000341415406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,64,balanced,0.16012266278266907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,64,power_law_1.01,0.1728320002555847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,64,power_law_1.01,0.16529279947280884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,64,balanced,0.1713599960009257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,64,power_law_1.01,0.17580800056457518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,64,power_law_1.01,0.19400320053100586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,64,power_law_1.01,0.2154304027557373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,64,balanced,0.20035199324289957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,64,power_law_1.01,0.23362560272216798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,64,power_law_1.01,0.31895039081573484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,64,power_law_1.01,0.3582207918167114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,64,balanced,0.2321066657702128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,64,power_law_1.01,0.4696191787719727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,64,power_law_1.01,0.5760960102081298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,64,balanced,0.2938613295555115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,64,power_law_1.01,0.8293439865112304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,64,power_law_1.01,1.0751232147216796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,64,power_law_1.01,1.2887680053710937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,64,power_law_1.01,2.029964828491211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,64,balanced,0.40429333845774335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,64,power_law_1.01,0.1107200026512146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,64,power_law_1.01,4.261420822143554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,64,power_law_1.01,0.10968960523605346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,64,power_law_1.01,0.10726399421691894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,64,power_law_1.01,0.07529600262641907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,64,power_law_1.01,0.08921599984169007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,64,power_law_1.01,0.09395840167999267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,64,balanced,0.5085920095443726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,64,power_law_1.01,0.096806401014328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,64,power_law_1.01,0.11032320261001587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,64,power_law_1.01,0.10698239803314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,64,power_law_1.01,0.10616320371627808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,64,power_law_1.01,0.11448320150375366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,64,power_law_1.01,0.11687040328979492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,64,power_law_1.01,0.10728319883346557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,64,power_law_1.01,0.11696000099182129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,64,balanced,0.6620586713155111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,64,power_law_1.01,0.1200063943862915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,64,power_law_1.01,0.13017599582672118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,64,power_law_1.01,0.12896640300750734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,64,power_law_1.01,0.1406272053718567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,64,power_law_1.01,0.15829119682312012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,64,power_law_1.01,0.1828927993774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,64,power_law_1.01,0.21358079910278321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,64,power_law_1.01,0.2763391971588135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,64,power_law_1.01,0.33972480297088625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,64,power_law_1.01,0.44871039390563966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,64,power_law_1.01,0.57259521484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,64,balanced,0.8369973500569662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,64,power_law_1.01,0.8303487777709961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,64,power_law_1.01,1.0498047828674317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,64,power_law_1.01,1.3106559753417968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,64,power_law_1.01,1.988256072998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,64,power_law_1.01,4.05676155090332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,64,power_law_1.01,0.11143679618835449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,64,balanced,1.2644000053405762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,64,power_law_1.01,0.10766719579696656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,64,power_law_1.01,0.10647679567337036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,64,power_law_1.01,0.07749119997024537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,64,power_law_1.01,0.08478720188140869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,64,power_law_1.01,0.08877440094947815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,64,power_law_1.01,0.10088959932327271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,64,power_law_1.01,0.09829760193824769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,64,power_law_1.01,0.10540800094604492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,64,power_law_1.01,0.10612479448318482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,64,power_law_1.01,0.11089279651641845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,64,power_law_1.01,0.10744960308074951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,64,power_law_1.01,0.11299840211868287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,64,power_law_1.01,0.11411199569702149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,64,power_law_1.01,0.11952639818191528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,64,power_law_1.01,0.11818239688873292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,64,power_law_1.01,0.11941759586334229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,64,power_law_1.2,0.1676543951034546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,64,power_law_1.2,0.11303679943084717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,64,power_law_1.01,0.1275712013244629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,64,power_law_1.2,0.11016319990158081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,64,power_law_1.2,0.16401920318603516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,64,power_law_1.01,0.13616000413894652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,64,power_law_1.2,0.16122879981994628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,64,power_law_1.2,0.10933120250701904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,64,power_law_1.2,0.10866559743881225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,64,power_law_1.01,0.16926079988479614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,64,power_law_1.2,0.10398720502853394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,64,power_law_1.2,0.10536320209503174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,64,power_law_1.2,0.10680960416793824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,64,power_law_1.2,0.1174847960472107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,64,power_law_1.01,0.18661760091781615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,64,power_law_1.2,0.08234239816665649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,64,power_law_1.2,0.12159359455108643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,64,power_law_1.2,0.07681919932365418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,64,power_law_1.01,0.23814399242401124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,64,power_law_1.2,0.11987839937210083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,64,power_law_1.2,0.08336640000343323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,64,power_law_1.2,0.08785920143127442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,64,power_law_1.2,0.1586176037788391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,64,power_law_1.01,0.2903104066848755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,64,power_law_1.2,0.10029439926147461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,64,power_law_1.2,0.13779200315475465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,64,power_law_1.2,0.09223679900169372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,32,balanced,0.06462933123111725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,64,power_law_1.2,0.09496960043907166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,64,power_law_1.2,0.16191359758377075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,64,power_law_1.01,0.37091200351715087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,64,power_law_1.2,0.10058239698410035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,64,power_law_1.2,0.16104960441589355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,32,balanced,0.06658666829268138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,64,power_law_1.2,0.10782079696655274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,64,power_law_1.2,0.15277440547943116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,64,power_law_1.2,0.10641920566558838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,64,power_law_1.01,0.4905280113220215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,32,balanced,0.06612800061702728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,64,power_law_1.2,0.11002240180969239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,64,balanced,2.4061333338419595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,64,power_law_1.2,0.16046719551086425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,64,power_law_1.2,0.1578112006187439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,32,balanced,0.07419200241565704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,64,power_law_1.2,0.11126400232315063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,64,power_law_1.2,0.09857919812202454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,64,power_law_1.01,0.6671232223510742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,64,power_law_1.2,0.1738752007484436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,32,balanced,0.09702400366465251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,64,power_law_1.2,0.10897279977798462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,32,balanced,0.11944533387819926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,64,power_law_1.2,0.1701696038246155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,64,power_law_1.2,0.10581120252609252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,64,power_law_1.01,0.8416000366210937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,32,balanced,0.16473066806793213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,64,power_law_1.2,0.11503360271453858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,64,power_law_1.2,0.1683135986328125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,32,balanced,0.15558399756749472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,64,power_law_1.2,0.10718079805374145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,64,power_law_1.2,0.19256319999694824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,64,power_law_1.01,1.0358336448669434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,64,power_law_1.2,0.1115007996559143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,32,balanced,0.15690132975578308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,64,power_law_1.2,0.2020672082901001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,32,balanced,0.1607039968172709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,64,power_law_1.2,0.1095039963722229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,64,power_law_1.2,0.10996479988098144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,64,power_law_1.2,0.2324160099029541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,32,balanced,0.16641066471735635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,64,power_law_1.01,1.6912576675415039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,32,balanced,0.1684053341547648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,64,power_law_1.2,0.27429120540618895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,64,power_law_1.2,0.1124351978302002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,64,power_law_1.2,0.12491519451141357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,32,balanced,0.16481600205103555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,64,power_law_1.2,0.33578879833221437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,64,power_law_1.2,0.12792320251464845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,64,power_law_1.2,0.10095360279083251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,32,balanced,0.1641813317934672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,64,power_law_1.01,3.3032894134521484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,64,power_law_1.2,0.4157567977905273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,32,balanced,0.17238932847976685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,64,power_law_1.2,0.12661759853363036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,64,power_law_1.2,0.11898880004882813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,64,power_law_1.2,0.5340991973876953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,32,balanced,0.17208532492319742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,64,power_law_1.2,0.14767999649047853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,64,power_law_1.2,0.7339327812194825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,64,power_law_1.2,0.11772799491882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,32,balanced,0.17467733224232992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,64,power_law_1.2,0.15947519540786742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,64,power_law_1.2,1.0475135803222657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,64,power_law_1.2,1.4613759994506836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,64,power_law_1.2,0.11827839612960815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,64,power_law_1.2,0.18702080249786376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,32,balanced,0.18033599853515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,64,power_law_1.2,0.2231679916381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,64,power_law_1.2,1.6083711624145507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,64,power_law_1.2,0.1423616051673889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,32,balanced,0.18598934014638266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,64,power_law_1.2,0.29663360118865967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,64,power_law_1.2,0.1584447979927063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,64,power_law_1.2,2.762713623046875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,64,power_law_1.2,0.35771520137786866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,64,power_law_1.2,0.1829632043838501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,32,balanced,0.20376000801722208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,64,power_law_1.2,0.5207808017730713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,64,power_law_1.2,0.2027008056640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,64,power_law_1.2,6.167225646972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,32,balanced,0.05417599777380625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,64,power_law_1.2,0.2670144081115723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,64,power_law_1.2,0.6161791801452636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,32,balanced,0.21582933266957602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,64,power_law_1.2,0.33181440830230713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,32,balanced,0.05606399973233541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,64,power_law_1.2,0.9291584014892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,32,balanced,0.05563200016816457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,64,power_law_1.2,0.4510848045349121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,32,balanced,0.24634667237599692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,64,power_law_1.2,1.2521280288696288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,32,balanced,0.06428800026575725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,32,balanced,0.07470933099587758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,64,power_law_1.2,0.5412928104400635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,64,power_law_1.2,1.4944064140319824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,32,balanced,0.0869653324286143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,32,balanced,0.11274666587511699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,64,power_law_1.2,0.7687551975250244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,64,power_law_1.2,2.45034236907959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,32,balanced,0.11293333768844604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,32,balanced,0.27587733666102093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,64,power_law_1.2,1.0858559608459473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,32,balanced,0.10920533537864685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,32,balanced,0.11286933223406474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,64,power_law_1.2,1.283801555633545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,64,power_law_1.2,5.091263961791992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,32,balanced,0.11283199985822041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,32,balanced,0.1132586697737376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,64,power_law_1.2,2.164159965515137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,32,balanced,0.11429333686828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,32,balanced,0.12010133266448975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,32,balanced,0.33463998635609943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,64,power_law_1.2,4.045484924316407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,32,balanced,0.12441600362459819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,32,balanced,0.1272053321202596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,32,balanced,0.1339359978834788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,32,balanced,0.14219733079274496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,32,balanced,0.15355733036994934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,32,balanced,0.4310293197631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,32,balanced,0.17605332533518472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,32,balanced,0.19902400175730386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,32,balanced,0.24088533719380698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,32,balanced,0.532426675160726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,32,balanced,0.2886613408724467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,32,balanced,0.05724266668160757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,32,balanced,0.05774933099746704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,32,balanced,0.37460800011952716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,32,balanced,0.058117335041364036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,32,balanced,0.06446933249632518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,32,balanced,0.07955733438332875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,32,balanced,0.09920000036557515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,32,balanced,0.12504000465075174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,32,balanced,0.683199961980184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,32,balanced,0.11966400345166524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,32,balanced,0.11333333452542622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,32,balanced,0.48046398162841797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,32,balanced,0.11769066254297893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,32,balanced,0.11169067025184631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,32,balanced,0.11556800206502278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,32,balanced,0.116047998269399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,32,balanced,0.12350933750470479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,32,balanced,0.12138133247693379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,32,balanced,0.6497706572214762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,32,balanced,0.1263146698474884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,32,balanced,0.127893328666687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,32,balanced,0.8865119616190592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,32,balanced,0.1320693294207255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,32,balanced,0.14057599504788718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,32,balanced,0.1545866628487905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,32,balanced,0.17247466246287027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,32,balanced,0.8410240014394125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,32,balanced,0.2032853364944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,32,balanced,0.2357813318570455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,32,balanced,0.301749328772227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,32,balanced,1.3183039824167888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,32,balanced,1.0518133640289307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,32,balanced,0.4033279816309611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,32,balanced,0.5254026651382446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,32,balanced,1.6227572758992512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,32,balanced,0.6824053128560384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,32,power_law_1.01,0.11901439428329467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,32,power_law_1.01,0.16370559930801393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,32,power_law_1.01,0.1546175956726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,32,power_law_1.01,0.08929920196533203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,32,power_law_1.01,0.1033471941947937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,32,power_law_1.01,0.1123263955116272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,32,power_law_1.01,0.08944000005722046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,32,power_law_1.01,0.11575039625167846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,32,power_law_1.01,0.1448639988899231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,32,power_law_1.01,0.10682239532470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,32,power_law_1.01,0.1448639988899231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,32,power_law_1.01,0.1044543981552124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,32,power_law_1.01,0.13532160520553588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,32,balanced,0.8588533401489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,32,power_law_1.01,0.06692479848861695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,32,power_law_1.01,0.14747519493103028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,32,power_law_1.01,0.07522559762001038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,32,power_law_1.01,0.09592959880828858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,32,power_law_1.01,0.14882559776306153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,32,power_law_1.01,0.07968639731407165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,32,power_law_1.01,0.1555840015411377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,32,power_law_1.01,0.09301120042800903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,32,power_law_1.01,0.11457279920578003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,32,power_law_1.01,0.15470720529556276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,32,power_law_1.01,0.15943039655685426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,32,power_law_1.01,0.10151040554046631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,32,balanced,2.4694453875223794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,32,power_law_1.01,0.11283199787139893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,32,power_law_1.01,0.16113920211791993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,32,power_law_1.01,0.1019711971282959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,32,power_law_1.01,0.0729088008403778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,32,power_law_1.01,0.1662976026535034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,32,power_law_1.01,0.0961023986339569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,32,power_law_1.01,0.08215680122375488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,32,power_law_1.01,0.17574399709701538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,32,power_law_1.01,0.10344959497451782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,32,power_law_1.01,0.09454079866409301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,32,power_law_1.2,0.11712640523910522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,32,power_law_1.01,0.18622080087661744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,32,power_law_1.01,0.10722559690475464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,32,power_law_1.01,0.09260799884796142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,32,power_law_1.01,0.20424320697784423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,32,power_law_1.01,0.22928640842437745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,32,power_law_1.2,0.16325759887695312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,32,power_law_1.01,0.09830399751663207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,32,power_law_1.01,0.10379519462585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,32,power_law_1.01,0.29023358821868894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,32,power_law_1.2,0.09048960208892823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,32,power_law_1.01,0.10465919971466064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,32,power_law_1.01,0.11571199893951416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,32,power_law_1.2,0.1427199959754944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,32,power_law_1.01,0.10463999509811402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,32,power_law_1.01,0.11744639873504639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,32,power_law_1.01,0.3554176092147827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,32,power_law_1.2,0.08765439987182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,32,power_law_1.2,0.11037440299987793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,32,power_law_1.01,0.10627199411392212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,32,power_law_1.01,0.12433919906616211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,32,power_law_1.2,0.09836159944534302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,32,power_law_1.01,0.13012479543685912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,32,power_law_1.01,0.11505919694900513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,32,power_law_1.01,0.44121599197387695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,32,power_law_1.2,0.102566397190094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,32,power_law_1.2,0.07171840071678162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,32,power_law_1.01,0.13505280017852783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,32,power_law_1.01,0.54301438331604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,32,power_law_1.01,0.10503040552139283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,32,power_law_1.2,0.10744960308074951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,32,balanced,1.3116587003072102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,32,power_law_1.2,0.08090879917144775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,32,power_law_1.01,0.152019202709198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,32,balanced,3.1506773630777993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,32,power_law_1.01,0.7361408233642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,32,power_law_1.2,0.11397119760513305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,32,power_law_1.01,0.11190400123596192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,32,power_law_1.2,0.08357759714126586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,32,power_law_1.01,0.17820160388946532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,32,power_law_1.01,0.9594047546386719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,32,power_law_1.2,0.1323456048965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,32,power_law_1.01,0.2001215934753418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,32,power_law_1.01,0.12017279863357544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,32,power_law_1.01,1.1611200332641602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,32,power_law_1.2,0.08814719915390015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,32,power_law_1.01,0.12088960409164429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,32,power_law_1.2,0.1065343976020813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,32,power_law_1.2,0.13097599744796753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,32,power_law_1.01,0.2651648044586182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,32,power_law_1.01,1.686310386657715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,32,power_law_1.01,0.12313599586486816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,32,power_law_1.2,0.10424319505691529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,32,power_law_1.2,0.1498944044113159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,32,power_law_1.01,0.3226815938949585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,32,power_law_1.2,0.09744639992713929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,32,power_law_1.2,0.132096004486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,32,power_law_1.01,0.1286784052848816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,32,power_law_1.01,0.4390143871307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,32,power_law_1.01,3.34716796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,32,power_law_1.2,0.09972479939460754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,32,power_law_1.2,0.1540287971496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,32,power_law_1.01,0.14353920221328736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,32,power_law_1.01,0.5572991847991944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,32,power_law_1.2,0.10271999835968018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,32,power_law_1.2,0.1542400002479553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,32,power_law_1.01,0.16622719764709473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,32,power_law_1.01,0.7582719802856446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,32,power_law_1.2,0.14766720533370972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,32,power_law_1.2,0.10837119817733765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,32,power_law_1.01,0.18871040344238282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,32,power_law_1.2,0.16113920211791993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,32,power_law_1.2,0.11608320474624634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,32,power_law_1.01,1.0005632400512696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,32,power_law_1.01,0.23420159816741942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,32,power_law_1.2,0.15277440547943116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,32,power_law_1.2,0.11775360107421876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,32,power_law_1.01,0.28480000495910646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,32,power_law_1.01,1.2554112434387208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,32,power_law_1.2,0.1767040014266968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,32,power_law_1.2,0.12189439535140992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,32,power_law_1.2,0.17029759883880616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,32,power_law_1.01,0.37382400035858154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,32,power_law_1.2,0.12895359992980956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,32,power_law_1.01,1.9042240142822267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,32,power_law_1.2,0.18655359745025635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,32,power_law_1.01,0.4743360042572021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,32,power_law_1.2,0.14529919624328613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,32,power_law_1.2,0.2192512035369873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,32,power_law_1.01,0.6590976238250732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,32,power_law_1.2,0.15815039873123168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,32,power_law_1.01,3.97955207824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,32,power_law_1.2,0.24191360473632811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,32,power_law_1.2,0.18366719484329225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,32,power_law_1.01,0.841817569732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,32,power_law_1.2,0.22230401039123535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,32,power_law_1.2,0.32901759147644044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,32,power_law_1.01,1.058240032196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,32,power_law_1.2,0.2827136039733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,32,power_law_1.2,0.3787712097167969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,32,power_law_1.01,1.568607997894287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,32,power_law_1.2,0.34272639751434325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,32,power_law_1.2,0.48735361099243163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,32,power_law_1.2,0.4847360134124756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,32,power_law_1.01,3.219136047363281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,32,power_law_1.2,0.6069119930267334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,32,power_law_1.2,0.6263167858123779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,32,power_law_1.2,0.8928768157958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,32,power_law_1.2,0.8000384330749511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,32,power_law_1.2,1.1238719940185546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,32,power_law_1.2,1.1432127952575684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,32,power_law_1.2,1.3097536087036132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,32,power_law_1.2,1.354259204864502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,32,balanced,2.520357290903727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,32,power_law_1.2,2.3136064529418947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,32,power_law_1.2,2.105900764465332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,32,power_law_1.2,4.925222396850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,32,power_law_1.2,4.558502578735352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,32,power_law_1.2,0.09411839842796325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,32,power_law_1.2,0.11227519512176513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,32,power_law_1.2,0.10893440246582031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,32,power_law_1.2,0.06762239933013917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,32,power_law_1.2,0.08142079710960388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,32,power_law_1.2,0.08591359853744507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,32,power_law_1.2,0.0944703996181488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,32,power_law_1.2,0.09745280146598816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,32,power_law_1.2,0.09518719911575317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,32,power_law_1.2,0.11063679456710815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,32,power_law_1.2,0.10952960252761841
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,32,power_law_1.2,0.10696959495544434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,32,power_law_1.2,0.10046080350875855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,32,power_law_1.2,0.10976639986038209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,32,power_law_1.2,0.10935039520263672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,32,power_law_1.2,0.11950080394744873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,32,power_law_1.2,0.12229119539260865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,32,power_law_1.2,0.13589760065078735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,32,power_law_1.2,0.1473471999168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,32,power_law_1.2,0.17141760587692262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,32,power_law_1.2,0.20348799228668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,32,power_law_1.2,0.25535359382629397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,32,power_law_1.2,0.3077824115753174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,32,power_law_1.2,0.41890559196472166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,32,power_law_1.2,0.5306111812591553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,32,power_law_1.2,0.7599103927612305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,32,power_law_1.2,0.9411775588989257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,32,power_law_1.2,1.1697216033935547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,32,power_law_1.2,1.9641279220581054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,32,power_law_1.2,3.8779457092285154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,16,balanced,0.05572799841562907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,16,balanced,0.056048000852266945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,16,balanced,0.054917335510253906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,16,balanced,0.05637866755326589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,16,balanced,0.060229331254959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,16,balanced,0.07046400010585785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,16,balanced,0.08874133229255676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,16,balanced,0.09119466940561931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,16,balanced,0.0918880005677541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,16,balanced,0.09187199672063191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,16,balanced,0.0940053363641103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,16,balanced,0.09288000067075093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,16,balanced,0.09589333335558574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,16,power_law_1.01,0.07713279724121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,16,balanced,0.0983733336130778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,16,balanced,0.1042133371035258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,16,power_law_1.01,0.07535359859466553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,16,balanced,0.10318932930628459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,16,power_law_1.01,0.08684800267219543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,16,balanced,0.11038399736086528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,16,power_law_1.01,0.059539198875427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,16,balanced,0.11995733777681987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,16,power_law_1.01,0.06339200139045716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,16,balanced,0.12781332929929098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,16,power_law_1.01,0.06812800168991089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,16,power_law_1.01,0.06967679858207702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,16,balanced,0.14864533146222433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,16,power_law_1.01,0.08028799891471863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,16,power_law_1.01,0.07747200131416321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,16,balanced,0.16364266475041708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,16,power_law_1.01,0.08153600096702576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,16,power_law_1.01,0.08437119722366333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,16,balanced,0.1999359925587972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,16,power_law_1.01,0.08477439880371093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,16,power_law_1.01,0.09063040018081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,16,power_law_1.01,0.09098240137100219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,16,power_law_1.01,0.09919360280036926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,16,balanced,0.23660266399383545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,16,power_law_1.01,0.10014079809188843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,16,power_law_1.01,0.10769920349121094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,16,power_law_1.01,0.12301440238952636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,16,power_law_1.01,0.13375999927520751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,16,balanced,0.30311999718348187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,16,power_law_1.01,0.15936000347137452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,16,power_law_1.01,0.1820032000541687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,16,power_law_1.01,0.24250879287719726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,16,power_law_1.01,0.27209599018096925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,16,power_law_1.01,0.3648576021194458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,16,balanced,0.39188265800476074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,16,power_law_1.01,0.4579904079437256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,16,power_law_1.01,0.639359998703003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,16,power_law_1.01,0.8188544273376465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,16,power_law_1.01,1.0522047996520996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,16,power_law_1.01,1.5167424201965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,16,power_law_1.01,3.1184127807617186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,16,balanced,0.5267413457234701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,16,balanced,0.6788372993469238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,16,balanced,0.8369440237681071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,16,balanced,1.2797760168711345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,16,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,16,balanced,0.044138665000597634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,16,balanced,0.04721599817276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,16,balanced,0.05532800157864889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,16,balanced,0.07008000214894612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,16,balanced,0.09874133268992107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,16,balanced,0.13010666767756143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,16,balanced,0.12379200259844463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,16,balanced,0.12542933225631714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,16,balanced,0.12532800436019897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,16,balanced,0.04740266501903534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,16,balanced,0.12706133723258972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,16,balanced,0.12717866897583008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,16,balanced,0.12972799936930338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,16,balanced,0.047770669062932335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,16,balanced,0.04595733185609182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,16,balanced,0.13147733608881632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,16,balanced,0.050053333242734276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,16,balanced,0.13823466499646506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,16,balanced,0.058101331194241844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,16,balanced,0.13913066188494363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,16,balanced,0.07821333408355713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,16,balanced,0.0946613351504008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,16,balanced,0.14500799775123596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,16,balanced,0.09472533067067464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,16,balanced,0.15110400319099426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,16,balanced,0.0953653355439504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,16,balanced,0.09498666723569234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,16,balanced,2.4989120165506997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,16,balanced,0.1572213371594747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,16,balanced,0.09900800387064616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,16,balanced,0.1014453371365865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,16,balanced,0.17759466171264648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,16,balanced,0.10016000270843506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,16,balanced,0.10457066694895427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,16,balanced,0.18850666284561157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,16,balanced,0.11012267072995503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,16,balanced,0.11174933115641277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,16,balanced,0.12160000205039978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,16,balanced,0.2217280069986979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,16,balanced,0.1332373321056366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,16,balanced,0.1431893308957418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,16,balanced,0.25918932755788165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,16,balanced,0.16874132553736368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,16,power_law_1.01,0.06988800168037415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,16,power_law_1.01,0.061286401748657224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,16,balanced,0.1892426609992981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,16,balanced,0.31861867507298786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,16,power_law_1.01,0.09679359793663025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,16,power_law_1.2,0.059987199306488034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,16,power_law_1.01,0.0815999984741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,16,balanced,0.23568532864252725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,16,power_law_1.2,0.06885120272636414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,16,power_law_1.01,0.12695679664611817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,16,power_law_1.01,0.08820480108261108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,16,power_law_1.2,0.07070080041885377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,16,power_law_1.01,0.061638402938842776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,16,power_law_1.2,0.09644160270690919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,16,power_law_1.01,0.05433599948883057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,16,power_law_1.2,0.06892160177230836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,16,power_law_1.01,0.0723136007785797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,16,power_law_1.2,0.08543999791145325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,16,balanced,0.3922773202260335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,16,power_law_1.01,0.05782399773597717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,16,balanced,0.28593067328135174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,16,power_law_1.01,0.0881663978099823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,16,power_law_1.2,0.05787519812583923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,16,power_law_1.2,0.058796799182891844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,16,power_law_1.01,0.07060480117797852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,16,power_law_1.01,0.09305599927902222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,16,power_law_1.2,0.07221119999885559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,16,power_law_1.2,0.06149119734764099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,16,power_law_1.01,0.07412480115890503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,16,power_law_1.01,0.10503040552139283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,16,power_law_1.2,0.08380799889564514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,16,power_law_1.2,0.06696320176124573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,16,power_law_1.01,0.0804032027721405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,16,power_law_1.01,0.1035264015197754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,16,balanced,0.3758773406346639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,16,power_law_1.01,0.08332160115242004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,16,power_law_1.2,0.08201599717140198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,16,power_law_1.2,0.0676800012588501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,16,balanced,0.5196906725565592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,16,power_law_1.01,0.11143679618835449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,16,power_law_1.01,0.0840511977672577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,16,power_law_1.2,0.101363205909729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,16,power_law_1.2,0.07788159847259521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,16,balanced,0.47462932268778485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,16,power_law_1.01,0.11130880117416382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,16,power_law_1.2,0.10287359952926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,16,power_law_1.01,0.08485760092735291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,16,power_law_1.01,0.11542400121688842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,16,power_law_1.2,0.10623999834060668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,16,power_law_1.01,0.0865664005279541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,16,power_law_1.2,0.07894399762153625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,16,power_law_1.01,0.11796480417251587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,16,power_law_1.2,0.10855040550231934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,16,power_law_1.01,0.0926144003868103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,16,power_law_1.01,0.11829760074615478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,16,power_law_1.2,0.07957760095596314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,16,power_law_1.2,0.11580159664154052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,16,power_law_1.01,0.09477760195732117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,16,power_law_1.01,0.12223360538482667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,16,balanced,0.6524639924367269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,16,power_law_1.01,0.10581120252609252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,16,power_law_1.01,0.1336192011833191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,16,power_law_1.2,0.11552000045776367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,16,power_law_1.2,0.08007680177688599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,16,balanced,0.6590613524119059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,16,power_law_1.01,0.10754560232162476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,16,power_law_1.01,0.1325376033782959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,16,power_law_1.2,0.12014720439910889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,16,power_law_1.2,0.08395519852638245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,16,power_law_1.01,0.11712000370025635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,16,power_law_1.01,0.14865920543670655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,16,power_law_1.2,0.12572159767150878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,16,power_law_1.2,0.09168000221252441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,16,power_law_1.01,0.13568639755249023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,16,power_law_1.01,0.1603327989578247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,16,power_law_1.2,0.13370879888534545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,16,power_law_1.2,0.09067519903182983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,16,power_law_1.01,0.1414463996887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,16,power_law_1.01,0.1931071996688843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,16,power_law_1.2,0.14035199880599974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,16,power_law_1.2,0.09895039796829223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,16,power_law_1.01,0.17456640005111695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,16,power_law_1.01,0.2110527992248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,16,power_law_1.2,0.154310405254364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,16,power_law_1.2,0.10379519462585449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,16,power_law_1.01,0.19623680114746095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,16,power_law_1.01,0.26606719493865966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,16,power_law_1.2,0.1648959994316101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,16,power_law_1.2,0.11136000156402588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,16,power_law_1.01,0.25816960334777833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,16,power_law_1.01,0.30750079154968263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,16,power_law_1.2,0.2067647933959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,16,power_law_1.2,0.12494720220565796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,16,power_law_1.01,0.31393918991088865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,16,power_law_1.01,0.4102464199066162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,16,power_law_1.2,0.21086080074310304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,16,power_law_1.2,0.13572479486465455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,16,power_law_1.01,0.4235392093658447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,16,power_law_1.2,0.2950079917907715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,16,power_law_1.2,0.16844799518585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,16,power_law_1.01,0.4824512004852295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,16,balanced,0.8481600284576416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,16,power_law_1.01,0.5555327892303467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,16,power_law_1.2,0.18792320489883424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,16,power_law_1.2,0.33683199882507325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,16,balanced,0.8097546895345052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,16,power_law_1.01,0.7946432113647461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,16,power_law_1.2,0.2520384073257446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,16,power_law_1.2,0.43985919952392577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,16,power_law_1.01,0.6740287780761719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,16,power_law_1.2,0.05738880038261414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,16,power_law_1.2,0.2930943965911865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,16,power_law_1.01,1.0008319854736327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,16,power_law_1.2,0.5207104206085205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,16,power_law_1.2,0.39502079486846925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,16,power_law_1.2,0.10243200063705445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,16,power_law_1.01,0.7966400146484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,16,power_law_1.2,0.7813951969146729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,16,power_law_1.01,1.2342016220092773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,16,power_law_1.2,0.5162176132202149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,16,power_law_1.2,0.06379520297050476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,16,power_law_1.2,0.9518976211547852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,16,power_law_1.01,1.9086847305297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,16,power_law_1.2,0.7121600151062012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,16,power_law_1.01,0.9621696472167969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,16,power_law_1.2,0.053465598821640016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,16,power_law_1.2,1.0749695777893067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,16,power_law_1.2,0.9018367767333985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,16,power_law_1.2,0.0568448007106781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,16,power_law_1.01,3.718841552734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,16,power_law_1.01,1.6099071502685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,16,power_law_1.2,0.0698303997516632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,16,power_law_1.2,1.8486656188964843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,16,power_law_1.2,1.1873279571533204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,16,power_law_1.2,0.07182719707489013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,16,power_law_1.2,1.8835391998291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,16,power_law_1.2,0.07975680232048035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,16,power_law_1.2,3.4475521087646483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,16,power_law_1.01,2.8808000564575194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,16,power_law_1.2,0.08295040130615235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,16,balanced,1.0464800198872883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,16,power_law_1.2,3.544550323486328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,16,power_law_1.2,0.08391039967536926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,16,power_law_1.2,0.0880832016468048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,16,power_law_1.2,0.09208319783210754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,16,power_law_1.2,0.0911296010017395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,16,power_law_1.2,0.09690240025520325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,16,power_law_1.2,0.10645760297775268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,16,balanced,1.1737653414408367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,16,power_law_1.2,0.10757759809494019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,16,power_law_1.2,0.11953279972076417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,16,power_law_1.2,0.1411520004272461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,16,power_law_1.2,0.15041279792785645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,16,power_law_1.2,0.18494720458984376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,16,power_law_1.2,0.2080064058303833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,16,power_law_1.2,0.2721856117248535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,16,balanced,1.5882612864176433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,16,power_law_1.2,0.32984960079193115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,16,power_law_1.2,0.4847231864929199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,16,power_law_1.2,0.6244927883148194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,16,power_law_1.2,0.8599231719970704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,16,power_law_1.2,1.0641152381896972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,16,power_law_1.2,1.3224063873291017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,16,power_law_1.2,2.143168067932129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,16,power_law_1.2,3.9580928802490236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,16,balanced,2.27894926071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,16,balanced,3.1070292790730796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,8,balanced,0.049829334020614624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,8,balanced,0.05186133086681366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,8,balanced,0.05622399846712748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,8,balanced,0.056159997979799904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,8,balanced,0.06483200192451477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,8,balanced,0.08480000495910645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,8,balanced,0.10593066612879436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,8,balanced,0.10662933190663655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,8,balanced,0.10718400279680888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,8,balanced,0.10694400469462077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,8,balanced,0.10959999759991963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,8,balanced,0.11157332857449849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,8,balanced,0.11107200384140015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,8,balanced,0.11458133657773335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,8,balanced,0.12152533729871114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,8,balanced,0.12314666310946147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,8,balanced,0.13334932923316956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,8,balanced,0.1422879993915558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,8,balanced,0.15009066462516785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,8,balanced,0.17126399278640747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,8,balanced,0.19009600083033243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,8,balanced,0.23240532477696738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,8,power_law_1.01,0.055692797899246214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,8,balanced,0.2797706723213196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,8,power_law_1.01,0.07926399707794189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,8,power_law_1.01,0.07873280048370361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,8,power_law_1.01,0.06042240262031555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,8,power_law_1.01,0.06807680130004883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,8,power_law_1.01,0.07504640221595764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,8,balanced,0.368127981821696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,8,power_law_1.01,0.07999359965324401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,8,power_law_1.01,0.08848000168800355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,8,power_law_1.01,0.08772479891777038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,8,power_law_1.01,0.09203839898109437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,8,power_law_1.01,0.09638400077819824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,8,balanced,0.4913546641667684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,8,power_law_1.01,0.0997376024723053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,8,power_law_1.01,0.10364160537719727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,8,power_law_1.01,0.10291839838027954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,8,power_law_1.01,0.11275520324707031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,8,power_law_1.01,0.11745280027389526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,8,power_law_1.01,0.12830079793930055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,8,power_law_1.01,0.14718079566955566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,8,power_law_1.01,0.16293120384216309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,8,power_law_1.01,0.19070719480514525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,8,balanced,0.6582613388697306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,8,power_law_1.01,0.2002880096435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,8,power_law_1.01,0.25248639583587645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,8,power_law_1.01,0.3040191888809204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,8,power_law_1.01,0.40903677940368655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,8,power_law_1.01,0.5711552143096924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,8,power_law_1.01,0.772166395187378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,8,power_law_1.01,1.0053888320922852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,8,power_law_1.01,1.2151103973388673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,8,power_law_1.01,1.8374080657958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,8,balanced,0.8598666985829672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,8,power_law_1.01,3.8997566223144533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,8,balanced,0.0481333335240682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,8,balanced,0.04866666595141093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,8,balanced,0.05004266897837321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,8,balanced,0.0562720000743866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,8,balanced,0.07625600198904674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,8,balanced,0.10518933335940044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,8,balanced,0.13664533694585165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,8,balanced,0.13190933068593344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,8,balanced,0.13593600193659464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,8,balanced,0.1341813306013743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,8,balanced,1.0648799737294514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,8,balanced,0.13476266463597616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,8,balanced,0.1381173332532247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,8,balanced,0.1384213368097941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,8,balanced,0.13805333773295084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,8,balanced,0.14537599682807922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,8,balanced,0.1471733351548513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,8,balanced,0.15482133626937866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,8,balanced,0.16546666622161865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,8,balanced,0.17482666174570718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,8,balanced,0.194922665754954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,8,balanced,0.2145706613858541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,8,balanced,0.2540213267008464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,8,balanced,0.303872009118398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,8,balanced,1.6406559944152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,8,balanced,0.38060800234476727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,8,balanced,0.4759840170542399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,8,balanced,0.6408693393071493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,8,balanced,0.8019999663035074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,8,balanced,0.9996639887491862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,8,balanced,3.2085227966308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,8,power_law_1.01,0.054054397344589236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,8,power_law_1.01,0.07125759720802308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,8,power_law_1.01,0.09829760193824769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,8,power_law_1.01,0.0584447979927063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,8,power_law_1.01,0.06740480065345764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,8,power_law_1.01,0.08781440258026123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,8,balanced,1.4801759719848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,8,power_law_1.01,0.09080960154533387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,8,balanced,0.04598933458328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,8,power_law_1.01,0.10362240076065063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,8,power_law_1.01,0.10321919918060303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,8,balanced,0.0467199981212616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,8,balanced,0.04781866570313772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,8,power_law_1.01,0.11020159721374512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,8,balanced,0.05087999999523163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,8,power_law_1.01,0.11121920347213746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,8,balanced,0.059845333298047386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,8,balanced,0.08249066770076752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,8,power_law_1.01,0.11813119649887086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,8,balanced,0.1030453344186147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,8,power_law_1.01,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,8,power_law_1.01,0.12405760288238525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,8,balanced,0.099973330895106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,8,power_law_1.01,0.1250815987586975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,8,balanced,0.10105066498120625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,8,power_law_1.01,0.057625597715377806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,8,balanced,0.10227200388908386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,8,power_law_1.01,0.13275519609451295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,8,power_law_1.01,0.07677440047264099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,8,balanced,0.1050933301448822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,8,power_law_1.01,0.13790719509124755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,8,power_law_1.01,0.053420799970626834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,8,balanced,0.10712533195813496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,8,power_law_1.01,0.061286401748657224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,8,power_law_1.01,0.1446272015571594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,8,balanced,0.1083679993947347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,8,power_law_1.01,0.06835839748382569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,8,balanced,0.11115733782450359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,8,power_law_1.01,0.16698240041732787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,8,power_law_1.01,0.07551360130310059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,8,balanced,0.11858133474985759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,8,power_law_1.01,0.16917760372161866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,8,power_law_1.01,0.08577920198440551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,8,balanced,0.11989866693814595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,8,power_law_1.01,0.2060096025466919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,8,power_law_1.01,0.08887040019035339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,8,balanced,0.12828266620635986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,8,power_law_1.01,0.22613120079040527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,8,power_law_1.2,0.053472000360488894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,8,power_law_1.01,0.08921599984169007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,8,balanced,0.144896000623703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,8,power_law_1.01,0.28268160820007326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,8,power_law_1.01,0.09220479726791382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,8,power_law_1.2,0.060249602794647215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,8,balanced,0.15661333004633585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,8,power_law_1.01,0.3589184045791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,8,power_law_1.01,0.09594240188598632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,8,power_law_1.2,0.05766400098800659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,8,power_law_1.01,0.09937919974327088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,8,power_law_1.01,0.44688639640808103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,8,power_law_1.2,0.059680002927780154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,8,balanced,0.18665599822998047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,8,power_law_1.01,0.10293760299682617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,8,power_law_1.01,0.5857024192810059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,8,power_law_1.2,0.06492159962654113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,8,power_law_1.01,0.11439360380172729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,8,balanced,0.21024533112843832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,8,power_law_1.01,0.7957888126373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,8,power_law_1.2,0.07022719979286193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,8,power_law_1.01,0.11811840534210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,8,power_law_1.01,0.9755904197692871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,8,power_law_1.2,0.07477759718894958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,8,power_law_1.01,0.13415039777755738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,8,power_law_1.01,1.1312191963195801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,8,power_law_1.2,0.08765439987182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,8,power_law_1.01,0.1542464017868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,8,balanced,0.26072533925374347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,8,power_law_1.2,0.08737279772758484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,8,power_law_1.01,0.15919359922409057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,8,power_law_1.01,1.8071935653686524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,8,power_law_1.2,0.08773760199546814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,8,power_law_1.01,0.19152640104293822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,8,power_law_1.01,3.730112075805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,8,power_law_1.2,0.09276800155639649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,8,power_law_1.01,0.22362880706787108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,8,balanced,0.33130133152008057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,8,power_law_1.2,0.09830399751663207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,8,power_law_1.01,0.29649279117584226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,8,power_law_1.2,0.10087039470672607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,8,power_law_1.01,0.3669183969497681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,8,power_law_1.2,0.05308799743652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,8,power_law_1.2,0.10373120307922364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,8,power_law_1.01,0.5159103870391846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,8,power_law_1.2,0.11137919425964356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,8,power_law_1.01,0.6353087902069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,8,power_law_1.2,0.06982399821281433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,8,balanced,0.4339413245519002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,8,power_law_1.2,0.11910400390625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,8,power_law_1.01,0.885638427734375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,8,power_law_1.2,0.06589440107345582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,8,power_law_1.2,0.13235199451446533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,8,power_law_1.2,0.05541759729385376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,8,power_law_1.01,1.1675583839416503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,8,power_law_1.2,0.06835200190544129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,8,power_law_1.2,0.14859520196914672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,8,power_law_1.01,1.3934720039367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,8,power_law_1.2,0.08205440044403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,8,power_law_1.2,0.15155839920043945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,8,power_law_1.2,0.08510079979896545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,8,balanced,0.5582506656646729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,8,power_law_1.01,2.2498111724853516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,8,balanced,2.86735471089681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,8,power_law_1.2,0.17861759662628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,8,power_law_1.2,0.0991807997226715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,8,power_law_1.01,4.342764663696289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,8,power_law_1.2,0.20740480422973634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,8,power_law_1.2,0.10784000158309937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,8,power_law_1.2,0.26895999908447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,8,power_law_1.2,0.1063423991203308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,8,balanced,0.7796320120493571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,8,power_law_1.2,0.05287039875984192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,8,power_law_1.2,0.32882559299468994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,8,power_law_1.2,0.11268479824066162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,8,power_law_1.2,0.06554239988327026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,8,power_law_1.2,0.11273599863052368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,8,power_law_1.2,0.47036161422729494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,8,power_law_1.2,0.0554751992225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,8,power_law_1.2,0.11747839450836181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,8,power_law_1.2,0.05418239831924439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,8,power_law_1.2,0.5923136234283447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,8,balanced,1.0004533131917317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,8,power_law_1.2,0.06301440000534057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,8,power_law_1.2,0.11756800413131714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,8,power_law_1.2,0.8208767890930175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,8,power_law_1.2,0.06944640278816223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,8,power_law_1.2,0.1307263970375061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,8,power_law_1.2,0.07342079877853394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,8,power_law_1.2,0.13589119911193848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,8,power_law_1.2,1.1604031562805175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,8,power_law_1.2,0.0825984001159668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,8,power_law_1.2,0.14601600170135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,8,power_law_1.2,0.0853056013584137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,8,power_law_1.2,1.3529088020324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,8,power_law_1.2,0.16766079664230346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,8,power_law_1.2,0.18935680389404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,8,power_law_1.2,0.08528640270233154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,8,power_law_1.2,2.1633792877197267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,8,power_law_1.2,0.22003839015960694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,8,power_law_1.2,0.09228159785270691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,8,power_law_1.2,0.23658881187438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,8,power_law_1.2,0.09722880125045777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,8,power_law_1.2,4.448819351196289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,8,power_law_1.2,0.31103360652923584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,8,power_law_1.2,0.0975488007068634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,8,power_law_1.2,0.37275519371032717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,8,power_law_1.2,0.10723199844360351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,8,balanced,1.2419733206431072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,8,power_law_1.2,0.11622400283813476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,8,power_law_1.2,0.4838399887084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,8,power_law_1.2,0.12067199945449829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,8,power_law_1.2,0.6247424125671387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,8,power_law_1.2,0.1347264051437378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,8,power_law_1.2,0.9084927558898925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,8,power_law_1.2,0.15838079452514647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,8,power_law_1.2,1.0230143547058106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,8,power_law_1.2,0.16447999477386474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,8,power_law_1.2,1.2905407905578614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,8,power_law_1.2,0.20778241157531738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,8,power_law_1.2,0.24021120071411134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,8,power_law_1.2,1.9078527450561524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,8,power_law_1.2,0.3050112009048462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,8,power_law_1.2,0.38530559539794923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,8,power_law_1.2,4.003257751464844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,8,power_law_1.2,0.5498432159423828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,8,power_law_1.2,0.696895980834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,8,power_law_1.2,0.9968447685241699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,8,power_law_1.2,1.2504128456115722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,8,balanced,1.9049919446309407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,8,power_law_1.2,1.4907135963439941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,8,power_law_1.2,2.4629375457763674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,8,power_law_1.2,4.919443130493164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,8,balanced,3.7244532903035483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,4,balanced,0.06234666705131531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,4,balanced,0.06307733555634816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,4,balanced,0.06634133557478587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,4,balanced,0.07226666808128357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,4,balanced,0.08340266346931458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,4,balanced,0.11198932925860088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,4,balanced,0.14410666624704996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,4,balanced,0.14613866806030273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,4,balanced,0.14633599917093912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,4,balanced,0.1464959979057312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,4,balanced,0.14380266269048056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,4,balanced,0.14619732896486917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,4,balanced,0.14891733725865683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,4,balanced,0.15196266770362854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,4,balanced,0.1601706643899282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,4,balanced,0.1609119971593221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,4,balanced,0.1692319909731547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,4,balanced,0.18382932742436728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,4,balanced,0.1967359979947408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,4,balanced,0.22452799479166666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,4,balanced,0.25281065702438354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,4,balanced,0.31310399373372394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,4,balanced,0.3782293399175008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,4,balanced,0.5061599810918173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,4,balanced,0.6810293197631836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,4,balanced,0.9341440200805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,4,power_law_1.01,0.061267197132110596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,4,power_law_1.01,0.0664255976676941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,4,power_law_1.01,0.0831167995929718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,4,power_law_1.01,0.06675840020179749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,4,power_law_1.01,0.08043519854545593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,4,power_law_1.01,0.09137279987335205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,4,balanced,1.2449066638946533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,4,power_law_1.01,0.10126719474792481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,4,power_law_1.01,0.11511039733886719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,4,power_law_1.01,0.1180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,4,power_law_1.01,0.12020479440689087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,4,power_law_1.01,0.12090879678726196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,4,power_law_1.01,0.12789119482040406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,4,power_law_1.01,0.12937599420547485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,4,power_law_1.01,0.13432320356369018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,4,power_law_1.01,0.14398080110549927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,4,power_law_1.01,0.14775680303573607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,4,power_law_1.01,0.15742720365524293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,4,power_law_1.01,0.1733888030052185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,4,power_law_1.01,0.19447679519653321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,4,power_law_1.01,0.2275775909423828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,4,balanced,1.5418027242024739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,4,power_law_1.01,0.26641919612884524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,4,power_law_1.01,0.34656639099121095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,4,power_law_1.01,0.4039487838745117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,4,power_law_1.01,0.5738687992095948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,4,power_law_1.01,0.7501120090484619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,4,power_law_1.01,1.0543935775756836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,4,power_law_1.01,1.3448960304260253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,4,power_law_1.01,1.6436159133911132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,4,power_law_1.01,2.714553642272949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,4,power_law_1.01,5.002048110961914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,4,balanced,2.377946694691976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,4,balanced,0.043951998154322304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,4,balanced,0.04828266799449921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,4,balanced,0.05153066913286845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,4,balanced,0.05788800120353699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,4,balanced,0.07941866914431255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,4,balanced,0.11349333326021831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,4,balanced,0.14601066708564758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,4,balanced,0.14362133542696634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,4,balanced,0.14523200194040933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,4,balanced,0.14617066582043967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,4,balanced,0.1455626686414083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,4,balanced,0.14991999665896097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,4,balanced,0.14805866281191507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,4,balanced,0.15272000432014465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,4,balanced,0.16103466351826987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,4,balanced,4.588405291239421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,4,balanced,0.16406933466593424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,4,balanced,0.17186667521794638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,4,balanced,0.18717332681020102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,4,balanced,0.20595200856526694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,4,balanced,0.23391467332839966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,4,balanced,0.25969066222508747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,4,balanced,0.31426666180292767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,4,balanced,0.38898666699727374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,4,balanced,0.49484264850616455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,4,balanced,0.6397866805394491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,4,balanced,0.8564746379852295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,4,balanced,0.04602666695912679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,4,balanced,0.04993600149949392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,4,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,4,balanced,0.05596800148487091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,4,balanced,0.06473599870999654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,4,balanced,0.09403733412424724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,4,balanced,0.11931733290354411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,4,balanced,0.11525332927703857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,4,balanced,0.11691733201344807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,4,balanced,0.11659199992815654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,4,balanced,0.11620266238848369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,4,balanced,0.1226026713848114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,4,balanced,1.119167963663737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,4,balanced,0.12330666184425354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,4,balanced,0.12610666950543722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,4,balanced,0.13428800304730734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,4,balanced,0.13776000340779623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,4,balanced,0.15077867110570273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,4,balanced,0.1686240037282308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,4,balanced,0.18306666612625122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,4,balanced,0.2169333299001058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,4,power_law_1.01,0.05125759840011597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,4,power_law_1.01,0.0531391978263855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,4,balanced,0.24939733743667603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,4,power_law_1.01,0.07710080146789551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,4,power_law_1.01,0.0575872004032135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,4,balanced,0.3193440039952596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,4,power_law_1.01,0.07088000178337098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,4,power_law_1.01,0.08647680282592773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,4,balanced,1.4087306658426921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,4,power_law_1.01,0.09873279929161072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,4,balanced,0.39981865882873535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,4,power_law_1.2,0.05797759890556335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,4,power_law_1.01,0.11240320205688477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,4,power_law_1.01,0.11514879465103149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,4,power_law_1.2,0.06386560201644897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,4,power_law_1.01,0.11808639764785767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,4,power_law_1.2,0.06324480175971985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,4,power_law_1.01,0.1241919994354248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,4,power_law_1.2,0.06804479956626892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,4,power_law_1.01,0.1308608055114746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,4,balanced,0.5370933214823405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,4,power_law_1.2,0.07870079874992371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,4,power_law_1.01,0.13471360206604005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,4,power_law_1.2,0.0873856008052826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,4,power_law_1.01,0.14032000303268433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,4,power_law_1.2,0.09622399806976319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,4,power_law_1.01,0.14700160026550294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,4,power_law_1.2,0.1098688006401062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,4,power_law_1.01,0.1537600040435791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,4,power_law_1.2,0.11077760457992554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,4,power_law_1.01,0.16090879440307618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,4,power_law_1.2,0.11763839721679688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,4,balanced,0.7178026835123698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,4,power_law_1.01,0.17750400304794312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,4,power_law_1.2,0.11877119541168213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,4,power_law_1.01,0.2055743932723999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,4,power_law_1.2,0.12616319656372071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,4,power_law_1.01,0.24223361015319825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,4,power_law_1.2,0.12908799648284913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,4,power_law_1.01,0.2775424003601074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,4,power_law_1.2,0.13435519933700563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,4,power_law_1.01,0.35520000457763673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,4,power_law_1.2,0.1438591957092285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,4,power_law_1.01,0.4427072048187256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,4,power_law_1.01,0.04894720017910004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,4,power_law_1.2,0.1519487977027893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,4,power_law_1.01,0.5766335964202881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,4,power_law_1.2,0.15897599458694459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,4,power_law_1.01,0.05120000243186951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,4,power_law_1.01,0.7233791828155518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,4,power_law_1.01,0.06408320069313049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,4,power_law_1.2,0.1761023998260498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,4,balanced,1.0140266418457031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,4,power_law_1.01,1.0345791816711425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,4,balanced,2.086437384287516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,4,power_law_1.01,0.05735039710998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,4,power_law_1.2,0.1947711944580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,4,power_law_1.01,1.4261247634887695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,4,power_law_1.01,0.06684799790382386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,4,power_law_1.2,0.2371392011642456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,4,power_law_1.01,1.5709759712219238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,4,power_law_1.01,0.07456640005111695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,4,power_law_1.2,0.27395200729370117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,4,power_law_1.01,0.08705919981002808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,4,power_law_1.01,2.2771711349487305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,4,power_law_1.2,0.35948801040649414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,4,power_law_1.01,0.09482880234718323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,4,power_law_1.2,0.4227456092834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,4,power_law_1.01,4.605107116699219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,4,power_law_1.01,0.09932799935340882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,4,power_law_1.2,0.5855167865753174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,4,power_law_1.01,0.1028607964515686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,4,power_law_1.2,0.8216959953308105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,4,power_law_1.01,0.10542720556259155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,4,power_law_1.2,1.1134719848632812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,4,power_law_1.01,0.10909440517425537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,4,power_law_1.01,0.11456639766693115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,4,balanced,1.323514699935913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,4,power_law_1.2,1.4462783813476563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,4,power_law_1.01,0.12335360050201416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,4,power_law_1.2,1.7971904754638672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,4,power_law_1.01,0.13438719511032104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,4,power_law_1.01,0.14187519550323485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,4,power_law_1.2,2.932524871826172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,4,power_law_1.01,0.1482751965522766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,4,power_law_1.01,0.16739200353622435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,4,power_law_1.2,0.05004799962043762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,4,power_law_1.2,5.189043045043945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,4,power_law_1.01,0.18661760091781615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,4,power_law_1.01,0.22927360534667968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,4,power_law_1.2,0.06259199976921082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,4,power_law_1.01,0.26638720035552976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,4,power_law_1.2,0.05941759943962097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,4,power_law_1.01,0.36509439945220945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,4,power_law_1.2,0.06037120223045349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,4,power_law_1.2,0.06974719762802124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,4,power_law_1.01,0.43181438446044923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,4,balanced,1.6558292706807454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,4,power_law_1.2,0.08279680013656616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,4,power_law_1.01,0.6023231983184815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,4,power_law_1.2,0.09509119987487794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,4,power_law_1.01,0.8021056175231933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,4,power_law_1.2,0.11227519512176513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,4,power_law_1.01,1.084607982635498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,4,power_law_1.2,0.1149440050125122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,4,power_law_1.2,0.11793279647827148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,4,power_law_1.01,1.5114879608154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,4,power_law_1.2,0.1271615982055664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,4,power_law_1.01,1.7992576599121093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,4,power_law_1.2,0.1275904059410095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,4,power_law_1.2,0.13841919898986815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,4,power_law_1.01,2.761324882507324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,4,power_law_1.2,0.1391808032989502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,4,power_law_1.2,0.14872959852218628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,4,power_law_1.01,5.309881591796875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,4,power_law_1.2,0.15259519815444947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,4,power_law_1.2,0.16499840021133422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,4,power_law_1.2,0.19107199907302858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,4,balanced,4.048437436421712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,4,power_law_1.2,0.21075839996337892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,4,power_law_1.2,0.24945919513702391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,4,power_law_1.2,0.2888063907623291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,4,power_law_1.2,0.35542399883270265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,4,balanced,2.494256019592285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,4,power_law_1.2,0.4672895908355713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,4,power_law_1.2,0.628768014907837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,4,power_law_1.2,0.7683199882507324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,4,power_law_1.2,1.060108757019043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,4,power_law_1.2,1.3055232048034668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,4,power_law_1.2,1.7263168334960937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,4,power_law_1.2,2.568704032897949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,4,power_law_1.2,4.742879867553711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,4,power_law_1.2,0.04837760031223297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,4,power_law_1.2,0.05148800015449524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,4,power_law_1.2,0.055769598484039305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,4,power_law_1.2,0.05661439895629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,4,power_law_1.2,0.06305279731750488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,4,power_law_1.2,0.0754368007183075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,4,power_law_1.2,0.07921919822692872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,4,power_law_1.2,0.09381759762763978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,4,power_law_1.2,0.09748479723930359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,4,power_law_1.2,0.09955840110778809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,4,power_law_1.2,0.1019327998161316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,4,power_law_1.2,0.11299200057983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,4,power_law_1.2,0.11156480312347412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,4,power_law_1.2,0.12611199617385865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,4,power_law_1.2,0.13527679443359375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,4,power_law_1.2,0.13882240056991577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,4,power_law_1.2,0.14431359767913818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,4,power_law_1.2,0.16769280433654785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,4,power_law_1.2,0.19088000059127808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,4,power_law_1.2,0.236844801902771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,4,power_law_1.2,0.2789504051208496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,4,power_law_1.2,0.3555583953857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,4,power_law_1.2,0.45311360359191893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,4,balanced,4.8815412521362305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,4,power_law_1.2,0.6319551944732666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,4,power_law_1.2,0.7771776199340821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,4,power_law_1.2,1.1800000190734863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,4,power_law_1.2,1.5904383659362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,4,power_law_1.2,1.8678207397460938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,4,power_law_1.2,2.8290687561035157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,4,power_law_1.2,5.400806427001953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,128,balanced,0.0461760014295578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,128,balanced,0.04643199841181437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,128,balanced,0.046767999728520714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,128,balanced,0.04781866570313772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,128,balanced,0.048101335763931274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,128,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,128,balanced,0.04972266654173533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,128,balanced,0.05148266752560934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,128,balanced,0.05000533163547516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,128,balanced,0.05051200091838837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,128,balanced,0.049728001157442726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,128,balanced,0.0499946673711141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,128,balanced,0.05148266752560934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,128,balanced,0.03982933362325033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,128,balanced,0.04996799925963084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,128,balanced,0.05187733471393585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,128,balanced,0.038805333276589714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,128,balanced,0.053898667295773826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,128,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,128,balanced,0.05186133086681366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,128,balanced,0.041759997606277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,128,balanced,0.0415040006240209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,128,balanced,0.05242133140563965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,128,balanced,0.0436160018046697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,128,balanced,0.04385599990685781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,128,balanced,0.0518453319867452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,128,balanced,0.04348800083001455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,128,balanced,0.043621331453323364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,128,balanced,0.05588266750176748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,128,balanced,0.04380266865094503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,128,balanced,0.04524800181388855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,128,balanced,0.045941332976023354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,128,balanced,0.05789333085219065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,128,balanced,0.04459733267625173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,128,balanced,0.046015997727712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,128,balanced,0.05006400247414907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,128,balanced,0.06406400104363759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,128,balanced,0.04786666731039683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,128,balanced,0.0496319979429245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,128,balanced,0.05198933184146881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,128,balanced,0.07425066828727722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,128,balanced,0.054485330979029335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,128,balanced,0.05784533421198527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,128,balanced,0.06400000055631001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,128,balanced,0.08676266670227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,128,balanced,0.07983466486136119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,128,balanced,0.10301333665847778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,128,balanced,0.10048533479372661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,128,balanced,0.13556266824404398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,128,balanced,0.12714133659998575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,128,balanced,0.16263999541600546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,128,balanced,0.22062400976816812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,128,balanced,0.15800000230471292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,128,balanced,0.28091200192769367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,128,balanced,0.18422933419545492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,128,balanced,0.34046932061513263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,128,balanced,0.26336532831192017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,128,balanced,0.5083893140157064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,2,balanced,0.05373333394527435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,2,balanced,0.05385600030422211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,2,balanced,0.059808000922203064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,2,balanced,0.0684853345155716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,2,balanced,0.09556800127029419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,2,balanced,0.14444266756375632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,2,balanced,0.1920479933420817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,2,balanced,0.17527467012405396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,2,balanced,0.17595199743906656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,2,balanced,0.17779733737309775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,2,balanced,0.1800853411356608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,2,balanced,0.18476800123850504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,2,balanced,0.18542399009068808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,2,balanced,0.18849066893259683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,2,balanced,0.1980266571044922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,2,balanced,0.19948800404866537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,128,balanced,0.47869332631429035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,2,balanced,0.211407999197642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,2,balanced,0.2332693338394165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,2,balanced,0.25340267022450763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,2,balanced,0.29983999331792194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,2,balanced,0.3360639810562134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,128,balanced,0.9661440054575602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,2,balanced,0.4240266482035319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,2,balanced,0.5369439919789633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,2,balanced,0.7047413190205892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,2,balanced,0.9197920163472494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,2,balanced,1.325269301732381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,2,balanced,1.738320032755534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,2,balanced,2.150325298309326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,2,balanced,0.05177066723505656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,2,balanced,0.05213866631189982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,2,balanced,0.05620799958705902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,2,balanced,0.06423466900984447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,2,balanced,0.08228800197442372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,2,balanced,0.11813333630561829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,2,balanced,0.14994133512179056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,2,balanced,0.15060266852378845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,2,balanced,0.152319997549057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,2,balanced,0.15429332852363586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,2,balanced,0.15502933661142984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,2,balanced,0.16114667057991028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,2,balanced,0.16244799892107645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,2,power_law_1.01,0.04831359982490539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,2,balanced,0.16657599806785583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,2,balanced,3.277104059855143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,2,balanced,0.17729065815607706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,2,power_law_1.01,0.053529602289199826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,2,balanced,0.17938133080800375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,2,balanced,0.19329599539438883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,2,power_law_1.01,0.06166399717330932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,2,power_law_1.01,0.06234239935874939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,2,balanced,0.22536534070968628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,2,power_law_1.01,0.08084480166435241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,2,balanced,0.24409067630767822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,2,power_law_1.01,0.10242559909820556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,2,balanced,0.29918932914733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,2,power_law_1.01,0.11856000423431397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,2,power_law_1.01,0.13276159763336182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,2,power_law_1.01,0.13927680253982544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,2,balanced,0.3394026756286621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,2,power_law_1.01,0.14416639804840087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,2,power_law_1.01,0.14724479913711547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,2,balanced,0.4359573523203532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,2,power_law_1.01,0.15300480127334595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,2,power_law_1.01,0.15571839809417726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,2,power_law_1.01,0.16536320447921754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,2,power_law_1.01,0.17781120538711548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,2,balanced,0.5662773450215658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,2,power_law_1.01,0.17887359857559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,2,power_law_1.01,0.19226880073547364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,2,power_law_1.01,0.22405760288238524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,2,power_law_1.01,0.25004799365997316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,2,power_law_1.2,0.04810880124568939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,2,balanced,0.757968028386434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,2,power_law_1.01,0.3096767902374268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,2,power_law_1.01,0.35735039710998534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,2,power_law_1.2,0.05310080051422119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,2,power_law_1.2,0.057740801572799684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,2,power_law_1.01,0.4577023983001709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,2,power_law_1.01,0.577619218826294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,2,power_law_1.2,0.06272000074386597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,2,power_law_1.2,0.07790079712867737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,2,power_law_1.01,0.7771455764770507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,2,power_law_1.2,0.0986240029335022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,2,power_law_1.01,0.9653887748718262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,2,balanced,1.036405324935913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,2,power_law_1.2,0.10978560447692871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,2,power_law_1.01,1.384224033355713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,2,power_law_1.2,0.1305408000946045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,2,power_law_1.01,0.04880000054836273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,2,power_law_1.2,0.13602559566497802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,2,power_law_1.01,1.8008127212524414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,2,power_law_1.2,0.13804800510406495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,2,power_law_1.01,0.05229439735412598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,2,power_law_1.01,2.234771156311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,2,power_law_1.2,0.14332799911499022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,2,power_law_1.01,0.06160640120506287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,2,power_law_1.2,0.15017600059509278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,2,power_law_1.01,3.472684860229492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,2,power_law_1.01,0.06280320286750793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,2,power_law_1.2,0.15523200035095214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,2,power_law_1.01,0.07882879972457886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,2,power_law_1.2,0.16596479415893556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,2,power_law_1.01,0.09275519847869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,2,power_law_1.2,0.17834240198135376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,2,power_law_1.01,6.986188507080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,2,balanced,1.4775999387105305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,2,power_law_1.01,0.10140800476074219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,2,power_law_1.2,0.1803455948829651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,2,power_law_1.2,0.1953984022140503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,2,power_law_1.01,0.11783039569854736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,2,power_law_1.2,0.22736001014709473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,2,power_law_1.01,0.1226304054260254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,2,power_law_1.2,0.25282559394836424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,128,balanced,0.039621333281199135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,2,power_law_1.01,0.1283903956413269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,128,balanced,0.019178666174411774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,2,power_law_1.2,0.3163264036178589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,2,power_law_1.01,0.13011200428009034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,128,balanced,0.04041599979003271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,2,power_law_1.2,0.3633536100387573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,128,balanced,0.019013332823912304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,2,power_law_1.01,0.14023679494857788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,128,balanced,0.04053866614898046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,2,power_law_1.2,0.46771841049194335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,2,power_law_1.01,0.1479871988296509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,2,power_law_1.2,0.04842239916324616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,128,balanced,0.021087999145189922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,128,balanced,0.04257066547870636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,2,balanced,6.374277114868164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,2,power_law_1.01,0.15907200574874877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,2,power_law_1.2,0.5854911804199219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,128,balanced,0.04045866678158442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,128,balanced,0.023365333676338196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,2,power_law_1.01,0.17100800275802613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,2,power_law_1.2,0.05237759947776795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,128,balanced,0.04393066465854645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,128,balanced,0.03956799954175949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,2,power_law_1.2,0.7703872203826905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,128,balanced,0.04404266675313314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,2,power_law_1.01,0.18101119995117188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,128,balanced,0.03341866781314214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,2,power_law_1.2,0.056620800495147706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,2,balanced,1.9116106033325195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,128,balanced,0.04362666606903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,2,power_law_1.2,1.000921630859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,128,balanced,0.033728001018365227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,128,balanced,0.043920000394185386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,2,power_law_1.01,0.2059391975402832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,2,power_law_1.2,0.06276479959487916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,128,balanced,0.03339199970165888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,128,balanced,0.04423466821511587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,2,power_law_1.2,1.4191167831420899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,128,balanced,0.034117333590984344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,2,power_law_1.01,0.2427135944366455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,2,power_law_1.2,0.07442560195922851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,128,balanced,0.03342399994532267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,128,balanced,0.04378666480382284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,2,power_law_1.01,0.24843521118164064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,2,power_law_1.2,1.8448896408081055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,2,power_law_1.01,0.3132031917572021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,128,balanced,0.03495466709136963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,128,balanced,0.04580266773700714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,2,power_law_1.2,0.09074559807777405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,2,power_law_1.01,0.36520960330963137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,128,balanced,0.03335466732581457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,2,power_law_1.2,2.2789632797241213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,128,balanced,0.045221333702405296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,2,power_law_1.2,0.0989184021949768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,128,balanced,0.034304000437259674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,2,power_law_1.01,0.4801152229309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,128,balanced,0.0454773356517156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,128,balanced,0.03519999980926514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,2,power_law_1.2,3.607283020019531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,2,power_law_1.01,0.6095295906066894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,2,power_law_1.2,0.11800960302352906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,128,balanced,0.050000001986821495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,128,balanced,0.03755199909210205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,2,power_law_1.01,0.8538687705993653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,128,balanced,0.0480373352766037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,2,power_law_1.2,0.12369920015335083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,128,balanced,0.038021333515644073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,2,power_law_1.01,1.0803775787353516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,2,power_law_1.2,0.12688640356063843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,128,balanced,0.049813335140546165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,2,power_law_1.2,7.404812622070312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,2,power_law_1.01,1.507033634185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,128,balanced,0.037589333951473236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,2,power_law_1.2,0.1299839973449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,128,balanced,0.051642666260401406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,2,balanced,2.388528029123942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,2,power_law_1.01,1.9803583145141601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,2,power_law_1.2,0.14039679765701293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,128,balanced,0.03956799954175949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,128,balanced,0.05411200225353241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,2,power_law_1.2,0.14956799745559693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,2,power_law_1.01,2.3995967864990235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,128,balanced,0.0425546665986379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,2,power_law_1.2,0.1587839961051941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,128,balanced,0.054661333560943604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,2,power_law_1.01,3.756460952758789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,2,power_law_1.2,0.17622400522232057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,128,balanced,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,2,power_law_1.2,0.18567680120468139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,128,balanced,0.06016000111897787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,2,power_law_1.01,7.707852935791015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,2,power_law_1.2,0.20702080726623534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,128,balanced,0.061887999375661217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,2,power_law_1.2,0.2499840021133423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,2,power_law_1.2,0.25125761032104493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,128,balanced,0.07025599976380666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,2,power_law_1.2,0.3189055919647217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,128,balanced,0.07562133173147838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,2,power_law_1.2,0.3787904024124146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,128,balanced,0.08018133540948232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,2,power_law_1.2,0.5013631820678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,128,balanced,0.09502399961153667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,2,power_law_1.2,0.629414415359497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,2,power_law_1.2,0.8526975631713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,128,balanced,0.10307733217875163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,2,power_law_1.2,1.1077823638916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,128,balanced,0.12803733348846436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,2,balanced,3.675509452819824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,2,power_law_1.2,1.6050432205200196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,2,power_law_1.2,1.9703424453735352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,2,power_law_1.2,2.5314239501953124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,128,balanced,0.12085333466529846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,128,balanced,0.16269333163897196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,2,power_law_1.2,3.8969280242919924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,2,power_law_1.2,7.895436859130859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,128,balanced,0.1567093332608541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,128,balanced,0.22769065697987875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,128,balanced,0.1938613255818685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,128,balanced,0.2934826612472534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,2,balanced,7.194143931070964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,128,balanced,0.2323573430379232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,128,balanced,0.35896531740824383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,128,balanced,0.3412959973017375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,128,balanced,0.5552106698354086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,128,power_law_1.01,0.045228800177574156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,128,power_law_1.01,0.04665600061416626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,128,power_law_1.01,0.044787201285362246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,128,power_law_1.01,0.04762240052223206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,128,power_law_1.01,0.047942399978637695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,128,power_law_1.01,0.04917120039463043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,128,power_law_1.01,0.0469760000705719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,128,power_law_1.01,0.04657920002937317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,128,power_law_1.01,0.04639999866485596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,128,power_law_1.01,0.045561599731445315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,128,power_law_1.01,0.046137601137161255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,128,power_law_1.01,0.04559360146522522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,128,power_law_1.01,0.04481280148029328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,128,power_law_1.01,0.044870400428771974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,128,power_law_1.01,0.046758401393890384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,128,power_law_1.01,0.0385919988155365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,128,power_law_1.01,0.05092480182647705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,128,power_law_1.01,0.050995200872421265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,128,power_law_1.01,0.03866879940032959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,128,power_law_1.01,0.05297920107841492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,128,power_law_1.01,0.03793280124664307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,128,power_law_1.01,0.05787519812583923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,128,power_law_1.01,0.04012799859046936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,128,power_law_1.01,0.07363839745521546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,128,power_law_1.01,0.040550398826599124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,128,power_law_1.01,0.08170239925384522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,128,power_law_1.01,0.042719998955726625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,128,power_law_1.01,0.1094655990600586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,128,power_law_1.01,0.04132480025291443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,128,power_law_1.01,0.1285248041152954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,128,power_law_1.01,0.04131200015544891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,128,power_law_1.01,0.04229759871959686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,128,power_law_1.01,0.18157440423965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,128,power_law_1.01,0.04222719967365265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,128,power_law_1.01,0.2262336015701294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,128,power_law_1.01,0.042316800355911253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,128,power_law_1.01,0.33935999870300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,128,power_law_1.01,0.042752000689506534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,128,power_law_1.01,0.03733119964599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,128,power_law_1.01,0.42465920448303224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,128,power_law_1.01,0.04368639886379242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,128,power_law_1.01,0.04323840141296387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,128,power_law_1.01,0.5414976119995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,128,power_law_1.01,0.03710080087184906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,128,power_law_1.01,0.04640640020370483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,128,power_law_1.01,0.0374208003282547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,128,power_law_1.01,0.9514559745788574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,128,power_law_1.01,0.048895999789237976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,128,power_law_1.01,0.039078399538993835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,128,power_law_1.01,0.04878079891204834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,128,power_law_1.01,0.0395583987236023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,128,power_law_1.01,1.938662338256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,128,power_law_1.01,0.05495679974555969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,128,power_law_1.01,0.04071039855480194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,128,power_law_1.01,0.05869439840316772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,128,power_law_1.01,0.04079360067844391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,128,power_law_1.01,0.06762880086898804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,128,power_law_1.01,0.04020479917526245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,128,power_law_1.01,0.08578559756278992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,128,power_law_1.01,0.03961600065231323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,128,power_law_1.01,0.11399680376052856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,128,power_law_1.01,0.040524798631668094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,128,power_law_1.01,0.1428287982940674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,128,power_law_1.01,0.018572799861431122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,128,power_law_1.01,0.19511040449142455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,128,balanced,1.082192023595174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,128,power_law_1.01,0.04132480025291443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,128,balanced,0.6337386767069498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,128,power_law_1.01,0.2415679931640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,128,power_law_1.01,0.04080640077590943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,128,power_law_1.01,0.01847680062055588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,128,power_law_1.01,0.3577215909957886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,128,power_law_1.01,0.04158720076084137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,128,power_law_1.01,0.47843198776245116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,128,power_law_1.01,0.01953279972076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,128,power_law_1.01,0.042412799596786496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,128,power_law_1.01,0.6119040012359619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,128,power_law_1.01,0.02253440022468567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,128,power_law_1.01,0.04332799911499023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,128,power_law_1.01,0.9374527931213379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,128,power_law_1.01,0.045151999592781066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,128,power_law_1.01,0.03895680010318756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,128,power_law_1.01,1.911801528930664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,128,power_law_1.01,0.04762240052223206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,128,power_law_1.01,0.031411200761795044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,128,power_law_1.01,0.053881597518920896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,128,power_law_1.01,0.03225600123405457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,128,power_law_1.01,0.032390400767326355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.01,0.059001600742340087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,128,power_law_1.01,0.0324288010597229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.01,0.06822400093078614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,128,power_law_1.01,0.03239679932594299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.01,0.07765120267868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,128,power_law_1.01,0.032358399033546446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.01,0.09633920192718506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,128,power_law_1.01,0.034169599413871765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.01,0.11019519567489625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,128,power_law_1.01,0.032416000962257385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.01,0.1561087965965271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,128,power_law_1.01,0.03277440071105957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,128,power_law_1.01,0.03430399894714355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.01,0.20018560886383058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.01,0.2799423933029175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,128,power_law_1.01,0.03537279963493347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,1,128,power_law_1.2,0.04480000138282776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.01,0.37932798862457273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,128,power_law_1.01,0.03669120073318481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,1,128,power_law_1.2,0.045491200685501096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,128,power_law_1.01,0.04010240137577057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,128,power_law_1.01,0.44337921142578124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,1,128,power_law_1.2,0.04485119879245758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.01,0.040659201145172116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,1,128,power_law_1.2,0.04737919867038727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,128,power_law_1.01,0.7077248096466064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.01,0.048230400681495665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,1,128,power_law_1.2,0.04649600088596344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.01,0.05509120225906372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,1,128,power_law_1.2,0.04815999865531921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,1,128,power_law_1.2,0.04734080135822296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,128,power_law_1.01,1.4415679931640626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.01,0.07062399983406067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,1,128,power_law_1.2,0.04733439981937408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.01,0.08674560189247131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,1,128,power_law_1.2,0.047737601399421695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,1,128,power_law_1.2,0.046649599075317384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.01,0.11975040435791015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,1,128,power_law_1.2,0.046265599131584165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,1,128,power_law_1.2,0.04575360119342804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.01,0.15039360523223877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,1,128,power_law_1.2,0.045516800880432126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,1,128,power_law_1.2,0.047366398572921756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.01,0.20921599864959717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,1,128,power_law_1.2,0.05177599787712097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.01,0.2717184066772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,1,128,power_law_1.2,0.050944000482559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,1,128,power_law_1.2,0.05489919781684875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,128,power_law_1.01,0.3385983943939209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,1,128,power_law_1.2,0.060761600732803345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,1,128,power_law_1.2,0.07128959894180298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,128,power_law_1.01,0.5212863922119141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,1,128,power_law_1.2,0.09032959938049316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,1,128,power_law_1.2,0.10336639881134033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,128,power_law_1.01,1.0302335739135742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,1,128,power_law_1.2,0.13957120180130006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,1,128,power_law_1.2,0.17921279668807982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,1,128,power_law_1.2,0.25231359004974363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,1,128,power_law_1.2,0.34377601146698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,1,128,power_law_1.2,0.5029248237609864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,1,128,power_law_1.2,0.6569600105285645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,1,128,power_law_1.2,0.03766399919986725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,1,128,power_law_1.2,0.815340805053711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,1,128,power_law_1.2,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,1,128,power_law_1.2,1.278444766998291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,1,128,power_law_1.2,0.03804160058498383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,1,128,power_law_1.2,2.5623424530029295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,1,128,power_law_1.2,0.039001598954200745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,1,128,power_law_1.2,0.03909119963645935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,1,128,power_law_1.2,0.04103040099143982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,1,128,power_law_1.2,0.04108160138130188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,1,128,power_law_1.2,0.040531200170516965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,1,128,power_law_1.2,0.04078719913959503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,1,128,power_law_1.2,0.041152000427246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,1,128,power_law_1.2,0.041126400232315063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,1,128,power_law_1.2,0.04178560078144074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,1,128,power_law_1.2,0.04230400025844574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,1,128,power_law_1.2,0.04351359903812409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,1,128,power_law_1.2,0.047276800870895384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,1,128,power_law_1.2,0.046751999855041505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,1,128,power_law_1.2,0.04935039877891541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,1,128,power_law_1.2,0.056383997201919556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,1,128,power_law_1.2,0.06543359756469727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,1,128,power_law_1.2,0.07784960269927979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,1,128,power_law_1.2,0.10798720121383668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,1,128,power_law_1.2,0.14076160192489623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,1,128,power_law_1.2,0.1748800039291382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,1,128,power_law_1.2,0.25414400100708007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,1,128,power_law_1.2,0.3033152103424072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,1,128,power_law_1.2,0.47587199211120607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,1,128,power_law_1.2,0.631935977935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,1,128,power_law_1.2,0.7834303855895997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,1,128,power_law_1.2,1.2146047592163085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,1,128,power_law_1.2,2.4063488006591798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,1,128,power_law_1.2,0.03709439933300018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,1,128,power_law_1.2,0.03769600093364715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,1,128,power_law_1.2,0.018585599958896637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,1,128,power_law_1.2,0.037676799297332766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,1,128,power_law_1.2,0.03929600119590759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,1,128,power_law_1.2,0.018118399381637573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,1,128,power_law_1.2,0.038796800374984744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,1,128,power_law_1.2,0.019244800508022308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,1,128,power_law_1.2,0.04044800102710724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,1,128,power_law_1.2,0.021984000504016877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,1,128,power_law_1.2,0.040403199195861814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,1,128,power_law_1.2,0.03818880021572113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,1,128,power_law_1.2,0.039929598569869995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,1,128,power_law_1.2,0.03247359991073608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,1,128,power_law_1.2,0.03983359932899475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,1,128,power_law_1.2,0.032742398977279666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,1,128,power_law_1.2,0.040428799390792844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,1,128,power_law_1.2,0.03292160034179688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,1,128,power_law_1.2,0.04043520092964172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,1,128,power_law_1.2,0.03304319977760315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,1,128,power_law_1.2,0.04040960073471069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,1,128,power_law_1.2,0.03322240114212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,1,128,power_law_1.2,0.04102399945259094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,1,128,power_law_1.2,0.03324800133705139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,1,128,power_law_1.2,0.044172799587249754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,1,128,power_law_1.2,0.0339711993932724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,1,128,power_law_1.2,0.0480320006608963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,1,128,power_law_1.2,0.03281280100345611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,1,128,power_law_1.2,0.0465472012758255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,1,128,power_law_1.2,0.03274880051612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,1,128,power_law_1.2,0.04965119957923889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,1,128,power_law_1.2,0.034841600060462954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,1,128,power_law_1.2,0.057811200618743896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,1,128,power_law_1.2,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.2,0.06301440000534057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,1,128,power_law_1.2,0.035872000455856326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.2,0.07565439939498901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,1,128,power_law_1.2,0.042361599206924436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.2,0.08983039855957031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.2,0.04079360067844391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.2,0.1141055941581726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.2,0.04751999974250794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.2,0.14211200475692748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.2,0.05836799740791321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.2,0.19556479454040526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.2,0.07264000177383423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.2,0.2674175977706909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.2,0.09086719751358033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.2,0.12242560386657715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.2,0.3815232038497925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.2,0.15790079832077025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.2,0.49319682121276853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.2,0.21976959705352783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,1,128,power_law_1.2,0.6061312198638916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.2,0.2904383897781372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,1,128,power_law_1.2,0.953593635559082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,1,128,power_law_1.2,0.35582718849182127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,1,128,power_law_1.2,1.8992128372192383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,1,128,power_law_1.2,0.5531775951385498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,1,128,power_law_1.2,1.0789119720458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,64,balanced,0.03341866781314214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,64,balanced,0.01937066639463107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,64,balanced,0.035418666899204254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,64,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,64,balanced,0.02109866589307785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,64,balanced,0.033626665671666466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,64,balanced,0.019776000330845516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,64,balanced,0.03575466573238373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,64,balanced,0.035786665976047516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,64,balanced,0.019280000279347103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,64,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,64,balanced,0.01974933346112569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,64,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,64,balanced,0.02314666658639908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,64,balanced,0.035743998984495796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,64,balanced,0.023242667317390442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,64,balanced,0.03769599894682566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,64,balanced,0.037434667348861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,64,balanced,0.03229333211978277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,64,balanced,0.037690666814645134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,64,balanced,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,64,balanced,0.0377813329299291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,64,balanced,0.0334346666932106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,64,balanced,0.03961066653331121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,64,balanced,0.03339733431736628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,64,balanced,0.0435146689414978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,64,balanced,0.03263466556866964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,64,balanced,0.04188266893227895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,64,balanced,0.034245334565639496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,64,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,64,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,64,balanced,0.045647998650868736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,64,balanced,0.0371573343873024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,64,balanced,0.04789333542188009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,64,balanced,0.03569599986076355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,64,balanced,0.03676266719897588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,64,balanced,0.04982399940490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,64,balanced,0.04172799984614054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,64,balanced,0.05382933219273885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,64,balanced,0.04565866788228353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,64,balanced,0.051776001850763954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,64,balanced,0.058559998869895935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,64,balanced,0.06380799909432729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,64,balanced,0.07045866549015045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,64,balanced,0.0790773332118988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,64,balanced,0.0902933379014333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,64,balanced,0.09893866380055745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,64,balanced,0.13057066996892294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,64,balanced,0.10125866532325745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,64,balanced,0.16683199008305868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,64,balanced,0.12754133343696594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,64,balanced,0.23287999629974365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,64,balanced,0.16314666469891867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,64,balanced,0.3002293308575948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,64,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,64,balanced,0.0373279998699824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,64,balanced,0.03159466634194056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,64,balanced,0.031445334355036415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,64,balanced,0.03379733363787333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,64,balanced,0.033386667569478355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,64,balanced,0.19176000356674194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,64,balanced,0.033530667424201965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,64,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,64,balanced,0.03366400053103765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,64,balanced,0.035391998787721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,64,balanced,0.03339733431736628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,64,balanced,0.03338133295377096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,64,balanced,0.031658666829268135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,64,balanced,0.03532266616821289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,64,balanced,0.034927998979886375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,64,balanced,0.035402665535608925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,64,balanced,0.03550933301448822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,64,balanced,0.035605333745479584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,64,balanced,0.03547733277082443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,64,balanced,0.03742400060097376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,64,balanced,0.3684213161468506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,64,balanced,0.033786666889985405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,64,balanced,0.03505066782236099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,64,balanced,0.0354666660229365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,64,balanced,0.03573333223660787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,64,balanced,0.03916800022125244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,64,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,64,power_law_1.01,0.03194240033626557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,64,balanced,0.03948266555865606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,64,balanced,0.03736000011364619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,64,balanced,0.040463998913764954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,64,balanced,0.03738133360942205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,64,power_law_1.01,0.03278720080852508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,64,balanced,0.03752533346414566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,64,power_law_1.01,0.03219839930534363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,64,balanced,0.0394400010506312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,64,power_law_1.01,0.03468799889087677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,64,balanced,0.04160533348719279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,64,power_law_1.01,0.0349375993013382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,64,balanced,0.041434665520985924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,64,balanced,0.04079466561476389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,64,power_law_1.01,0.035673600435256955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,64,power_law_1.01,0.03561600148677826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,64,balanced,0.04258666435877482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,64,balanced,0.04377066592375437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,64,power_law_1.01,0.03580160140991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,64,balanced,0.043477331598599754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,64,power_law_1.01,0.0357120007276535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,64,power_law_1.01,0.03656960129737854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,64,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,64,balanced,0.046581332882245384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,64,power_law_1.01,0.03729279935359955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,64,power_law_1.01,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,64,balanced,0.04985600213209788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,64,power_law_1.01,0.03898879885673523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,64,balanced,0.05315199991067251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,64,power_law_1.01,0.04008319973945618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,64,power_law_1.01,0.017715199291706084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,64,balanced,0.2637866735458374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,64,power_law_1.01,0.04245760142803192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,64,balanced,0.05414933462937673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,64,power_law_1.01,0.04381439983844757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,64,power_law_1.01,0.017868800461292265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,64,power_law_1.01,0.04671359956264496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,64,power_law_1.01,0.017785599827766417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,64,power_law_1.01,0.04997760057449341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,64,balanced,0.05572799841562907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,64,power_law_1.01,0.01849599927663803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.01,0.05394560098648071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,64,balanced,0.0729013333717982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,64,power_law_1.01,0.022368000447750093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.01,0.06233599781990051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,64,power_law_1.01,0.038099199533462524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.01,0.07164160013198853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,64,power_law_1.01,0.032441601157188416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.01,0.0887615978717804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,64,power_law_1.01,0.03285120129585266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.01,0.10374399423599243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,64,balanced,0.577397346496582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,64,balanced,0.09166399637858073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,64,power_law_1.01,0.03214080035686493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,64,balanced,0.07092800239721934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.01,0.13969919681549073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,64,power_law_1.01,0.032652801275253295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.01,0.181113600730896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,64,power_law_1.01,0.032569599151611325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,64,power_law_1.01,0.03258239924907684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.01,0.22608640193939208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,64,power_law_1.01,0.03166080117225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.01,0.2742719888687134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,64,balanced,0.11954133709271748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,64,power_law_1.01,0.03153280019760132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,64,power_law_1.01,0.3930624008178711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,64,power_law_1.01,0.032915198802948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,64,balanced,0.07865599791208903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,64,power_law_1.01,0.5507967948913575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,64,power_law_1.01,0.033632001280784606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,64,power_law_1.01,0.036006399989128114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,64,power_law_1.01,0.9266240119934082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,64,power_law_1.01,0.03941119909286499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.01,0.0426367998123169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.01,0.048492801189422605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,64,power_law_1.01,0.03389439880847931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,64,balanced,0.14628799756368002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.01,0.05605120062828064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.01,0.07279999852180481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,64,power_law_1.01,0.03663359880447388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.01,0.09108480215072631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,64,power_law_1.01,0.031327998638153075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,64,power_law_1.01,0.03329919874668121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.01,0.11356159448623657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.01,0.14844160079956054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,64,power_law_1.01,0.03327359855175018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,64,balanced,0.09501866499582927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.01,0.21068799495697021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,64,power_law_1.01,0.03377279937267304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.01,0.26711039543151854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,64,power_law_1.01,0.03342080116271973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,64,power_law_1.01,0.033542400598526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,64,power_law_1.01,0.3153408050537109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,64,power_law_1.01,0.03377279937267304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,64,power_law_1.01,0.48453121185302733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,64,power_law_1.01,0.033888000249862674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,64,balanced,0.1943946679433187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,64,power_law_1.01,0.03529599905014038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,64,power_law_1.01,1.0112832069396973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,64,power_law_1.01,0.03553279936313629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,64,power_law_1.01,0.03632639944553375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,64,power_law_1.01,0.03906559944152832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,64,power_law_1.01,0.041503998637199405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,64,power_law_1.01,0.042752000689506534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,64,power_law_1.01,0.0466048002243042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,64,balanced,0.11348266402880351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,64,power_law_1.01,0.05382400155067444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,64,power_law_1.01,0.05667200088500977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,64,power_law_1.01,0.03134079873561859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,64,power_law_1.01,0.07018240094184876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,64,power_law_1.01,0.06982399821281433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,64,power_law_1.01,0.03206399977207184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,64,power_law_1.01,0.09063040018081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,64,power_law_1.01,0.031020799279212953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,64,power_law_1.01,0.12250239849090576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,64,balanced,0.24899733066558838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,64,power_law_1.01,0.031641599535942075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,64,power_law_1.01,0.14929280281066895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,64,balanced,0.4822346766789754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,64,power_law_1.01,0.032287999987602234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,64,power_law_1.01,0.16559360027313233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,64,power_law_1.01,0.033004799485206605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,64,power_law_1.01,0.2316864013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,64,power_law_1.01,0.03299199938774109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,64,power_law_1.01,0.3039103984832764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,64,power_law_1.01,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,64,power_law_1.01,0.3940095901489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,64,power_law_1.01,0.0335999995470047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,64,power_law_1.01,0.03457280099391937
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,64,power_law_1.01,0.5035136222839356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,64,power_law_1.01,0.035622400045394895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,64,balanced,0.13193066914876303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,64,power_law_1.01,0.036959999799728395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,64,power_law_1.01,1.1749631881713867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,64,power_law_1.01,0.037868800759315493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,64,balanced,1.1251253286997478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,64,power_law_1.01,0.03957119882106781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,64,power_law_1.01,0.043059200048446655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,64,power_law_1.01,0.04482559859752655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,64,balanced,0.30155734221140545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,64,power_law_1.01,0.04601599872112274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,64,power_law_1.01,0.0514240026473999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,64,power_law_1.01,0.057171201705932616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,64,power_law_1.01,0.06940159797668458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,64,power_law_1.01,0.07559679746627808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,64,power_law_1.01,0.10661760568618775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,64,power_law_1.01,0.12859519720077514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,64,power_law_1.01,0.19804160594940184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,64,power_law_1.01,0.2290112018585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,64,power_law_1.01,0.33978240489959716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,64,power_law_1.01,0.4087039947509766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,64,power_law_1.01,0.4828927993774414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,64,balanced,0.17523199319839478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,64,power_law_1.01,0.7936448097229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,64,power_law_1.01,1.7191743850708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,64,balanced,0.43647468090057373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,64,balanced,0.29948800802230835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,64,balanced,0.8195466995239258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,64,power_law_1.2,0.03342719972133636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,64,power_law_1.2,0.03455359935760498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,64,power_law_1.2,0.02927359938621521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,64,power_law_1.2,0.03240320086479187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,64,power_law_1.2,0.03133440017700195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,64,power_law_1.2,0.03290880024433136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,64,power_law_1.2,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,64,power_law_1.2,0.032332798838615416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,64,power_law_1.2,0.03341439962387085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,64,power_law_1.2,0.03346560001373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,64,power_law_1.2,0.03498240113258362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,64,power_law_1.2,0.03592320084571839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,64,power_law_1.2,0.03160319924354553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,64,power_law_1.2,0.03712640106678009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,64,power_law_1.2,0.03848319947719574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,64,power_law_1.2,0.027923199534416198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,64,power_law_1.2,0.04258559942245484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,64,power_law_1.2,0.029382398724555968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,64,power_law_1.2,0.04352000057697296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,64,power_law_1.2,0.031014400720596313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,64,power_law_1.2,0.04796159863471985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,64,power_law_1.2,0.0316864013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,64,power_law_1.2,0.05633919835090637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,64,power_law_1.2,0.03272320032119751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,64,power_law_1.2,0.05978879928588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,64,power_law_1.2,0.03271040022373199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,64,power_law_1.2,0.07896959781646729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,64,power_law_1.2,0.03308799862861633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,64,power_law_1.2,0.08855040073394775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,64,power_law_1.2,0.03324800133705139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,64,power_law_1.2,0.13022079467773437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,64,power_law_1.2,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,64,power_law_1.2,0.13139840364456176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,64,power_law_1.2,0.03516159951686859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,64,power_law_1.2,0.2144767999649048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,64,power_law_1.2,0.03607679903507233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,64,power_law_1.2,0.031001600623130798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,64,power_law_1.2,0.038124799728393555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,64,power_law_1.2,0.23912959098815917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,64,power_law_1.2,0.031763198971748355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,64,power_law_1.2,0.03871999979019165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,64,power_law_1.2,0.3776895999908447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,64,power_law_1.2,0.03149439990520477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,64,power_law_1.2,0.04411520063877106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,64,power_law_1.2,0.0341376006603241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,64,power_law_1.2,0.4949952125549316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,64,power_law_1.2,0.04645119905471802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,64,power_law_1.2,0.03478400111198425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,64,power_law_1.2,0.7785280227661133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,64,power_law_1.2,0.050316798686981204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,64,power_law_1.2,0.03491199910640717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,64,power_law_1.2,0.05589119791984558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,64,power_law_1.2,1.0230208396911622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,64,power_law_1.2,0.03544319868087768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,64,power_law_1.2,0.0623744010925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,64,power_law_1.2,0.035750401020050046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,64,power_law_1.2,0.07323520183563233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,64,power_law_1.2,2.0238847732543945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,64,power_law_1.2,0.035980799794197084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,64,power_law_1.2,0.09100800156593322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,64,power_law_1.2,0.036339199542999266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,64,power_law_1.2,0.12481919527053834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,64,power_law_1.2,0.03713279962539673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,64,power_law_1.2,0.15186560153961182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,64,power_law_1.2,0.038150399923324585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,64,power_law_1.2,0.26150400638580323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,64,power_law_1.2,0.03927040100097656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,64,power_law_1.2,0.29384961128234866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,64,power_law_1.2,0.040268799662590025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,64,power_law_1.2,0.018617600202560425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,64,power_law_1.2,0.5129151821136475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,64,power_law_1.2,0.04361599981784821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,64,power_law_1.2,0.0438975989818573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,64,power_law_1.2,0.6111551761627197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,64,power_law_1.2,0.018592000007629395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,64,power_law_1.2,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,64,power_law_1.2,0.018592000007629395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,64,power_law_1.2,0.7243199825286866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,64,power_law_1.2,0.050470399856567386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,64,power_law_1.2,0.02011519968509674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,64,power_law_1.2,0.023046399652957916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.2,0.06020479798316956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,64,power_law_1.2,1.2681344032287598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,64,power_law_1.2,0.03895680010318756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.2,0.06610559821128845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,64,power_law_1.2,0.03224959969520569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.2,0.07652480006217957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,64,power_law_1.2,0.03292160034179688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,64,power_law_1.2,2.5629951477050783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.2,0.10658559799194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,64,power_law_1.2,0.0323199987411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.2,0.12376320362091064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,64,power_law_1.2,0.0321152001619339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.2,0.17255680561065673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,64,power_law_1.2,0.03221760094165802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.2,0.2484287977218628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,64,power_law_1.2,0.03312639892101288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.2,0.3801215887069702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,64,power_law_1.2,0.03204480111598969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,64,power_law_1.2,0.032492798566818235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.2,0.4275263786315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,64,power_law_1.2,0.033766400814056394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,64,power_law_1.2,0.6086592197418212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,64,power_law_1.2,0.03480319976806641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,64,power_law_1.2,0.03587839901447296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,64,power_law_1.2,0.8887104034423828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,64,power_law_1.2,0.040012800693511964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.2,0.043609601259231565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,64,power_law_1.2,2.199558448791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.2,0.04830079972743988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.2,0.06110079884529114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.2,0.07704960107803345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.2,0.0950655996799469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.2,0.12668800354003906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.2,0.16360960006713868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.2,0.232422399520874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.2,0.30023040771484377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,64,power_law_1.2,0.3646143913269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,64,power_law_1.2,0.5799871921539307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,64,power_law_1.2,1.1101696014404296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,32,balanced,0.019082666685183842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,32,balanced,0.03583466758330663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,32,balanced,0.021061333517233532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,32,balanced,0.037461332976818085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,32,balanced,0.018976000448067982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,32,balanced,0.0365226666132609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,32,balanced,0.020618667205174763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,32,balanced,0.03584533433119456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,32,balanced,0.03877866764863332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,32,balanced,0.021157334248224895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,32,balanced,0.037776000797748566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,32,balanced,0.020949333906173706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,32,balanced,0.03958400090535482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,32,balanced,0.021061333517233532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,32,balanced,0.03863999992609024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,32,balanced,0.023386667172114056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,32,balanced,0.03807466725508372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,32,balanced,0.02513599892457326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,32,balanced,0.039594667653242745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,32,balanced,0.025087999800841015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,32,balanced,0.03956266740957896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,32,balanced,0.03325333446264267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,32,balanced,0.04161600023508072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,32,balanced,0.03305600086847941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,32,balanced,0.03956266740957896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,32,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,32,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,32,balanced,0.03345600018898646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,32,balanced,0.04580266773700714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,32,balanced,0.038378665844599404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,32,balanced,0.044079999128977455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,32,balanced,0.03937600056330363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,32,balanced,0.04569066564242045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,32,balanced,0.03940266619126002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,32,balanced,0.04786133269468943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,32,balanced,0.04359999795754751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,32,balanced,0.05014933149019877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,32,balanced,0.047728002071380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,32,balanced,0.05209066470464071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,32,balanced,0.057914664347966514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,32,balanced,0.05874133110046387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,32,balanced,0.07301866511503856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,32,balanced,0.06611200173695882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,32,balanced,0.09061333537101746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,32,balanced,0.07780799766381581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,32,balanced,0.11293866237004598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,32,balanced,0.0997973382472992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,32,balanced,0.15242133537928262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,32,balanced,0.11969066659609477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,32,balanced,0.19788267215092978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,32,balanced,0.03336533407370249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,32,balanced,0.03190933416287104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,32,balanced,0.02918400118748347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,32,balanced,0.031194667021433514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,32,balanced,0.0335359995563825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,32,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,32,balanced,0.15169066190719604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,32,balanced,0.033610666791598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,32,balanced,0.2757546703020732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,32,balanced,0.03438399980465571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,32,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,32,balanced,0.03545066714286804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,32,balanced,0.03535466641187668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,32,balanced,0.03502399971087774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,32,balanced,0.03568000098069509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,32,balanced,0.03734400123357773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,32,balanced,0.039077334105968475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,32,balanced,0.039808000127474465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,32,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,32,balanced,0.029631999631722767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,32,balanced,0.04177066683769226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,32,balanced,0.031194667021433514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,32,balanced,0.19379733006159464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,32,balanced,0.04370133578777313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,32,balanced,0.031210665901501972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,32,balanced,0.3588533401489258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,32,balanced,0.031471999982992806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,32,power_law_1.01,0.04007039964199066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,32,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,32,balanced,0.03357866654793421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,32,balanced,0.03328000009059906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,32,power_law_1.01,0.03707520067691803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,32,balanced,0.03538133452335993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,32,balanced,0.03438399980465571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,32,power_law_1.01,0.035724800825119016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,32,balanced,0.04797866443792979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,32,balanced,0.0346666673819224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,32,power_law_1.01,0.03920640051364899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,32,balanced,0.03542399903138479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,32,power_law_1.01,0.03919360041618347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,32,balanced,0.036389333506425224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,32,power_law_1.01,0.04173440039157868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,32,balanced,0.037104000647862755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,32,balanced,0.05693866809209188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,32,power_law_1.01,0.04257279932498932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,32,balanced,0.036389333506425224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,32,balanced,0.03775999943415324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,32,power_law_1.01,0.04318720102310181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,32,power_law_1.01,0.018534399569034576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,32,balanced,0.04178133110205332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,32,power_law_1.01,0.042694398760795595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,32,balanced,0.04010133445262909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,32,power_law_1.01,0.04344959855079651
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,32,balanced,0.06195733447869619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,32,power_law_1.01,0.020326399803161622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,32,balanced,0.04358399907747904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,32,power_law_1.01,0.043673598766326906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,32,power_law_1.01,0.018508799374103546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,32,balanced,0.043807998299598694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,32,power_law_1.01,0.04493440091609955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,32,power_law_1.01,0.018572799861431122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,32,power_law_1.01,0.046060800552368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,32,balanced,0.04758933186531067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,32,power_law_1.01,0.018400000035762788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,32,balanced,0.22617600361506143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,32,power_law_1.01,0.04536960124969482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,32,power_law_1.01,0.019091199338436126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,32,balanced,0.43742398420969647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,32,balanced,0.05067733426888784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,32,power_law_1.01,0.04930559992790222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,32,power_law_1.01,0.01973759979009628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,32,balanced,0.08004799981911977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,32,power_law_1.01,0.049721598625183105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,32,power_law_1.01,0.02184319943189621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,32,balanced,0.05820799867312113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,32,power_law_1.01,0.05310080051422119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,32,power_law_1.01,0.022336000204086305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,32,power_law_1.01,0.05614719986915588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,32,power_law_1.01,0.022886399924755097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.01,0.060844802856445314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,32,power_law_1.01,0.03272320032119751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,32,balanced,0.07675200204054515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.01,0.0718720018863678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,32,power_law_1.01,0.03409920036792755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.01,0.0839680016040802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,32,power_law_1.01,0.032601600885391234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,32,balanced,0.09672000010808308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.01,0.1004032015800476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,32,power_law_1.01,0.03256320059299469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,32,balanced,0.09761599699656169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,32,power_law_1.01,0.034835198521614076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.01,0.11958400011062623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,32,power_law_1.01,0.03447040021419525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.01,0.15868159532546997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,32,power_law_1.01,0.03683840036392212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.01,0.195961594581604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,32,power_law_1.01,0.04071039855480194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.01,0.30429439544677733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.01,0.04345600008964538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,32,balanced,0.12776000301043192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.01,0.38543360233306884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.01,0.04896000027656555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.01,0.05421440005302429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,32,power_law_1.01,0.4911935806274414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.01,0.07744640111923218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,32,power_law_1.01,0.7669824123382568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.01,0.09137920141220093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,32,balanced,0.1165120005607605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.01,0.1137984037399292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,32,power_law_1.01,1.406707191467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.01,0.1495296001434326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,32,balanced,0.16337600350379944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.01,0.2024319887161255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.01,0.28032000064849855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,32,balanced,0.32495999336242676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,32,power_law_1.01,0.3213248014450073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,32,balanced,0.6960266431172689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,32,power_law_1.01,0.48784642219543456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,32,power_law_1.01,0.9726655960083008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,32,balanced,0.2189226746559143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,32,balanced,0.14497599999109903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,32,balanced,0.27429866790771484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,32,balanced,0.1634666621685028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,32,power_law_1.01,0.03617919981479645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,32,balanced,0.33021867275238037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,32,power_law_1.01,0.030585598945617676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,32,power_law_1.01,0.029919999837875366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,32,power_law_1.01,0.02988159954547882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,32,power_law_1.01,0.032102400064468385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,32,power_law_1.01,0.032767999172210696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,32,power_law_1.01,0.03288959860801697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,32,power_law_1.01,0.03270399868488312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,32,power_law_1.01,0.03391999900341034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,32,power_law_1.01,0.03322240114212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,32,power_law_1.01,0.03492479920387268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,32,power_law_1.01,0.035692799091339114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,32,power_law_1.01,0.036083200573921205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,32,power_law_1.01,0.036908799409866334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,32,balanced,0.5981599887212118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,32,power_law_1.01,0.040659201145172116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,32,power_law_1.01,0.041791999340057374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,32,balanced,0.22524799903233847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,32,balanced,1.3672000567118328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,32,power_law_1.01,0.04325119853019714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,32,power_law_1.01,0.04716159999370575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,32,power_law_1.01,0.056383997201919556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,32,power_law_1.01,0.063264000415802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,32,power_law_1.01,0.07523840069770812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,32,power_law_1.01,0.08762239813804626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,32,power_law_1.01,0.0970304012298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,32,balanced,0.4865386486053467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,32,power_law_1.01,0.13351039886474608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,32,power_law_1.01,0.16532479524612426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,32,power_law_1.01,0.22590720653533936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,32,power_law_1.01,0.31693439483642577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,32,power_law_1.01,0.3708224058151245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,32,power_law_1.01,0.5803135871887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,32,power_law_1.01,0.9143936157226562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,32,power_law_1.01,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,32,power_law_1.01,0.03127039968967438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,32,power_law_1.01,0.030822399258613586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,32,power_law_1.01,0.030726400017738343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,32,power_law_1.01,0.03112959861755371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,32,power_law_1.01,0.032364800572395325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,32,power_law_1.01,0.03280639946460724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,32,power_law_1.01,0.03281919956207276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,32,power_law_1.01,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,32,power_law_1.01,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,32,power_law_1.01,0.03527039885520935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,32,power_law_1.01,0.036550399661064145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,32,power_law_1.01,0.03816959857940674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,32,power_law_1.01,0.03823359906673431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,32,power_law_1.01,0.041791999340057374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,32,power_law_1.01,0.04430719912052154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,32,power_law_1.01,0.04903680086135864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,32,power_law_1.01,0.05048959851264954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,32,power_law_1.01,0.05691519975662231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,32,power_law_1.01,0.06718080043792725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,32,power_law_1.01,0.07687039971351624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,32,power_law_1.01,0.10691200494766236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,32,power_law_1.01,0.13671679496765138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,32,power_law_1.01,0.17789440155029296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,32,power_law_1.01,0.23637120723724364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,32,balanced,0.388592004776001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,32,power_law_1.01,0.31226239204406736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,32,power_law_1.01,0.4219967842102051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,32,power_law_1.01,0.5594367980957031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,32,power_law_1.01,0.8731583595275879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,32,balanced,0.9188799858093262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,32,power_law_1.01,1.6220479965209962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,32,power_law_1.2,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,32,power_law_1.2,0.029721599817276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,32,power_law_1.2,0.02728320062160492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,32,power_law_1.2,0.03107840120792389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,32,power_law_1.2,0.030393600463867188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,32,power_law_1.2,0.03318400084972382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,32,power_law_1.2,0.03403519988059998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,32,power_law_1.2,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,32,power_law_1.2,0.03313280045986176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,32,power_law_1.2,0.03333120048046112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,32,power_law_1.2,0.03495039939880371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,32,power_law_1.2,0.03654400110244751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,32,power_law_1.2,0.035776001214981076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,32,power_law_1.2,0.03142400085926056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,32,power_law_1.2,0.03795199990272522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,32,power_law_1.2,0.04314239919185638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,32,power_law_1.2,0.030643200874328612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,32,power_law_1.2,0.04227199852466583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,32,power_law_1.2,0.030086401104927062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,32,power_law_1.2,0.045516800880432126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,32,power_law_1.2,0.03170560002326965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,32,power_law_1.2,0.050521600246429446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,32,power_law_1.2,0.03201920092105866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,32,power_law_1.2,0.05860480070114136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,32,power_law_1.2,0.03400959968566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,32,power_law_1.2,0.07267839908599853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,32,power_law_1.2,0.0336896002292633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,32,power_law_1.2,0.08024320006370544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,32,power_law_1.2,0.03425279855728149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,32,power_law_1.2,0.03432320058345795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,32,power_law_1.2,0.10700160264968872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,32,power_law_1.2,0.03513599932193756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,32,power_law_1.2,0.12517759799957276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,32,power_law_1.2,0.0367680013179779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,32,power_law_1.2,0.19794559478759766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,32,power_law_1.2,0.03816959857940674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,32,power_law_1.2,0.2393343925476074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,32,power_law_1.2,0.04037120044231415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,32,power_law_1.2,0.338809609413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,32,power_law_1.2,0.04139519929885864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,32,power_law_1.2,0.475980806350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,32,power_law_1.2,0.044921600818634035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,32,power_law_1.2,0.5516287803649902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,32,power_law_1.2,0.04733439981937408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,32,power_law_1.2,1.231935977935791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,32,power_law_1.2,0.04741120040416717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,32,power_law_1.2,0.0543936014175415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,32,power_law_1.2,1.560582447052002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,32,power_law_1.2,0.06050559878349304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,32,power_law_1.2,0.07369599938392639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,32,power_law_1.2,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,32,power_law_1.2,0.11978880167007447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,32,power_law_1.2,0.1559167981147766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,32,power_law_1.2,0.20682880878448487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,32,power_law_1.2,0.2953088045120239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,32,power_law_1.2,0.5016895771026612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,32,power_law_1.2,0.6039936065673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,32,power_law_1.2,0.0420415997505188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,32,power_law_1.2,0.018649600446224213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,32,power_law_1.2,0.7287680149078369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,32,power_law_1.2,0.036671999096870425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,32,power_law_1.2,0.0383103996515274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,32,power_law_1.2,1.3070719718933106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,32,power_law_1.2,0.02024320065975189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,32,power_law_1.2,0.040524798631668094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,32,power_law_1.2,0.018592000007629395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,32,power_law_1.2,0.040915200114250184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,32,power_law_1.2,2.659199905395508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,32,power_law_1.2,0.01860480010509491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,32,power_law_1.2,0.043756800889968875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,32,power_law_1.2,0.018745599687099455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,32,power_law_1.2,0.043750399351119997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,32,power_law_1.2,0.04411520063877106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,32,power_law_1.2,0.01966720074415207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,32,power_law_1.2,0.04503040015697479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,32,power_law_1.2,0.019840000569820403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,32,power_law_1.2,0.022291199862957002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,32,power_law_1.2,0.043296000361442565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,32,power_law_1.2,0.022278399765491487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,32,power_law_1.2,0.0451200008392334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,32,power_law_1.2,0.022918400168418885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,32,power_law_1.2,0.045388799905776975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,32,power_law_1.2,0.03253119885921478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,32,power_law_1.2,0.04611839950084686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,32,power_law_1.2,0.034118399024009705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,32,power_law_1.2,0.04647679924964905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,32,power_law_1.2,0.032927998900413515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,32,power_law_1.2,0.05133439898490906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,32,power_law_1.2,0.05156480073928833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,32,power_law_1.2,0.03240320086479187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,32,power_law_1.2,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,32,power_law_1.2,0.05570560097694397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,32,power_law_1.2,0.03460479974746704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,32,power_law_1.2,0.058310401439666745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.2,0.06555520296096802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,32,power_law_1.2,0.03623040020465851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.2,0.08030080199241638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,32,power_law_1.2,0.04047360122203827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.2,0.08748160004615783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.2,0.04306559860706329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.2,0.11539839506149292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.2,0.048588800430297854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.2,0.13990399837493897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.2,0.05664640069007874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.2,0.20901761054992676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.2,0.07859839797019959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.2,0.23995521068572997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.2,0.08997759819030762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.2,0.34885759353637696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.2,0.5387328147888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.2,0.12581759691238403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,32,power_law_1.2,0.6138304233551025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.2,0.1646399974822998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.2,0.2220992088317871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,32,power_law_1.2,1.041990375518799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.2,0.2996992111206055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,32,power_law_1.2,0.34551680088043213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,32,power_law_1.2,2.4718143463134767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,32,power_law_1.2,0.5762432098388672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,32,power_law_1.2,1.0995200157165528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,16,balanced,0.029909332593282063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,16,balanced,0.029648000995318096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,16,balanced,0.029311999678611755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,16,balanced,0.03013866643110911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,16,balanced,0.03143999973932902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,16,balanced,0.03323200096686681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,16,balanced,0.03541333228349686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,16,balanced,0.033386667569478355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,16,balanced,0.033370666205883026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,16,balanced,0.03358400116364161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,16,balanced,0.035455999275048576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,16,balanced,0.03364799916744232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,16,balanced,0.03565866748491923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,16,balanced,0.0354720006386439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,16,balanced,0.03775466730197271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,16,balanced,0.03944533318281174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,16,balanced,0.04185600082079569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,16,balanced,0.042591998974482216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,16,balanced,0.04598933458328247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,16,balanced,0.049839998284975685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,16,balanced,0.053743998209635414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,16,balanced,0.06639466683069865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,16,balanced,0.0749013324578603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,16,balanced,0.09607999523480733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,16,balanced,0.11928000052769978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,16,balanced,0.1524799962838491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,16,balanced,0.031354665756225586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,16,power_law_1.01,0.03558399975299835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,16,balanced,0.031583999594052635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,16,power_law_1.01,0.03051519989967346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,16,balanced,0.0315786674618721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,16,balanced,0.03156800071398417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,16,power_law_1.01,0.029580798745155335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,16,balanced,0.035461333890755974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,16,power_law_1.01,0.029600000381469725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,16,balanced,0.03570666660865148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,16,power_law_1.01,0.03113600015640259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,16,balanced,0.18838399648666382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,16,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,16,power_law_1.01,0.03268480002880096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,16,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,16,power_law_1.01,0.03256320059299469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,16,balanced,0.035743998984495796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,16,power_law_1.01,0.03283199965953827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,16,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,16,power_law_1.01,0.03282560110092163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,16,balanced,0.03736533224582672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,16,power_law_1.01,0.033292800188064575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,16,balanced,0.039173332353432976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,16,power_law_1.01,0.03598720133304596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,16,power_law_1.01,0.034892800450325015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,16,balanced,0.03783999880154928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,16,power_law_1.01,0.036134400963783266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,16,balanced,0.03946666667858759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,16,power_law_1.01,0.033792001008987424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,16,power_law_1.01,0.03607040047645569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,16,balanced,0.043935999274253845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,16,power_law_1.01,0.029836800694465638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,16,power_law_1.01,0.03785600066184998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,16,balanced,0.04201066493988037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,16,power_law_1.01,0.03163520097732544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,16,power_law_1.01,0.0414463996887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,16,power_law_1.01,0.03281919956207276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,16,balanced,0.04494399825731913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,16,power_law_1.01,0.04266240000724793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,16,power_law_1.01,0.03400320112705231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,16,power_law_1.01,0.04570240080356598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,16,balanced,0.0476746658484141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,16,power_law_1.01,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,16,power_law_1.01,0.04842239916324616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,16,power_law_1.01,0.03527680039405823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,16,power_law_1.01,0.05978879928588867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,16,balanced,0.05250666538874308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,16,power_law_1.01,0.035622400045394895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,16,power_law_1.01,0.07530879974365234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,16,balanced,0.22617600361506143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,16,power_law_1.01,0.035571199655532834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,16,power_law_1.01,0.07875199913978577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,16,power_law_1.01,0.03754880130290985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,16,balanced,0.05793066819508871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,16,power_law_1.01,0.10250879526138305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,16,power_law_1.01,0.038412800431251524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,16,power_law_1.01,0.1192639946937561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,16,power_law_1.01,0.039059200882911684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,16,power_law_1.01,0.038540801405906676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,16,power_law_1.01,0.1607103943824768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,16,balanced,0.06438933312892914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,16,power_law_1.01,0.04394879937171936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,16,power_law_1.01,0.21230719089508057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,16,power_law_1.01,0.04589439928531647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,16,power_law_1.01,0.26695680618286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,16,power_law_1.01,0.04830079972743988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,16,balanced,0.09009599685668945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,16,power_law_1.01,0.28592638969421386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,16,power_law_1.01,0.05219200253486633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,16,power_law_1.01,0.48504958152770994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,16,power_law_1.01,0.06097919940948486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,16,power_law_1.01,0.074508798122406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,16,power_law_1.01,0.6405888080596924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,16,power_law_1.01,0.08859519958496094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,16,balanced,0.11098133524258931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,16,power_law_1.01,0.11663999557495117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,16,power_law_1.01,1.4270591735839844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,16,power_law_1.01,0.14172159433364867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,16,power_law_1.01,0.2202303886413574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,16,power_law_1.01,0.2565887928009033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,16,power_law_1.01,0.3677824020385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,16,balanced,0.14827199776967367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,16,power_law_1.01,0.447603178024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,16,power_law_1.01,0.6138559818267822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,16,power_law_1.01,0.9453056335449219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,16,balanced,0.3221919933954875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,16,power_law_1.01,1.9994495391845704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,16,power_law_1.2,0.034534400701522826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,16,balanced,0.18916799624760947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,16,power_law_1.2,0.032467201352119446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,16,power_law_1.2,0.029183998703956604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,16,power_law_1.2,0.02887679934501648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,16,power_law_1.2,0.031200000643730165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,16,power_law_1.2,0.03226880133152008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,16,power_law_1.2,0.03252480030059814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,16,power_law_1.2,0.032652801275253295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,16,power_law_1.2,0.0324288010597229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,16,power_law_1.2,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,16,power_law_1.2,0.035359999537467955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,16,power_law_1.2,0.036032000184059144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,16,balanced,0.2547840078671773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,16,power_law_1.2,0.0367935985326767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,16,power_law_1.2,0.03743360042572021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,16,power_law_1.2,0.0438400000333786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,16,power_law_1.2,0.04312959909439087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,16,power_law_1.2,0.04640640020370483
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,16,power_law_1.2,0.05304960012435913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,16,power_law_1.2,0.06183040142059326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,16,power_law_1.2,0.08133119940757752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,16,power_law_1.2,0.09507840275764465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,16,power_law_1.2,0.11191040277481079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,16,power_law_1.2,0.13081599473953248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,16,power_law_1.2,0.20344960689544678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,16,power_law_1.2,0.23935360908508302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,16,balanced,0.3219359914461772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,16,power_law_1.2,0.3660928010940552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,16,power_law_1.2,0.520249605178833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,16,power_law_1.2,0.03717760145664215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,16,power_law_1.2,0.7379136085510254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,16,power_law_1.2,0.9866559982299805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,16,power_law_1.2,0.03326080143451691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,16,power_law_1.2,0.03127039968967438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,16,power_law_1.2,2.3651712417602537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,16,power_law_1.2,0.031673601269721983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,16,power_law_1.2,0.03351039886474609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,16,power_law_1.2,0.03524479866027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,16,power_law_1.2,0.035257598757743834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,16,power_law_1.2,0.03606399893760681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,16,power_law_1.2,0.036671999096870425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,16,power_law_1.2,0.03744640052318573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,16,power_law_1.2,0.03850240111351013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,16,power_law_1.2,0.04002560079097748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,16,balanced,0.3991413513819377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,16,power_law_1.2,0.04028159976005554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,16,power_law_1.2,0.04115839898586273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,16,power_law_1.2,0.04689919948577881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,16,power_law_1.2,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,16,power_law_1.2,0.050425601005554196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,16,power_law_1.2,0.055731201171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,16,balanced,0.5845546722412109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,16,power_law_1.2,0.06439039707183838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,16,power_law_1.2,0.07580800056457519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,16,power_law_1.2,0.09274240136146546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,16,power_law_1.2,0.14325759410858155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,16,power_law_1.2,0.17025920152664184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,16,power_law_1.2,0.25640320777893066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,16,power_law_1.2,0.30951039791107177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,16,power_law_1.2,0.46693758964538573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,16,power_law_1.2,0.6738431930541993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,16,power_law_1.2,0.8533632278442382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,16,power_law_1.2,1.2919551849365234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,16,power_law_1.2,2.6718015670776367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,16,balanced,0.5951093435287476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,8,balanced,0.03150933235883713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,8,balanced,0.031248000760873158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,8,balanced,0.031152000029881794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,8,balanced,0.031189332405726116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,8,balanced,0.035599999129772186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,8,balanced,0.03557866563399633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,8,balanced,0.03568533311287562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,8,balanced,0.03696000079313914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,8,balanced,0.035546667873859406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,8,balanced,0.037802666425704956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,8,balanced,0.037834666669368744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,8,balanced,0.0374293327331543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,8,balanced,0.037658666570981346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,8,balanced,0.03969600051641464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,8,balanced,0.04181866844495138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,8,balanced,0.0313226655125618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,8,balanced,0.041877334316571556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,8,balanced,0.044165333112080894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,8,balanced,0.032960000137488045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,8,balanced,0.03359466542800268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,8,balanced,0.04824000100294749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,8,balanced,0.033359999457995095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,8,balanced,0.0377866675456365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,8,balanced,0.05589333176612854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,8,balanced,0.03809600075085958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,8,balanced,0.03972800076007843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,8,balanced,0.03972266614437103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,8,balanced,0.06163200239340464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,8,balanced,0.0396373321612676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,8,balanced,0.039461334546407066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,8,balanced,0.03979733337958654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,8,balanced,0.06974400083223979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,8,balanced,0.0418453315893809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,8,balanced,0.04222933451334635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,8,balanced,0.04505600035190582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,8,balanced,0.04797866443792979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,8,balanced,0.08927466471989949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,8,balanced,0.049679999550183616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,8,balanced,0.0499946673711141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,8,balanced,0.05377600093682607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,8,balanced,0.10668266812960307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,16,balanced,1.1404799620310466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,8,balanced,0.05607999861240387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,8,balanced,0.07041599849859874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,8,balanced,0.0763626645008723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,8,power_law_1.01,0.03353599905967712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,8,balanced,0.13407466808954874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,8,power_law_1.01,0.029708799719810487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,8,power_law_1.01,0.02741760015487671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,8,balanced,0.11130133271217346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,8,power_law_1.01,0.030451199412345885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,8,power_law_1.01,0.0319680005311966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,8,power_law_1.01,0.03528960049152374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,8,power_law_1.01,0.03558399975299835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,8,balanced,0.13892799615859985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,8,power_law_1.01,0.03555839955806732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,8,balanced,0.16385066509246826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,8,power_law_1.01,0.03678080141544342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,8,power_law_1.01,0.03696640133857727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,8,power_law_1.01,0.03746559917926788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,8,power_law_1.01,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,8,power_law_1.01,0.03918080031871796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,8,balanced,0.18932799498240152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,8,power_law_1.01,0.041868799924850465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,8,power_law_1.01,0.04678399860858917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,8,power_law_1.01,0.04921599924564361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,8,power_law_1.01,0.05059199929237366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,8,power_law_1.01,0.05738880038261414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,8,power_law_1.01,0.07069439888000488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,8,balanced,0.2326080004374186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,8,power_law_1.01,0.08204159736633301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,8,balanced,0.23633599281311035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,8,power_law_1.01,0.09303039908409119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,8,power_law_1.01,0.030943998694419862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,8,power_law_1.01,0.11550079584121704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,8,power_law_1.01,0.1406399965286255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,8,power_law_1.01,0.029971200227737426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,8,power_law_1.01,0.19001599550247192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,8,power_law_1.01,0.030156800150871278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,8,power_law_1.01,0.24443519115447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,8,power_law_1.01,0.03251200020313263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,8,power_law_1.01,0.34675838947296145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,8,power_law_1.01,0.0333759993314743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,8,power_law_1.01,0.03628160059452057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,8,power_law_1.01,0.4286975860595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,8,power_law_1.01,0.03715200126171112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,8,power_law_1.01,0.548031997680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,8,power_law_1.01,0.03751679956912994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,8,balanced,0.3344053427378337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,8,power_law_1.01,0.851801586151123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,8,power_law_1.01,0.0385343998670578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,8,balanced,0.277402659257253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,8,power_law_1.01,0.03898240029811859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,8,power_law_1.01,0.040403199195861814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,8,power_law_1.01,1.799558448791504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,8,power_law_1.01,0.041894400119781496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,8,power_law_1.01,0.04520959854125976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,8,power_law_1.01,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,8,power_law_1.01,0.05059840083122254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,8,power_law_1.01,0.05268480181694031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,8,power_law_1.01,0.0537280023097992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,8,power_law_1.01,0.061452800035476686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,8,power_law_1.01,0.06915839910507202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,8,balanced,0.42980265617370605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,8,power_law_1.01,0.0827072024345398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,8,power_law_1.2,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,8,power_law_1.01,0.10888320207595825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,8,power_law_1.2,0.028038400411605834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,8,power_law_1.01,0.14163199663162232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,8,power_law_1.2,0.027635198831558228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,8,power_law_1.01,0.19589120149612427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,8,balanced,0.3493066628774007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,8,power_law_1.2,0.029235199093818665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,8,power_law_1.01,0.2594559907913208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,8,power_law_1.2,0.03244799971580505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,8,power_law_1.01,0.3190975904464722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,8,power_law_1.2,0.0353408008813858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,8,power_law_1.01,0.4306431770324707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,8,power_law_1.2,0.0353408008813858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,8,power_law_1.01,0.5692863941192627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,8,power_law_1.2,0.03526400029659271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,8,power_law_1.01,0.7075007915496826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,8,power_law_1.2,0.03709439933300018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,8,power_law_1.2,0.03697279989719391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,8,power_law_1.01,1.2764927864074707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,8,power_law_1.2,0.03703039884567261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,8,power_law_1.01,2.48864631652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,8,power_law_1.2,0.038649600744247434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,8,power_law_1.2,0.03992320001125336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,8,power_law_1.2,0.04245119988918304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,8,balanced,0.5280426740646362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,8,power_law_1.2,0.048767998814582825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,8,power_law_1.2,0.052147197723388675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,8,power_law_1.2,0.052774399518966675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,8,power_law_1.2,0.058790397644042966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,8,power_law_1.2,0.07256320118904114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,8,power_law_1.2,0.08863999843597412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,8,power_law_1.2,0.09927039742469787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,8,power_law_1.2,0.031404799222946166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,8,power_law_1.2,0.12182400226593018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,8,power_law_1.2,0.16280319690704345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,8,power_law_1.2,0.028415998816490172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,8,power_law_1.2,0.21512959003448487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,8,power_law_1.2,0.02967039942741394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,8,power_law_1.2,0.032671999931335446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,8,power_law_1.2,0.28767359256744385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,8,balanced,0.5144319931666056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,8,power_law_1.2,0.03387520015239716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,8,power_law_1.2,0.4779071807861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,8,power_law_1.2,0.03736959993839264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,8,power_law_1.2,0.5265024185180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,8,power_law_1.2,0.03764480054378509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,8,power_law_1.2,0.03792639970779419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,8,power_law_1.2,0.684665584564209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,8,power_law_1.2,0.038515201210975646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,8,power_law_1.2,0.9035455703735351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,8,power_law_1.2,0.03937279880046844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,8,power_law_1.2,0.041171199083328246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,8,power_law_1.2,0.04293760061264038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,8,power_law_1.2,2.9755584716796877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,8,power_law_1.2,0.04562560021877289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,8,power_law_1.2,0.047865599393844604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,8,balanced,0.8014026482899984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,8,power_law_1.2,0.052985602617263795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,8,power_law_1.2,0.05459840297698974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,8,power_law_1.2,0.054176002740859985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,8,power_law_1.2,0.06460800170898437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,8,power_law_1.2,0.07307519912719726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,8,power_law_1.2,0.0900223970413208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,8,power_law_1.2,0.10853760242462158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,8,power_law_1.2,0.1464192032814026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,8,power_law_1.2,0.1968832015991211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,8,power_law_1.2,0.2851263999938965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,8,power_law_1.2,0.37855360507965086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,8,power_law_1.2,0.5013631820678711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,8,power_law_1.2,0.8102527618408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,8,power_law_1.2,0.8795392036437988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,8,power_law_1.2,1.5079232215881349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,8,power_law_1.2,2.9604543685913085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,8,balanced,0.9603786468505859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,8,balanced,1.5573919614156086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,4,balanced,0.031258667508761086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,4,balanced,0.03362133353948593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,4,balanced,0.03324266771475474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,4,balanced,0.03548266738653183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,4,balanced,0.03972266614437103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,4,balanced,0.039813332259655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,4,balanced,0.041690667470296226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,4,balanced,0.04178666571776072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,4,balanced,0.039818666875362396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,4,balanced,0.04181333382924398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,4,balanced,0.04190933207670847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,4,balanced,0.043866669138272606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,4,balanced,0.04382933179537455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,4,balanced,0.043824002146720886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,4,balanced,0.04826666911443075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,4,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,4,balanced,0.053818667928377785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,4,balanced,0.05815466741720835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,4,balanced,0.06307733555634816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,4,balanced,0.08481599887212117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,4,balanced,0.09471999605496724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,4,balanced,0.12077333529790242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,4,balanced,0.1602720022201538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,4,balanced,0.23825067281723022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,4,balanced,0.28456000487009686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,4,balanced,0.4110879898071289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,4,balanced,0.031343999008337654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,4,balanced,0.5284586747487386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,4,balanced,0.031530665854612984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,4,balanced,0.03363733241955439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,4,balanced,0.03572266548871994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,4,balanced,0.041434665520985924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,4,balanced,0.043925335009892784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,4,balanced,0.04413333535194397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,4,balanced,0.04576533536116282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,4,balanced,0.044495999813079834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,4,balanced,0.04492799937725067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,4,balanced,0.045791998505592346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,4,balanced,0.04794133206208547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,4,balanced,0.0499839981396993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,4,balanced,0.04987733562787374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,4,balanced,0.05605866511662801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,4,balanced,0.058058664202690125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,4,balanced,0.05839466551939646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,4,balanced,0.06625600159168243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,4,balanced,0.6503146489461263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,4,balanced,0.07035199801127116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,4,balanced,0.09315733114878337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,4,power_law_1.01,0.03368319869041443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,4,balanced,0.11340266466140747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,4,power_law_1.01,0.0308351993560791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,4,power_law_1.01,0.03192960023880005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,4,power_law_1.01,0.033215999603271484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,4,balanced,0.15556800365447998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,4,power_law_1.01,0.03452799916267395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,4,power_law_1.01,0.03802880048751831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,4,power_law_1.01,0.03874559998512268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,4,power_law_1.01,0.0394239991903305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,4,power_law_1.01,0.039647999405860904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,4,power_law_1.01,0.0401856005191803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,4,balanced,0.18920000394185385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,4,power_law_1.01,0.041529598832130435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,4,power_law_1.01,0.042668798565864564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,4,power_law_1.01,0.04299519956111908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,4,power_law_1.01,0.04625279903411865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,4,power_law_1.01,0.05006080269813538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,4,power_law_1.01,0.052236801385879515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,4,power_law_1.01,0.06108800172805786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,4,balanced,0.2690879901250203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,4,power_law_1.01,0.0703935980796814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,4,power_law_1.01,0.0785215973854065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,4,power_law_1.01,0.0974399983882904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,4,power_law_1.01,0.11232000589370728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,4,balanced,1.00654935836792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,4,power_law_1.01,0.1457535982131958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,4,power_law_1.01,0.2186431884765625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,4,power_law_1.01,0.2958336114883423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,4,power_law_1.01,0.3619136095046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,4,power_law_1.01,0.5180607795715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,4,balanced,0.3377866744995117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,4,power_law_1.01,0.6532544136047364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,4,power_law_1.01,0.7609216213226319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,4,power_law_1.01,1.2961024284362792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,4,power_law_1.01,2.3128192901611326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,4,power_law_1.01,0.031116798520088196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,4,power_law_1.01,0.030425599217414855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,4,power_law_1.01,0.03172479867935181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,4,power_law_1.01,0.03473280072212219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,4,balanced,0.49262932936350506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,4,power_law_1.01,0.03601279854774475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,4,power_law_1.01,0.040870401263237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,4,power_law_1.2,0.03338879942893982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,4,power_law_1.01,0.04168320000171662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,4,power_law_1.01,0.04273920059204102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,4,power_law_1.2,0.030329599976539612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,4,power_law_1.2,0.030873599648475646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,4,power_law_1.01,0.04355199933052063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,4,power_law_1.2,0.03281280100345611
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,4,power_law_1.01,0.04485119879245758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,4,power_law_1.2,0.03381119966506958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,4,power_law_1.01,0.04718720018863678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,4,power_law_1.2,0.03760640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,32,balanced,0.056176001826922096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,4,power_law_1.01,0.0495743989944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,4,power_law_1.2,0.03874559998512268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,4,power_law_1.01,0.051686400175094606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,32,balanced,0.04602666695912679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,4,power_law_1.01,0.05493119955062866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,32,balanced,0.0476746658484141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,4,power_law_1.2,0.0394239991903305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,4,power_law_1.01,0.06108160018920898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,32,balanced,0.04770133395989736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,4,power_law_1.2,0.039392000436782836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,4,power_law_1.01,0.06330239772796631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,32,balanced,0.05178666611512502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,4,power_law_1.01,0.06407679915428162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,4,power_law_1.2,0.039084801077842714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,32,balanced,0.06090133388837179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,4,power_law_1.01,0.07093120217323304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,4,power_law_1.2,0.04151679873466492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,4,balanced,0.640501340230306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,4,power_law_1.01,0.08375040292739869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,32,balanced,0.06201600035031637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,4,power_law_1.01,0.10736639499664306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,4,power_law_1.2,0.041843199729919435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,32,balanced,0.04571199913819631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,4,power_law_1.2,0.04320000112056732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,4,power_law_1.01,0.13406720161437988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,32,balanced,0.06439466774463654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,4,power_law_1.2,0.04588159918785095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,4,power_law_1.01,0.1794816017150879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,32,balanced,0.045754666129748024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,32,balanced,0.06436266501744588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,4,power_law_1.01,0.22512640953063964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,4,power_law_1.2,0.052723199129104614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,32,balanced,0.04587733248869578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,32,balanced,0.06397333244482677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,4,power_law_1.01,0.3143232107162476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,4,power_law_1.2,0.05569919943809509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,32,balanced,0.04562666515509287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,4,balanced,1.859242598215739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,32,balanced,0.0666293352842331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,4,balanced,0.7976693312327067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,4,power_law_1.01,0.4060671806335449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,4,power_law_1.01,0.5941887855529785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,32,balanced,0.06783466537793477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,32,balanced,0.047775998711586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,4,power_law_1.2,0.06033920049667359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,32,balanced,0.06741333504517873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,4,power_law_1.01,0.7729343891143798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,32,balanced,0.05008533100287119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,4,power_law_1.2,0.07004799842834472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,4,power_law_1.2,0.03165439963340759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,32,balanced,0.06835733354091644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,4,power_law_1.01,1.0472000122070313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,32,balanced,0.05004799862702688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,32,balanced,0.07433066765467326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,4,power_law_1.2,0.07803519964218139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,4,power_law_1.01,1.7254720687866212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,32,balanced,0.05216533442338308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,4,power_law_1.2,0.030630400776863097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,32,balanced,0.07357866565386455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,4,power_law_1.2,0.09984639883041382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,32,balanced,0.05202133456865946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,4,power_law_1.01,3.0938751220703127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,4,power_law_1.2,0.03201279938220978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,32,balanced,0.052154665191968284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,4,power_law_1.2,0.11709439754486084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,32,balanced,0.0745066652695338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,4,power_law_1.2,0.034431999921798705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,32,balanced,0.05426666637261709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,4,power_law_1.2,0.15326080322265626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,32,balanced,0.0860533316930135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,4,power_law_1.2,0.03585920035839081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,32,balanced,0.055215999484062195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,4,power_law_1.2,0.2217087984085083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,4,power_law_1.2,0.040761598944664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,32,balanced,0.08456533153851827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,32,balanced,0.0539680023988088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,4,power_law_1.2,0.04181120097637177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,4,power_law_1.2,0.33679358959197997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,32,balanced,0.058320000767707825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,4,balanced,1.2452747027079265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,4,power_law_1.2,0.04351359903812409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,32,balanced,0.095551997423172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,32,balanced,0.0601440022389094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,4,power_law_1.2,0.3954240083694458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,4,power_law_1.2,0.043993601202964784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,4,power_law_1.2,0.04650880098342895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,32,balanced,0.0611413319905599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,4,power_law_1.2,0.5622591972351074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,4,power_law_1.2,0.04948480129241943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,32,balanced,0.09923733274141948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,32,balanced,0.06472533444563548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,4,power_law_1.2,0.759993600845337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,4,power_law_1.2,0.051072001457214355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,32,balanced,0.07939200103282928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,4,power_law_1.2,1.1029439926147462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,4,power_law_1.2,0.053273600339889524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,4,power_law_1.2,0.05686399936676025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,32,balanced,0.12484799822171529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,4,power_law_1.2,1.5056447982788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,32,balanced,0.07563733557860057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,4,power_law_1.2,0.062118399143219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,4,power_law_1.2,0.0661952018737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,4,power_law_1.2,2.775103950500488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,32,balanced,0.10822932918866475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,4,power_law_1.2,0.06531199812889099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,4,power_law_1.2,0.07520639896392822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,32,balanced,0.1359999974568685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,4,power_law_1.2,0.0880511999130249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,32,balanced,0.11563199758529663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,4,power_law_1.2,0.11546239852905274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,4,power_law_1.2,0.14325759410858155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,32,balanced,0.15121066570281982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,4,power_law_1.2,0.18861440420150757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,32,balanced,0.18897066513697305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,4,power_law_1.2,0.24344959259033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,4,power_law_1.2,0.3322495937347412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,32,balanced,0.18133334318796793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,4,power_law_1.2,0.46367359161376953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,4,power_law_1.2,0.690009593963623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,4,power_law_1.2,0.929695987701416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,32,balanced,0.1998080015182495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,32,balanced,0.243066668510437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,4,power_law_1.2,1.1183744430541993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,4,power_law_1.2,1.7078208923339844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,4,power_law_1.2,3.996166229248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,32,balanced,0.29055466254552204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,32,balanced,0.29453333218892414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,32,balanced,0.4002186854680379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,4,balanced,2.416698614756266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,32,balanced,0.3314453363418579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,32,balanced,0.5094506740570068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,32,balanced,0.38676265875498456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,32,balanced,0.616320013999939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,32,balanced,0.5722453196843466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,32,balanced,0.04814399778842926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,32,balanced,0.05022933085759481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,32,balanced,0.04852266609668732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,32,balanced,0.05284800132115682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,32,balanced,0.05384000142415365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,32,balanced,0.9434026877085367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,32,balanced,0.05416533350944519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,32,balanced,0.05564799904823303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,32,balanced,0.05619733532269796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,32,balanced,0.05619200070699056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,32,balanced,0.05628266433874766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,32,balanced,0.058229332168896995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,32,balanced,0.056143999099731445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,32,balanced,0.05825600028038025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,32,balanced,0.05993066728115082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,32,balanced,0.06400533517201741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,32,balanced,0.06279466549555461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,32,balanced,0.023354666928450268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,32,balanced,0.06647466619809468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,32,balanced,0.02327466756105423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,32,balanced,0.07051733136177063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,32,power_law_1.01,0.08419200181961059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,32,balanced,0.023226665953795116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,32,balanced,0.023189333577950794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,32,balanced,0.07467733323574066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,32,balanced,0.024847999215126038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,32,balanced,0.029477333029111225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,32,power_law_1.01,0.056441599130630495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,32,balanced,0.09090666969617207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,32,balanced,0.0506879985332489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,32,balanced,0.050026665131251015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,32,power_law_1.01,0.047891199588775635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,32,balanced,0.05000533163547516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,32,power_law_1.01,0.051475197076797485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,32,balanced,0.10379733641942342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,32,balanced,0.04675200084845225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,32,balanced,0.047797332207361855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,32,power_law_1.01,0.0547327995300293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,32,balanced,0.04675200084845225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,32,power_law_1.01,0.057126402854919434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,32,balanced,0.04402133325735728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,32,balanced,0.13250133395195007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,32,balanced,0.04366933306058248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,32,power_law_1.01,0.06236799955368042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,32,balanced,0.05148266752560934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,32,power_law_1.01,0.06371200084686279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,32,balanced,0.05526400109132131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,32,balanced,0.14809067050615946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,32,power_law_1.01,0.05923839807510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,32,balanced,0.05003733436266581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,32,power_law_1.01,0.06272000074386597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,32,balanced,0.05499733487764994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,32,power_law_1.01,0.06581760048866273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,32,balanced,0.07473599910736084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,32,power_law_1.01,0.06823679804801941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,32,balanced,0.08707200487454732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,32,balanced,0.18875199556350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,32,power_law_1.01,0.0680512011051178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,32,balanced,0.1093386709690094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,32,balanced,1.07423996925354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,32,power_law_1.01,0.07434239983558655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,32,power_law_1.01,0.07733759880065919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,32,power_law_1.01,0.07920640110969543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,32,balanced,0.14231466253598532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,32,power_law_1.01,0.08407679796218873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,32,balanced,0.22180267175038657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,32,power_law_1.01,0.09918720126152039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,32,power_law_1.01,0.09572479724884034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,32,power_law_1.01,0.11331199407577515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,32,balanced,0.17867734034856161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,32,power_law_1.01,0.1291584014892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,32,power_law_1.01,0.1555392026901245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,32,power_law_1.01,0.18729599714279174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,32,balanced,1.8307466506958008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,32,power_law_1.01,0.2577663898468018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,32,balanced,0.25280533234278363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,32,power_law_1.01,0.04863359928131104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,32,power_law_1.01,0.3251904010772705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,32,balanced,0.3142079909642537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,32,power_law_1.01,0.45546879768371584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,32,power_law_1.01,0.042668798565864564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,32,power_law_1.01,0.5364927768707275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,32,power_law_1.01,0.04040960073471069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,32,power_law_1.01,0.6703999996185303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,32,power_law_1.01,0.043724799156188966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,32,balanced,0.3240586717923482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,32,power_law_1.01,0.04451839923858643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,32,power_law_1.01,1.1364543914794922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,32,power_law_1.01,0.0473471999168396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,32,power_law_1.01,0.04776960015296936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,32,power_law_1.01,2.2760383605957033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,32,power_law_1.01,0.04958719909191132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,32,power_law_1.01,0.04986880123615265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,32,power_law_1.01,0.049772799015045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,32,power_law_1.01,0.04831359982490539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,32,power_law_1.01,0.05070080161094666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,32,power_law_1.01,0.0536191999912262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,32,power_law_1.01,0.04663679897785187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,32,power_law_1.01,0.05427839756011963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,32,power_law_1.01,0.04966399967670441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,32,balanced,0.3744586706161499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,32,power_law_1.01,0.05543680191040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,32,power_law_1.01,0.050323200225830075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,32,balanced,0.4607253472010295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,32,power_law_1.01,0.060262399911880496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,32,power_law_1.01,0.05204480290412903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,32,power_law_1.01,0.06625279784202576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,32,power_law_1.01,0.05263360142707825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,32,power_law_1.01,0.07070080041885377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,32,power_law_1.01,0.054451197385787964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,32,power_law_1.01,0.07470719814300537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,32,power_law_1.01,0.05448319911956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,32,power_law_1.01,0.08959360122680664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,32,power_law_1.01,0.054425597190856934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,32,power_law_1.01,0.09523839950561523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,32,power_law_1.01,0.055257600545883176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,32,power_law_1.01,0.1219648003578186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,32,power_law_1.01,0.056652802228927615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,32,power_law_1.01,0.13907840251922607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,32,power_law_1.01,0.057734400033950806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,32,power_law_1.01,0.17637120485305785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,32,power_law_1.01,0.0595583975315094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,32,power_law_1.01,0.22627840042114258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,32,power_law_1.01,0.0646399974822998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,32,power_law_1.01,0.3029632091522217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,32,power_law_1.01,0.0674239993095398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,32,power_law_1.01,0.06862080097198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,32,power_law_1.01,0.3823744058609009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,32,balanced,0.6145120064417521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,32,power_law_1.01,0.07959039807319641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,32,power_law_1.01,0.5396096229553222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,32,balanced,0.4482453266779582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.01,0.08712319731712341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,32,power_law_1.01,0.7077631950378418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.01,0.10633599758148193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,32,power_law_1.01,0.9060544013977051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,32,power_law_1.01,0.021644799411296843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.01,0.1263167977333069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.01,0.16030720472335816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,32,power_law_1.01,1.3548864364624023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,32,power_law_1.01,0.021350400149822236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.01,0.19544320106506347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,32,power_law_1.01,0.021184000372886657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.01,0.24728960990905763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,32,power_law_1.01,2.605459213256836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,32,power_law_1.01,0.021209600567817687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.01,0.32160000801086425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,32,power_law_1.01,0.023321600258350374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.01,0.453715181350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,32,power_law_1.01,0.02826879918575287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.01,0.5871679782867432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,32,power_law_1.01,0.04937599897384644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,32,power_law_1.01,0.6579264163970947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,32,power_law_1.01,0.04990079998970032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,32,power_law_1.01,1.0633536338806153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,32,power_law_1.01,0.04981760084629059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,32,balanced,0.7605226834615072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,32,power_law_1.01,2.3763071060180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,32,power_law_1.01,0.046367999911308286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,32,power_law_1.01,0.04755840003490448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,32,power_law_1.01,0.0472896009683609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,32,power_law_1.01,0.046540799736976626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,32,power_law_1.01,0.047007998824119566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,32,power_law_1.01,0.04542079865932465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,32,power_law_1.01,0.047328001260757445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,32,balanced,0.677344004313151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,32,power_law_1.01,0.04741120040416717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,32,power_law_1.01,0.04744960069656372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.01,0.05095040202140808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.01,0.06465920209884643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.01,0.08379520177841186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.01,0.09963520169258118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.01,0.12168960571289063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.01,0.1597440004348755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.01,0.20778241157531738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.01,0.2843712091445923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.01,0.39847679138183595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,32,power_law_1.01,0.4677055835723877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,32,balanced,1.2047733465830486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,32,power_law_1.01,0.7900991916656495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,32,power_law_1.01,1.4617792129516602
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,32,balanced,1.3123946984608967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,4,32,power_law_1.2,0.08357759714126586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,4,32,power_law_1.2,0.04644480049610138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,4,32,power_law_1.2,0.047040000557899475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,4,32,power_law_1.2,0.05175679922103882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,32,balanced,2.380805333455404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,4,32,power_law_1.2,0.05008000135421753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,4,32,power_law_1.2,0.0576960027217865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,4,32,power_law_1.2,0.05839359760284424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,4,32,power_law_1.2,0.05144320130348205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,4,32,power_law_1.2,0.0637503981590271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,4,32,power_law_1.2,0.06318719983100891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,4,32,power_law_1.2,0.06323199868202209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,4,32,power_law_1.2,0.0450111985206604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,4,32,power_law_1.2,0.06580479741096497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,4,32,power_law_1.2,0.04302079975605011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,4,32,power_law_1.2,0.06898559927940369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,4,32,power_law_1.2,0.06823679804801941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,4,32,power_law_1.2,0.046028798818588255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,4,32,power_law_1.2,0.04686079919338226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,4,32,power_law_1.2,0.0738048017024994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,4,32,power_law_1.2,0.049132800102233885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,4,32,power_law_1.2,0.07863680124282837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,4,32,power_law_1.2,0.0502016007900238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,4,32,power_law_1.2,0.08074880242347718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,4,32,power_law_1.2,0.05143679976463318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,4,32,power_law_1.2,0.08789759874343872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,6144,2560,8,160,4,32,power_law_1.2,0.0517632007598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,4,32,power_law_1.2,0.052832001447677614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,4,32,power_law_1.2,0.10145280361175538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,4,32,power_law_1.2,0.0537663996219635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,6144,2560,8,160,4,32,power_law_1.2,0.04938240051269531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,4,32,power_law_1.2,0.10364160537719727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,6144,2560,8,160,4,32,power_law_1.2,0.048895999789237976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,4,32,power_law_1.2,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,4,32,power_law_1.2,0.12039680480957031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,6144,2560,8,160,4,32,power_law_1.2,0.052185600996017455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,4,32,power_law_1.2,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,4,32,power_law_1.2,0.13968000411987305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,6144,2560,8,160,4,32,power_law_1.2,0.052665597200393675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,4,32,power_law_1.2,0.056364798545837404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,6144,2560,8,160,4,32,power_law_1.2,0.0543936014175415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,4,32,power_law_1.2,0.18494720458984376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,6144,2560,8,160,4,32,power_law_1.2,0.054579198360443115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,4,32,power_law_1.2,0.06367999911308289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,4,32,power_law_1.2,0.23123838901519775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,6144,2560,8,160,4,32,power_law_1.2,0.05600000023841858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,4,32,power_law_1.2,0.06906880140304565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,6144,2560,8,160,4,32,power_law_1.2,0.055769598484039305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,4,32,power_law_1.2,0.3066431999206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,4,32,power_law_1.2,0.07247999906539918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,6144,2560,8,160,4,32,power_law_1.2,0.05596799850463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,4,32,power_law_1.2,0.4292287826538086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,6144,2560,8,160,4,32,power_law_1.2,0.057017600536346434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,4,32,power_law_1.2,0.08193280100822449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,6144,2560,8,160,4,32,power_law_1.2,0.058348798751831056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,4,32,power_law_1.2,0.5879039764404297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,4,32,power_law_1.2,0.09792640209197997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,6144,2560,8,160,4,32,power_law_1.2,0.05846400260925293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,4,32,power_law_1.2,0.10517760515213012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,6144,2560,8,160,4,32,power_law_1.2,0.06110079884529114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,4,32,power_law_1.2,0.7345536231994629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,6144,2560,8,160,4,32,power_law_1.2,0.06700800061225891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,4,32,power_law_1.2,0.1369279980659485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,6144,2560,8,160,4,32,power_law_1.2,0.06837120056152343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,4,32,power_law_1.2,0.8445247650146485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,4,32,power_law_1.2,0.1583295941352844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,6144,2560,8,160,4,32,power_law_1.2,0.07246720194816589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,6144,2560,8,160,4,32,power_law_1.2,0.08621439933776856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,4,32,power_law_1.2,0.19721599817276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,4,32,power_law_1.2,1.558016014099121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.2,0.09523199796676636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,4,32,power_law_1.2,0.2358720064163208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.2,0.12688000202178956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.2,0.1415743947029114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,4,32,power_law_1.2,2.8511871337890624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,4,32,power_law_1.2,0.3619584083557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.2,0.17457280158996583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,4,32,power_law_1.2,0.4868351936340332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.2,0.22138240337371826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,4,32,power_law_1.2,0.6704512119293213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.2,0.2845184087753296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.2,0.39416320323944093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,4,32,power_law_1.2,0.9261311531066895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.2,0.5487616062164307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,4,32,power_law_1.2,1.0108799934387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.2,0.8457599639892578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,6144,2560,8,160,4,32,power_law_1.2,0.9641280174255371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,4,32,power_law_1.2,1.9498367309570312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,6144,2560,8,160,4,32,power_law_1.2,1.4247296333312989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,4,32,power_law_1.2,3.7406654357910156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,6144,2560,8,160,4,32,power_law_1.2,2.860639953613281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,6144,2560,8,160,4,32,power_law_1.2,0.021459199488162994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,6144,2560,8,160,4,32,power_law_1.2,0.020505599677562714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,6144,2560,8,160,4,32,power_law_1.2,0.021171200275421142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,6144,2560,8,160,4,32,power_law_1.2,0.021241599321365358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,6144,2560,8,160,4,32,power_law_1.2,0.022777600586414336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,6144,2560,8,160,4,32,power_law_1.2,0.02901119887828827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,6144,2560,8,160,4,32,power_law_1.2,0.04927999973297119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,6144,2560,8,160,4,32,power_law_1.2,0.049958398938179015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,6144,2560,8,160,4,32,power_law_1.2,0.04990079998970032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,6144,2560,8,160,4,32,power_law_1.2,0.04807040095329285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,6144,2560,8,160,4,32,power_law_1.2,0.04894720017910004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,6144,2560,8,160,4,32,power_law_1.2,0.047635200619697574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,6144,2560,8,160,4,32,power_law_1.2,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,6144,2560,8,160,4,32,power_law_1.2,0.04635519981384277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,6144,2560,8,160,4,32,power_law_1.2,0.04570240080356598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,6144,2560,8,160,4,32,power_law_1.2,0.04620159864425659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,6144,2560,8,160,4,32,power_law_1.2,0.048153600096702574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,6144,2560,8,160,4,32,power_law_1.2,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.2,0.05325440168380737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.2,0.0739135980606079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.2,0.08376960158348083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.2,0.10879360437393189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.2,0.12774399518966675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.2,0.18469120264053346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.2,0.23313279151916505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.2,0.32868480682373047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.2,0.46120319366455076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,6144,2560,8,160,4,32,power_law_1.2,0.5456960201263428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,6144,2560,8,160,4,32,power_law_1.2,0.851257610321045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,6144,2560,8,160,4,32,power_law_1.2,1.9146560668945312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,16,balanced,0.040576001008351646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,16,balanced,0.04369066655635834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,16,balanced,0.043466667334238686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,16,balanced,0.04553066690762838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,16,balanced,0.05558399856090546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,16,balanced,0.06417066852251689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,16,balanced,0.06357333560784657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,16,balanced,0.0655626654624939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,16,balanced,0.06563200056552887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,16,balanced,0.06602666775385539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,16,balanced,0.06540800134340923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,16,balanced,0.06710400183995564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,16,balanced,0.06761066615581512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,16,balanced,0.06979733208815257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,16,balanced,0.07317866881688435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,16,balanced,0.07577066620190938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,16,balanced,0.0767626663049062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,16,balanced,0.0849173367023468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,16,balanced,0.08483200271924336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,16,balanced,0.10667199889818828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,16,balanced,0.10981333255767822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,16,balanced,0.14457600315411887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,16,balanced,0.14948266744613647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,16,balanced,0.21171200275421143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,16,balanced,0.23943465948104858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,16,balanced,0.0458133320013682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,16,balanced,0.04567466676235199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,16,balanced,0.04633066554864248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,16,balanced,0.046394666035970054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,16,balanced,0.04993066688378652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,16,balanced,0.053770666321118675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,16,balanced,0.05384533107280731
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,16,balanced,0.053946668903032936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,16,balanced,0.05589866638183594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,16,balanced,0.05584533512592316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,16,balanced,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,16,balanced,0.05791999896367391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,16,balanced,0.32928532361984253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,16,balanced,0.05808533231417338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,16,balanced,0.060165335734685264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,16,balanced,0.06426666676998138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,16,balanced,0.06410666803518932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,16,balanced,0.06753600140412648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,16,balanced,0.0755626658598582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,16,balanced,0.0815413345893224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,16,balanced,0.11503466963768005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,16,balanced,0.13084266583124796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,16,power_law_1.01,0.05382400155067444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,16,balanced,0.3994293212890625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,16,power_law_1.01,0.058316802978515624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,16,balanced,0.17323199907938638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,16,power_law_1.01,0.047443199157714847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,16,power_law_1.01,0.04860160052776337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,16,power_law_1.01,0.05211520195007324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,16,power_law_1.01,0.05613440275192261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,16,balanced,0.1970026691754659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,16,power_law_1.01,0.05749120116233826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,16,power_law_1.01,0.05950719714164734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,16,power_law_1.01,0.06223359704017639
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,16,power_law_1.01,0.061299198865890504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,16,power_law_1.01,0.05107839703559876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,16,power_law_1.01,0.062745600938797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,16,power_law_1.01,0.06639999747276307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,16,balanced,0.27291733026504517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,16,power_law_1.01,0.05627520084381103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,16,power_law_1.01,0.06928640007972717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,16,power_law_1.01,0.04428800046443939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,16,power_law_1.01,0.07355520129203796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,16,power_law_1.01,0.04633600115776062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,16,power_law_1.01,0.07896959781646729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,16,power_law_1.01,0.04688000082969666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,16,power_law_1.01,0.08236160278320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,16,power_law_1.01,0.050342398881912234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,16,power_law_1.01,0.08577280044555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,16,power_law_1.01,0.05079039931297302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,16,power_law_1.01,0.09571840167045594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,16,power_law_1.01,0.05261440277099609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,16,power_law_1.01,0.09713919758796692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,16,power_law_1.01,0.05248000025749207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,16,power_law_1.01,0.11416319608688355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,16,balanced,0.32279467582702637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,16,power_law_1.01,0.05375999808311462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,16,balanced,0.46697600682576496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,16,power_law_1.01,0.14179840087890624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,16,power_law_1.01,0.17710720300674437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,16,power_law_1.01,0.05580800175666809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,16,power_law_1.01,0.05724800229072571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,16,power_law_1.01,0.20429439544677735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,16,power_law_1.01,0.06114559769630432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,16,power_law_1.01,0.2730432033538818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,16,power_law_1.01,0.06496000289916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,16,power_law_1.01,0.33060479164123535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,16,power_law_1.01,0.07043840289115906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,16,power_law_1.01,0.46213760375976565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,16,power_law_1.01,0.0757695972919464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,16,power_law_1.01,0.5890431880950928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,16,power_law_1.01,0.07440000176429748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,16,power_law_1.01,0.7896383762359619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,16,power_law_1.01,0.09039999842643738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,16,power_law_1.01,1.1804351806640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,16,balanced,0.4588640133539836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,16,power_law_1.01,0.09894400238990783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,16,power_law_1.01,2.1965120315551756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,16,power_law_1.01,0.12424319982528687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,16,power_law_1.01,0.15025919675827026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,16,power_law_1.01,0.1992192029953003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,16,power_law_1.01,0.2507328033447266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,16,power_law_1.01,0.324780797958374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,16,power_law_1.01,0.4182784080505371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,16,power_law_1.01,0.6105535984039306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,16,power_law_1.01,0.7910848140716553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,16,power_law_1.01,0.9091648101806641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,16,power_law_1.01,1.5052736282348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,16,power_law_1.2,0.059443199634552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,16,balanced,0.5803786516189575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,16,power_law_1.01,3.0000511169433595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,16,power_law_1.2,0.04992640018463135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,16,power_law_1.2,0.049158400297164916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,16,balanced,0.7017599741617838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,16,power_law_1.2,0.05138559937477112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,16,power_law_1.2,0.052288001775741576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,16,power_law_1.2,0.06090239882469177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,16,power_law_1.2,0.06231679916381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,16,power_law_1.2,0.06275839805603027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,16,power_law_1.2,0.06279680132865906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,16,power_law_1.2,0.06483200192451477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,16,power_law_1.2,0.06715520024299622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,16,power_law_1.2,0.06832640171051026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,16,power_law_1.2,0.07093120217323304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,16,power_law_1.2,0.07618560194969178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,16,balanced,0.7025706768035889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,16,power_law_1.2,0.08290560245513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,16,power_law_1.2,0.0525056004524231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,16,power_law_1.2,0.08323839902877808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,16,power_law_1.2,0.08528640270233154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,16,power_law_1.2,0.047142401337623596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,16,power_law_1.2,0.10278400182723998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,16,power_law_1.2,0.045158401131629944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,16,power_law_1.2,0.1082111954689026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,16,power_law_1.2,0.045849600434303285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,16,power_law_1.2,0.12425600290298462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,16,power_law_1.2,0.0471231997013092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,16,power_law_1.2,0.14293760061264038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,16,power_law_1.2,0.050911998748779295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,16,power_law_1.2,0.19127680063247682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,16,power_law_1.2,0.05214080214500427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,16,power_law_1.2,0.23928959369659425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,16,power_law_1.2,0.05358080267906189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,16,power_law_1.2,0.32434558868408203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,16,power_law_1.2,0.05377920269966126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,16,power_law_1.2,0.4151872158050537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,16,power_law_1.2,0.055731201171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,16,power_law_1.2,0.057657599449157715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,16,power_law_1.2,0.6000639915466308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,16,power_law_1.2,0.060115200281143186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,16,power_law_1.2,0.7081344127655029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,16,power_law_1.2,0.06310399770736694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,8,balanced,0.045594667394955955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,16,power_law_1.2,0.9704319953918457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,16,power_law_1.2,0.06814720034599304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,8,balanced,0.04778666794300079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,16,power_law_1.2,0.07685760259628296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,8,balanced,0.04831466575463613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,16,power_law_1.2,1.7730047225952148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,16,power_law_1.2,0.07823359966278076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,8,balanced,0.049733335773150124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,16,power_law_1.2,0.0796288013458252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,8,balanced,0.057999998331069946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,16,power_law_1.2,0.096288001537323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,8,balanced,0.0682239979505539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,16,power_law_1.2,3.2133056640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,8,balanced,0.06886933247248332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,16,power_law_1.2,0.10526080131530761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,8,balanced,0.07032000025113423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,16,power_law_1.2,0.13045120239257812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,16,balanced,1.077888011932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,8,balanced,0.07215466598669688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,16,power_law_1.2,0.16019840240478517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,8,balanced,0.07131200035413106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,16,power_law_1.2,0.21641600131988525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,8,balanced,0.07195200026035309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,16,power_law_1.2,0.2707135915756226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,8,balanced,0.07217599948247273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,8,balanced,0.07361066838105519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,16,power_law_1.2,0.37396481037139895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,8,balanced,0.07669866581757863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,16,power_law_1.2,0.4886591911315918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,8,balanced,0.07914666831493378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,8,balanced,0.08112533390522003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,16,power_law_1.2,0.7189055919647217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,8,balanced,0.08397333820660909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,16,power_law_1.2,0.9126144409179687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,16,balanced,1.3372693061828613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,8,balanced,0.09384000301361084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,16,power_law_1.2,1.0947327613830566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,8,balanced,0.09545066952705383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,16,power_law_1.2,1.7016063690185548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,8,balanced,0.11930666367212932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,16,power_law_1.2,4.179110336303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,8,balanced,0.12643200159072876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,8,balanced,0.16680532693862915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,8,balanced,0.18869332472483316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,8,balanced,0.26479466756184894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,8,balanced,0.31038399537404376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,8,balanced,0.044309332966804504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,8,balanced,0.0440586656332016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,8,balanced,0.046623999873797096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,16,balanced,2.109210650126139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,8,balanced,0.04766400158405304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,8,balanced,0.05271466573079427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,8,balanced,0.05823466678460439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,8,balanced,0.0582239975531896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,8,balanced,0.06001066664854685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,8,balanced,0.06029866635799408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,8,balanced,0.06016000111897787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,8,balanced,0.06225599845250448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,8,balanced,0.42075200875600177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,8,balanced,0.06098666787147522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,8,balanced,0.0643146683772405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,8,balanced,0.0660693347454071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,8,balanced,0.07015466690063477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,8,balanced,0.07251200079917908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,8,balanced,0.07461333274841309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,8,balanced,0.08666666348775227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,8,balanced,0.09697066744168599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,8,balanced,0.12980799873669943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,8,balanced,0.533402681350708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,8,balanced,0.1506613294283549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,8,balanced,0.20287466049194336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,8,balanced,0.22963200012842813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,8,balanced,0.3288639982541402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,8,power_law_1.01,0.04955520033836365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,8,power_law_1.01,0.047065600752830505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,8,power_law_1.01,0.04356479942798615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,8,balanced,0.647653341293335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,8,power_law_1.01,0.052127999067306516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,8,power_law_1.01,0.05416319966316223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,8,power_law_1.01,0.06069759726524353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,8,balanced,0.39907201131184894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,8,power_law_1.01,0.06120960116386413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,8,power_law_1.01,0.06494719982147217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,8,power_law_1.01,0.06451200246810913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,8,power_law_1.01,0.0677183985710144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,8,power_law_1.01,0.06753280162811279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,8,power_law_1.01,0.07075200080871583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,8,power_law_1.01,0.07301759719848633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,8,balanced,0.5650346676508585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,8,power_law_1.01,0.0756608009338379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,8,power_law_1.01,0.0823743999004364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,8,power_law_1.01,0.09038079977035522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,8,power_law_1.01,0.0882752001285553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,8,power_law_1.01,0.04540160000324249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,8,power_law_1.01,0.09956480264663696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,8,power_law_1.01,0.10841599702835084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,8,power_law_1.01,0.04656000137329101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,8,power_law_1.01,0.1387712001800537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,8,power_law_1.01,0.04147840142250061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,8,power_law_1.01,0.15902719497680665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,8,power_law_1.01,0.04602240025997162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,8,power_law_1.01,0.19227520227432252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,8,power_law_1.01,0.04947839975357056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,8,power_law_1.01,0.24041600227355958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,8,power_law_1.01,0.05315200090408325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,8,power_law_1.01,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,8,power_law_1.01,0.3179647922515869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,8,balanced,0.9663786888122559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,8,balanced,0.724618673324585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,8,power_law_1.01,0.05550079941749573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,8,power_law_1.01,0.41093759536743163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,8,power_law_1.01,0.05708799958229065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,8,power_law_1.01,0.5633791923522949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,8,power_law_1.01,0.05868160128593445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,8,power_law_1.01,0.7272575855255127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,8,power_law_1.01,0.06139519810676575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,8,power_law_1.01,0.06622080206871032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,8,power_law_1.01,0.8762368202209473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,8,power_law_1.01,0.06567680239677429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,8,power_law_1.01,1.4603263854980468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,8,power_law_1.01,0.0723904013633728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,8,power_law_1.01,0.07882879972457886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,8,power_law_1.01,0.08245760202407837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,8,power_law_1.01,2.9060287475585938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,8,power_law_1.01,0.08216959834098816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,8,power_law_1.01,0.0971455991268158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,8,power_law_1.01,0.11369600296020507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,8,power_law_1.01,0.1446079969406128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,8,power_law_1.01,0.17315200567245484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,8,balanced,0.8821973005930582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,8,power_law_1.01,0.2246783971786499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,8,power_law_1.01,0.2838912010192871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,8,power_law_1.01,0.38074240684509275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,8,power_law_1.01,0.4850304126739502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,8,power_law_1.01,0.6875135898590088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,8,power_law_1.2,0.050169599056243894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,8,power_law_1.01,0.949465560913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,8,power_law_1.2,0.04405120015144348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,8,power_law_1.01,1.1468992233276367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,8,power_law_1.2,0.043635201454162595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,8,power_law_1.01,1.8279104232788086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,8,power_law_1.2,0.05046399831771851
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,8,power_law_1.01,3.5510974884033204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,8,power_law_1.2,0.05314559936523437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,8,power_law_1.2,0.06157439947128296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,8,power_law_1.2,0.06227200031280518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,8,power_law_1.2,0.06646400094032287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,8,power_law_1.2,0.04462080001831055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,8,power_law_1.2,0.06528000235557556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,8,power_law_1.2,0.06613119840621948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,8,power_law_1.2,0.0664192020893097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,8,power_law_1.2,0.042656001448631284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,8,power_law_1.2,0.0705024003982544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,8,power_law_1.2,0.0721343994140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,8,power_law_1.2,0.04175359904766083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,8,power_law_1.2,0.07813119888305664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,8,balanced,1.3635360399882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,8,power_law_1.2,0.04561919867992401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,8,power_law_1.2,0.08966400027275086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,8,power_law_1.2,0.047968000173568726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,8,power_law_1.2,0.09132800102233887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,8,power_law_1.2,0.05440639853477478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,8,power_law_1.2,0.09519360065460206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,8,power_law_1.2,0.05603839755058289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,8,power_law_1.2,0.10238080024719239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,8,power_law_1.2,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,8,power_law_1.2,0.11726720333099365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,8,balanced,1.8699520428975422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,8,power_law_1.2,0.14216320514678954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,8,power_law_1.2,0.058297598361968996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,8,power_law_1.2,0.1785920023918152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,8,power_law_1.2,0.06064640283584595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,8,power_law_1.2,0.06355839967727661
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,8,power_law_1.2,0.21270399093627929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,8,power_law_1.2,0.06736639738082886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,8,power_law_1.2,0.2630784034729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,8,power_law_1.2,0.06896640062332153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,8,power_law_1.2,0.36735999584198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,8,power_law_1.2,0.07184000015258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,8,power_law_1.2,0.4576064109802246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,8,power_law_1.2,0.07939199805259704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,8,power_law_1.2,0.6958847999572754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,8,power_law_1.2,0.08785920143127442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,8,power_law_1.2,0.9645055770874024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,8,power_law_1.2,0.08686720132827759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,8,power_law_1.2,0.1040768027305603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,8,power_law_1.2,1.1421759605407715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,8,power_law_1.2,0.12223360538482667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,8,power_law_1.2,0.15073920488357545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,8,power_law_1.2,1.798111915588379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,8,power_law_1.2,0.19100799560546874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,8,power_law_1.2,3.5285312652587892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,8,power_law_1.2,0.23418879508972168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,8,power_law_1.2,0.30091519355773927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,8,power_law_1.2,0.4120639801025391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,8,power_law_1.2,0.5209663867950439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,8,power_law_1.2,0.7501567840576172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,8,power_law_1.2,0.9668095588684082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,8,power_law_1.2,1.449612808227539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,8,power_law_1.2,2.2339712142944337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,8,power_law_1.2,3.929587173461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,8,balanced,2.6701014836629233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,128,balanced,0.0642133355140686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,128,balanced,0.08178133269151051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,128,balanced,0.06347733239332835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,128,balanced,0.08124266564846039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,128,balanced,0.06376533210277557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,128,balanced,0.08247999846935272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,128,balanced,0.08326399823029836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,128,balanced,0.0651039977868398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,128,balanced,0.08669867118199666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,128,balanced,0.06643733382225037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,128,balanced,0.09684266646703084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,128,balanced,0.0795253316561381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,128,balanced,0.11157332857449849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,128,balanced,0.086517333984375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,128,balanced,0.11343999703725179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,128,balanced,0.08522666494051616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,128,balanced,0.12060266733169556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,128,balanced,0.08614400029182434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,128,balanced,0.11390399932861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,128,balanced,0.08875733613967896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,128,balanced,0.12009599804878235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,128,balanced,0.08875733613967896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,128,balanced,0.12306132912635803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,128,balanced,0.08661333719889323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,128,balanced,0.12170666456222534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,128,balanced,0.09264000256856282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,128,balanced,0.12866133451461792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,128,balanced,0.0906986693541209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,128,balanced,0.12229866782824199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,128,balanced,0.09929066896438599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,128,balanced,0.12338133653004964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,128,balanced,0.10198400417963664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,128,balanced,0.1328053375085195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,128,balanced,0.10947199662526448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,128,balanced,0.11524266997973125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,128,balanced,0.14498666922251383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,128,balanced,0.12418133020401001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,128,balanced,0.13757866621017456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,128,balanced,0.14828266700108847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,128,balanced,0.16200000047683716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,128,balanced,0.16714133818944296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,128,balanced,0.16288000345230103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,128,balanced,0.19556266069412231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,128,balanced,0.21257599194844565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,128,balanced,0.21909334262212118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,128,balanced,0.25388266642888385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,128,balanced,0.33685866991678876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,128,balanced,0.2736213405927022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,128,balanced,0.43327999114990234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,128,balanced,0.3592746655146281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,4,balanced,0.04402133325735728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,4,balanced,0.047226667404174805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,4,balanced,0.04970666766166687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,4,balanced,0.05381333331267039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,4,balanced,0.06224533418814341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,4,balanced,0.07819733520348866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,4,balanced,0.07863466441631317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,4,balanced,0.07928533355395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,4,balanced,0.07939733564853668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,4,balanced,0.08163733283678691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,128,balanced,0.4555893341700236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,128,balanced,0.5951253175735474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,4,balanced,0.08067733546098073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,4,balanced,0.0817493349313736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,4,balanced,0.08303466439247131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,4,balanced,0.08578667044639587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,4,balanced,0.09204799930254619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,4,balanced,0.09293333689371745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,4,balanced,0.09776533643404643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,4,balanced,0.10916800300280254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,4,balanced,0.11343466242154439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,4,balanced,0.15334933002789816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,128,balanced,0.5740586519241333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,128,balanced,0.7611093521118164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,4,balanced,0.16849066813786825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,4,balanced,0.24392000834147134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,4,balanced,0.26134934027989704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,4,balanced,0.3768320083618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,128,balanced,0.7249706586201986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,128,balanced,0.9469866752624512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,4,balanced,0.46329065163930255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,4,balanced,0.04366933306058248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,4,balanced,0.04561600089073181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,4,balanced,0.04757333298524221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,4,balanced,0.053871999184290566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,4,balanced,0.059978668888409935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,4,balanced,0.06676800052324931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,4,balanced,0.0688266654809316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,4,balanced,0.070592001080513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,4,balanced,0.06985066831111908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,4,balanced,0.0709440012772878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,4,balanced,0.07087466617425282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,4,balanced,0.072202667593956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,4,balanced,0.6352479855219523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,4,balanced,0.0753653347492218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,4,balanced,0.07644266883532207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,4,balanced,0.08238400022188823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,4,balanced,0.08449600140253703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,4,balanced,0.08867733677228291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,4,balanced,0.10662933190663655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,128,balanced,1.082085371017456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,4,balanced,0.11822400490442912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,128,balanced,1.460858662923177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,4,balanced,0.16241600116093954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,4,balanced,0.8024906317392985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,4,balanced,0.18905067443847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,4,balanced,0.25971200068791706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,4,balanced,0.3017653425534566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,4,balanced,0.43325332800547284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,4,balanced,0.987775961558024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,4,balanced,0.5502880016962687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,4,power_law_1.01,0.04891520142555237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,4,power_law_1.01,0.05055999755859375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,4,power_law_1.01,0.04744960069656372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,4,power_law_1.01,0.05367680191993714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,4,power_law_1.01,0.057811200618743896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,4,power_law_1.01,0.06685439944267273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,4,power_law_1.01,0.07031040191650391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,4,power_law_1.01,0.07419520020484924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,4,power_law_1.01,0.07795199751853943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,4,balanced,0.7672106424967448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,4,power_law_1.01,0.07622399926185608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,4,power_law_1.01,0.0800000011920929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,4,power_law_1.01,0.08168320059776306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,4,power_law_1.01,0.08497920036315917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,4,power_law_1.01,0.08444160223007202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,4,power_law_1.01,0.09221119880676269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,4,power_law_1.01,0.09648640155792236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,4,power_law_1.01,0.10366079807281495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,4,power_law_1.01,0.11526399850845337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,4,balanced,1.4881547292073567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,4,power_law_1.01,0.13282560110092162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,4,power_law_1.01,0.1673151969909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,4,power_law_1.01,0.20463359355926514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,4,power_law_1.01,0.2543999910354614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,4,balanced,1.0055413246154785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,4,power_law_1.01,0.3085119962692261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,4,power_law_1.01,0.456217622756958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,128,balanced,2.0666240056355796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,4,power_law_1.01,0.5419583797454834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,128,balanced,2.847087860107422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,4,power_law_1.01,0.71113600730896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,4,power_law_1.01,1.0277631759643555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,4,power_law_1.01,1.2228863716125489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,4,power_law_1.01,0.044659200310707095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,4,power_law_1.01,1.7377151489257812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,4,power_law_1.01,0.048518401384353635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,4,power_law_1.01,0.045126399397850035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,4,power_law_1.01,0.04961279928684235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,4,power_law_1.01,3.5172351837158202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,4,power_law_1.01,0.05237759947776795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,4,power_law_1.01,0.05979520082473755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,4,power_law_1.01,0.06354560256004334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,4,power_law_1.01,0.06564480066299438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,4,balanced,1.233130693435669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,4,power_law_1.01,0.06666880249977111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,4,power_law_1.01,0.06878079771995545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,4,power_law_1.2,0.04760960042476654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,4,power_law_1.01,0.07232000231742859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,4,power_law_1.01,0.07679359912872315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,4,power_law_1.01,0.08101119995117187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,4,power_law_1.2,0.04755200147628784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,4,power_law_1.01,0.07769600152969361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,4,power_law_1.2,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,4,power_law_1.2,0.05243520140647888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,4,power_law_1.01,0.08750079870223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,4,power_law_1.2,0.058323198556900026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,4,power_law_1.01,0.09248639941215515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,4,power_law_1.2,0.06749439835548401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,4,power_law_1.01,0.10312319993972778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,4,power_law_1.2,0.07008640170097351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,4,power_law_1.01,0.12213759422302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,4,power_law_1.2,0.07557759881019592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,4,power_law_1.01,0.13691519498825072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,4,power_law_1.2,0.07545599937438965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,4,power_law_1.01,0.1754240036010742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,4,power_law_1.2,0.07735040187835693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,4,power_law_1.01,0.215065598487854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,4,power_law_1.2,0.08065279722213745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,4,power_law_1.01,0.2784895896911621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,4,power_law_1.2,0.08480640053749085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,4,power_law_1.01,0.34979839324951173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,4,power_law_1.2,0.08432000279426574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,4,power_law_1.01,0.4869247913360596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,4,power_law_1.2,0.08714240193367004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,4,power_law_1.2,0.09663360118865967
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,4,power_law_1.01,0.624889612197876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,4,power_law_1.2,0.1008255958557129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,4,power_law_1.01,0.9414912223815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,4,power_law_1.2,0.10606080293655396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,4,power_law_1.01,1.1648320198059081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,4,power_law_1.2,0.12066559791564942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,4,balanced,1.9142667452494304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,4,balanced,2.8538134892781577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,4,power_law_1.2,0.13276159763336182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,4,power_law_1.01,1.607583999633789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,4,power_law_1.2,0.17249280214309692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,4,power_law_1.01,2.330112075805664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,4,power_law_1.2,0.2170880079269409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,4,power_law_1.2,0.3065023899078369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,4,power_law_1.01,4.594387054443359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,4,power_law_1.2,0.3377919912338257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,4,power_law_1.2,0.048870399594306946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,4,power_law_1.2,0.47716479301452636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,4,power_law_1.2,0.04689919948577881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,4,power_law_1.2,0.6254271984100341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,4,power_law_1.2,0.04869759976863861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,4,power_law_1.2,0.8817919731140137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,4,power_law_1.2,0.05187839865684509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,4,power_law_1.2,0.054688000679016115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,4,power_law_1.2,1.0595711708068847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,4,power_law_1.2,0.06295040249824524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,4,power_law_1.2,1.2803968429565429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,4,power_law_1.2,0.06403200030326843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,4,power_law_1.2,0.06699519753456115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,4,power_law_1.2,2.1668800354003905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,4,power_law_1.2,0.06967679858207702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,4,power_law_1.2,0.07002239823341369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,4,power_law_1.2,3.873273468017578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,4,power_law_1.2,0.07490559816360473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,4,power_law_1.2,0.07772799730300903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,4,power_law_1.2,0.0807039976119995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,4,power_law_1.2,0.08127999901771546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,4,power_law_1.2,0.09202560186386108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,4,power_law_1.2,0.09183359742164612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,4,power_law_1.2,0.10183039903640748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,4,power_law_1.2,0.12551039457321167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,4,power_law_1.2,0.1483199954032898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,4,power_law_1.2,0.189792001247406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,4,power_law_1.2,0.22391040325164796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,4,power_law_1.2,0.30088961124420166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,4,power_law_1.2,0.3705343961715698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,4,power_law_1.2,0.514079999923706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,4,power_law_1.2,0.6712063789367676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,4,power_law_1.2,0.9759936332702637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,4,power_law_1.2,1.259648036956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,4,power_law_1.2,1.6360319137573243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,4,power_law_1.2,2.977382469177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,4,balanced,3.7460800806681314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,4,power_law_1.2,5.14071044921875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,128,balanced,0.05804799993832906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,128,balanced,0.06010133524735769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,128,balanced,0.05827199916044871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,128,balanced,0.06035733222961426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,128,power_law_1.01,0.10864640474319458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,128,balanced,0.06442133088906606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,128,balanced,0.07298133273919423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,128,balanced,0.08469333251317342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,128,balanced,0.08256533245245616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,128,balanced,0.08682133754094441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,128,power_law_1.01,0.11103359460830689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,128,balanced,0.0869653324286143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,128,balanced,0.08865599830945332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,128,power_law_1.01,0.10636160373687745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,128,balanced,0.08835200468699138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,128,balanced,0.08691199620564778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,128,power_law_1.01,0.09649279713630676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,128,balanced,0.09274133046468098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,128,power_law_1.01,0.10353280305862426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,128,balanced,0.10309333602587382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,128,power_law_1.01,0.07671679854393006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,128,power_law_1.01,0.11101440191268921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,128,balanced,0.0969493289788564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,128,power_law_1.01,0.10928640365600586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,128,balanced,0.10105066498120625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,128,power_law_1.01,0.07471359968185425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,128,power_law_1.01,0.11146240234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,128,power_law_1.01,0.07343360185623168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,128,balanced,0.10678933064142863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,128,power_law_1.01,0.11159039735794067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,128,power_law_1.01,0.0713919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,128,balanced,0.1174720029036204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,128,power_law_1.01,0.11421439647674561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,128,power_law_1.01,0.07448959946632386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,128,power_law_1.01,0.11912959814071655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,128,power_law_1.01,0.08170239925384522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,128,balanced,0.13294933239618936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,128,power_law_1.01,0.11358079910278321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,128,power_law_1.01,0.0831488013267517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,128,power_law_1.01,0.11720960140228272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,128,power_law_1.01,0.08572800159454345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,128,balanced,0.15050666530927023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,128,power_law_1.01,0.11575679779052735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,128,power_law_1.01,0.08523520231246948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,128,power_law_1.01,0.12298239469528198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,128,power_law_1.01,0.08722559809684753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,128,power_law_1.01,0.12297600507736206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,128,power_law_1.01,0.08631680011749268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,128,balanced,0.1787733236948649
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,128,power_law_1.01,0.1254207968711853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,128,power_law_1.01,0.08694400191307068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,128,power_law_1.01,0.1260607957839966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,128,power_law_1.01,0.08737279772758484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,128,power_law_1.01,0.14563839435577391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,128,power_law_1.01,0.09066240191459655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,128,power_law_1.01,0.16814719438552855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,128,balanced,0.20950400829315186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,128,power_law_1.01,0.09946879744529724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,128,power_law_1.01,0.19630080461502075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,128,power_law_1.01,0.09786239862442017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,128,power_law_1.01,0.24078080654144288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,128,power_law_1.01,0.10423040390014648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,128,power_law_1.01,0.28318080902099607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,128,power_law_1.01,0.1146623969078064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,128,power_law_1.01,0.12838399410247803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,128,power_law_1.01,0.38588800430297854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,128,balanced,0.2696213324864705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,128,power_law_1.01,0.15916800498962402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,128,power_law_1.01,0.5027904033660888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,128,power_law_1.01,0.08481280207633972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,128,power_law_1.01,0.18456319570541382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,128,power_law_1.01,0.6734975814819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,128,power_law_1.01,0.2360447883605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,128,power_law_1.01,0.9661120414733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,128,power_law_1.01,0.08542720079421998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,128,power_law_1.01,0.2949311971664429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,128,power_law_1.01,0.08204159736633301
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,128,power_law_1.01,1.145241641998291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,128,power_law_1.01,0.39448959827423097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,128,power_law_1.01,0.07370880246162415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,128,power_law_1.01,0.5143680095672607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,128,power_law_1.01,1.7427328109741211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,128,balanced,0.3517386515935262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,128,power_law_1.01,0.0763264000415802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,128,power_law_1.01,0.7372223854064941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,128,power_law_1.01,0.07596799731254578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,128,power_law_1.01,3.571257781982422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,128,power_law_1.01,0.9449472427368164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,128,power_law_1.01,0.0832256019115448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,128,power_law_1.01,0.08545920252799988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,128,power_law_1.01,1.203878402709961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,128,power_law_1.01,0.08261119723320007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,128,power_law_1.01,1.8319168090820312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,128,power_law_1.01,0.08197759985923767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,128,power_law_1.01,3.747660827636719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,128,power_law_1.01,0.0836736023426056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,1,128,power_law_1.2,0.10785919427871704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,128,balanced,0.4644000132878621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,128,power_law_1.01,0.08092799782752991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,1,128,power_law_1.2,0.10940159559249878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,128,power_law_1.01,0.0859008014202118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,128,power_law_1.01,0.08759040236473084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,1,128,power_law_1.2,0.10481280088424683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,128,power_law_1.01,0.09232640266418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,1,128,power_law_1.2,0.0987775981426239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,128,power_law_1.01,0.09319679737091065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,1,128,power_law_1.2,0.10492160320281982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,128,power_law_1.01,0.09957119822502136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,1,128,power_law_1.2,0.10369280576705933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,1,128,power_law_1.2,0.11258879899978638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,128,power_law_1.01,0.10767359733581543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,1,128,power_law_1.2,0.12295039892196655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,128,power_law_1.01,0.12360960245132446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,1,128,power_law_1.2,0.0788927972316742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,1,128,power_law_1.2,0.11550079584121704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,128,power_law_1.01,0.14436479806900024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,128,balanced,0.6016106605529785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,1,128,power_law_1.2,0.10738559961318969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,128,power_law_1.01,0.16362240314483642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,1,128,power_law_1.2,0.07762560248374939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,1,128,power_law_1.2,0.10911359786987304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,128,power_law_1.01,0.2128063917160034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,1,128,power_law_1.2,0.07396479845046997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,1,128,power_law_1.2,0.1131775975227356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,1,128,power_law_1.2,0.07170559763908387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,128,power_law_1.01,0.2566015958786011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,1,128,power_law_1.2,0.07923840284347534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,1,128,power_law_1.2,0.11694079637527466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,128,power_law_1.01,0.3436863899230957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,1,128,power_law_1.2,0.11214720010757447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,1,128,power_law_1.2,0.07928320169448852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,128,power_law_1.01,0.42666239738464357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,1,128,power_law_1.2,0.11599359512329102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,1,128,power_law_1.2,0.08333439826965332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,128,power_law_1.01,0.597331190109253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,1,128,power_law_1.2,0.11882239580154419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,1,128,power_law_1.2,0.0877888023853302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,1,128,power_law_1.2,0.12285439968109131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,128,power_law_1.01,0.7831808090209961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,1,128,power_law_1.2,0.08684800267219543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,1,128,power_law_1.2,0.13614720106124878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,1,128,power_law_1.2,0.08692479729652405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,128,power_law_1.01,0.957759952545166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,1,128,power_law_1.2,0.14298880100250244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,1,128,power_law_1.2,0.08224639892578126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,1,128,power_law_1.2,0.1836351990699768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,128,power_law_1.01,1.5481216430664062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,1,128,power_law_1.2,0.08897280097007751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,128,balanced,0.7486080328623453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,1,128,power_law_1.2,0.2207871913909912
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,1,128,power_law_1.2,0.08950399756431579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,128,power_law_1.01,3.0959871292114256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,1,128,power_law_1.2,0.2572736024856567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,1,128,power_law_1.2,0.09068800210952759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,64,balanced,0.0666133314371109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,1,128,power_law_1.2,0.32067840099334716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,1,128,power_law_1.2,0.10034559965133667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,64,balanced,0.06625066697597504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,1,128,power_law_1.2,0.10315519571304321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,1,128,power_law_1.2,0.46353921890258787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,64,balanced,0.06620266536871593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,1,128,power_law_1.2,0.08240640163421631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,64,balanced,0.06865066786607106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,1,128,power_law_1.2,0.10557440519332886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,1,128,power_law_1.2,0.630131196975708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,64,balanced,0.07614933451016744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,64,balanced,0.09873066345850627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,1,128,power_law_1.2,0.0835968017578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,1,128,power_law_1.2,0.11957759857177734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,64,balanced,0.1079253355662028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,1,128,power_law_1.2,0.8173184394836426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,64,balanced,0.107424000898997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,1,128,power_law_1.2,0.13676799535751344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,1,128,power_law_1.2,0.07360640168190002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,64,balanced,0.11104533076286316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,1,128,power_law_1.2,1.2128447532653808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,64,balanced,0.1072106659412384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,64,balanced,0.11451199650764465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,1,128,power_law_1.2,0.1723647952079773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,1,128,power_law_1.2,0.06997759938240052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,1,128,power_law_1.2,1.3590592384338378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,64,balanced,0.11720533172289531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,1,128,power_law_1.2,0.07356799840927124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,64,balanced,0.1114453375339508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,1,128,power_law_1.2,0.1962623953819275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,64,balanced,0.11124799648920695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,1,128,power_law_1.2,2.1611072540283205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,1,128,power_law_1.2,0.07454079985618592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,1,128,power_law_1.2,0.2535936117172241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,64,balanced,0.11673600474993388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,64,balanced,0.1179093321164449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,1,128,power_law_1.2,0.07838720083236694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,1,128,power_law_1.2,5.158745574951172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,1,128,power_law_1.2,0.326963210105896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,64,balanced,0.12178132931391399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,1,128,power_law_1.2,0.0834176003932953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,128,balanced,1.1412586371103923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,64,balanced,0.13014400005340576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,1,128,power_law_1.2,0.4509632110595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,64,balanced,0.1337493360042572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,1,128,power_law_1.2,0.08511999845504761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,1,128,power_law_1.2,0.5949567794799805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,1,128,power_law_1.2,0.07923840284347534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,64,balanced,0.15414933363596597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,1,128,power_law_1.2,0.8307968139648437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,1,128,power_law_1.2,0.08280320167541504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,64,balanced,0.16196266810099283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,1,128,power_law_1.2,1.2352448463439942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,1,128,power_law_1.2,0.08050559759140015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,1,128,power_law_1.2,0.08886399865150452
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,64,balanced,0.1905226707458496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,1,128,power_law_1.2,1.376863956451416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,1,128,power_law_1.2,0.08703359961509705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,1,128,power_law_1.2,0.09258239865303039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,1,128,power_law_1.2,2.2566272735595705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,1,128,power_law_1.2,0.09566079974174499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,64,balanced,0.21914132436116537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,1,128,power_law_1.2,0.0976960003376007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,1,128,power_law_1.2,0.1124608039855957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,1,128,power_law_1.2,4.4171905517578125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,1,128,power_law_1.2,0.12083840370178223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,1,128,power_law_1.2,0.15205760002136232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,64,balanced,0.27376532554626465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,1,128,power_law_1.2,0.16764160394668579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,1,128,power_law_1.2,0.21418240070343017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,1,128,power_law_1.2,0.26975359916687014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,1,128,power_law_1.2,0.3877824068069458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,64,balanced,0.3594133456548055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,1,128,power_law_1.2,0.486246395111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,1,128,power_law_1.2,0.709164810180664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,1,128,power_law_1.2,0.9112832069396972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,1,128,power_law_1.2,1.0784064292907716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,1,128,power_law_1.2,1.756083106994629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,128,power_law_1.2,3.7068161010742187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,64,balanced,0.46133331457773846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,64,balanced,0.57477335135142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,1,128,balanced,2.211104075113932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,64,balanced,0.728928009668986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,64,balanced,0.05481066803137461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,64,balanced,0.056287998954455055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,64,balanced,0.055530667304992676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,64,balanced,0.05729066828886668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,64,balanced,0.0652213344971339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,64,balanced,0.07267199953397115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,64,balanced,0.0825386643409729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,64,balanced,0.08247466882069905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,64,balanced,0.08326399823029836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,64,balanced,0.08474133412043254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,64,balanced,0.0869706670443217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,64,balanced,0.08673066894213359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,64,balanced,0.09178133805592854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,64,balanced,0.05398933092753092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,64,balanced,0.0900266667207082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,64,balanced,0.05600533386071523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,64,balanced,0.09507200121879578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,64,balanced,0.05551466842492422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,64,balanced,0.05624533196290334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,64,balanced,0.09716799855232239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,64,balanced,0.06614399949709575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,64,balanced,0.10334400335947673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,64,balanced,0.07030933101971944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,64,balanced,0.08735466996828715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,64,balanced,0.11508267124493916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,64,balanced,0.08477866649627686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,64,balanced,0.08477333188056946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,64,balanced,0.12247999509175618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,64,balanced,0.08544533451398213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,64,balanced,0.08689066767692566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,64,balanced,0.14601600170135498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,64,balanced,0.08661333719889323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,64,balanced,1.0876853466033936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,64,balanced,0.08698667089144389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,64,balanced,0.08741333087285359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,64,balanced,0.1669386625289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,64,balanced,0.09134933352470398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,64,balanced,0.09304533402125041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,64,balanced,0.2116853396097819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,64,balanced,0.10169066985448201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,64,balanced,0.1086346705754598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,64,balanced,0.25467199087142944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,64,balanced,0.11727999647458394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,64,balanced,0.1353600025177002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,64,power_law_1.01,0.10501760244369507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,64,power_law_1.01,0.10763520002365112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,64,balanced,0.338373343149821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,64,balanced,0.1467359960079193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,64,power_law_1.01,0.1001855969429016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,64,power_law_1.01,0.08060799837112427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,64,balanced,0.18004266421000162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,64,power_law_1.01,0.08900480270385742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,64,power_law_1.01,0.09280639886856079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,64,power_law_1.01,0.10164480209350586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,64,balanced,0.2116053303082784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,64,power_law_1.01,0.1056831955909729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,64,balanced,0.43648000558217365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,64,power_law_1.01,0.10156799554824829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,64,power_law_1.01,0.10503040552139283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,64,power_law_1.01,0.10661760568618775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,64,balanced,0.27508266766866046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,64,power_law_1.01,0.10862720012664795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,64,power_law_1.01,0.10705920457839965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,64,power_law_1.01,0.10878080129623413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,64,power_law_1.01,0.11662720441818238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,64,balanced,0.6020106474558512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,64,power_law_1.01,0.11366399526596069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,64,balanced,0.3555999994277954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,64,power_law_1.01,0.12360320091247559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,64,power_law_1.01,0.1242751955986023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,64,power_law_1.01,0.13271039724349976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,64,power_law_1.01,0.15697280168533326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,64,power_law_1.01,0.17197439670562745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,64,power_law_1.01,0.23320960998535156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,64,power_law_1.01,0.07909119725227357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,64,power_law_1.01,0.2627455949783325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,64,power_law_1.01,0.36192638874053956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,64,power_law_1.01,0.0765824019908905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,64,balanced,0.4739786783854167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,64,power_law_1.01,0.46587519645690917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,64,power_law_1.01,0.07462400197982788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,64,power_law_1.01,0.6411263942718506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,64,balanced,0.7701546351114908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,64,power_law_1.01,0.06419839859008789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,64,power_law_1.01,0.07180160284042358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,64,balanced,2.0958666801452637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,64,power_law_1.01,0.8210047721862793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,64,power_law_1.01,0.07461760044097901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,64,power_law_1.01,0.9829504013061523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,64,power_law_1.01,0.0746944010257721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,64,power_law_1.01,0.0775551974773407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,64,power_law_1.01,1.524294376373291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,64,power_law_1.01,0.07953919768333435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,64,power_law_1.01,0.08319360017776489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,64,power_law_1.01,2.982348823547363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,64,power_law_1.01,0.08243200182914734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,64,balanced,0.6125333309173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,64,power_law_1.01,0.08366079926490784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,64,power_law_1.01,0.0866047978401184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,64,power_law_1.01,0.08739200234413147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,64,power_law_1.01,0.09289600253105164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,64,power_law_1.01,0.09455999732017517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,64,balanced,0.9578773180643717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,64,power_law_1.01,0.10107519626617431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,64,power_law_1.01,0.11331199407577515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,64,power_law_1.01,0.08688639998435974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,64,power_law_1.01,0.1250496029853821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,64,power_law_1.01,0.08271359801292419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,64,power_law_1.01,0.14982399940490723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,64,power_law_1.01,0.08095359802246094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,64,power_law_1.01,0.18030719757080077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,64,power_law_1.01,0.06731520295143127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,64,balanced,0.7649172941843668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,64,power_law_1.01,0.23410561084747314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,64,power_law_1.01,0.06663680076599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,64,power_law_1.01,0.27644801139831543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,64,power_law_1.01,0.07475839853286743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,64,power_law_1.01,0.38775680065155027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,64,power_law_1.01,0.07879679799079894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,64,power_law_1.01,0.4931839942932129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,64,power_law_1.01,0.07882879972457886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,64,power_law_1.2,0.10582400560379028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,64,power_law_1.01,0.7010240077972412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,64,power_law_1.01,0.082259202003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,64,power_law_1.01,0.8998656272888184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,64,power_law_1.2,0.10769280195236205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,64,power_law_1.01,0.08095999956130981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,64,power_law_1.2,0.09813119769096375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,64,power_law_1.01,1.1403840065002442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,64,power_law_1.01,0.08468480110168457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,64,power_law_1.2,0.0812608003616333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,64,power_law_1.01,0.07828480005264282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,64,power_law_1.01,1.7788288116455078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,64,power_law_1.2,0.08851839900016785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,64,power_law_1.01,0.08548480272293091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,64,power_law_1.2,0.0927616000175476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,64,power_law_1.01,0.0880128026008606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,64,power_law_1.2,0.09742720127105713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,64,power_law_1.01,3.6367679595947267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,64,power_law_1.01,0.09039360284805298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,64,power_law_1.2,0.10328960418701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,64,power_law_1.2,0.0791167974472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,64,power_law_1.2,0.10416640043258667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,64,power_law_1.01,0.09533439874649048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,64,power_law_1.2,0.10535039901733398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,64,balanced,1.4739680290222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,64,power_law_1.2,0.07637760043144226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,64,power_law_1.01,0.10088959932327271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,64,power_law_1.2,0.1067199945449829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,64,power_law_1.2,0.07521920204162598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,64,power_law_1.2,0.0645632028579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,64,power_law_1.2,0.10690560340881347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,64,power_law_1.01,0.11050239801406861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,64,power_law_1.2,0.10489599704742432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,64,power_law_1.01,0.12013440132141114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,64,power_law_1.2,0.07097600102424621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,64,balanced,1.1661439736684163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,64,power_law_1.2,0.10593279600143432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,64,power_law_1.01,0.13778560161590575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,64,power_law_1.2,0.07109119892120361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,64,power_law_1.2,0.11518720388412476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,64,power_law_1.01,0.16035840511322022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,64,power_law_1.2,0.07555199861526489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,64,power_law_1.2,0.11303679943084717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,64,power_law_1.01,0.20006399154663085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,64,power_law_1.2,0.07619199752807618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,64,power_law_1.2,0.12248320579528808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,64,power_law_1.2,0.13182079792022705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,64,power_law_1.2,0.0824512004852295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,64,power_law_1.01,0.24292480945587158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,64,power_law_1.2,0.1375167965888977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,64,power_law_1.2,0.07965440154075623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,64,power_law_1.01,0.33207681179046633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,64,power_law_1.2,0.17093119621276856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,64,power_law_1.2,0.08190720081329346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,64,power_law_1.01,0.4174464225769043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,64,power_law_1.2,0.1995200037956238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,64,power_law_1.2,0.08652799725532531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,64,power_law_1.2,0.26070399284362794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,64,power_law_1.01,0.584281587600708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,64,power_law_1.2,0.0824512004852295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,64,power_law_1.2,0.29818880558013916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,64,power_law_1.01,0.746073579788208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,64,power_law_1.2,0.08903679847717286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,64,power_law_1.2,0.3993727922439575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,64,power_law_1.2,0.5030079841613769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,64,power_law_1.2,0.09358720183372497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,64,power_law_1.01,0.927609634399414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,64,power_law_1.2,0.09644799828529357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,64,power_law_1.2,0.7949823856353759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,64,power_law_1.01,1.4692864418029785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,64,power_law_1.2,0.10254080295562744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,64,power_law_1.2,1.0201343536376952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,64,power_law_1.2,0.1171839952468872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,64,power_law_1.2,1.2808639526367187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,64,power_law_1.01,2.8898880004882814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,64,power_law_1.2,0.12641919851303102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,64,power_law_1.2,1.9651264190673827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,64,power_law_1.2,0.16311039924621581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,64,power_law_1.2,3.8948734283447264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,64,power_law_1.2,0.19026559591293335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,64,power_law_1.2,0.24179201126098632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,64,power_law_1.2,0.2993088006973267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,64,power_law_1.2,0.4346496105194092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,64,power_law_1.2,0.5502655982971192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,64,power_law_1.2,0.7833407878875732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,64,power_law_1.2,1.0862848281860351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,64,power_law_1.2,1.3041152000427245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,64,power_law_1.2,2.0765247344970703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,64,balanced,2.8985493977864585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,64,power_law_1.2,4.212486267089844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,64,balanced,2.2695840199788413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,64,power_law_1.2,0.08669440150260925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,64,power_law_1.2,0.08152959942817688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,64,power_law_1.2,0.07959039807319641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,64,power_law_1.2,0.06581119894981384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,64,power_law_1.2,0.06732800006866455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,64,power_law_1.2,0.07473919987678528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,64,power_law_1.2,0.0791167974472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,64,power_law_1.2,0.0816703975200653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,64,power_law_1.2,0.0832319974899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,32,balanced,0.06190933287143707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,64,power_law_1.2,0.082259202003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,64,power_law_1.2,0.08410879969596863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,32,balanced,0.062405332922935486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,32,balanced,0.0624533345301946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,64,power_law_1.2,0.08502399921417236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,32,balanced,0.06653866668542226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,64,power_law_1.2,0.08458880186080933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,32,balanced,0.08052266637484233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,64,power_law_1.2,0.08236160278320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,32,balanced,0.094842662413915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,64,power_law_1.2,0.0894976019859314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,32,balanced,0.10737066467603047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,32,balanced,0.10605333248774211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,64,power_law_1.2,0.09315840005874634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,32,balanced,0.10704533259073894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,64,power_law_1.2,0.09875839948654175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,32,balanced,0.10797333717346191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,64,power_law_1.2,0.11032320261001587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,32,balanced,0.11101866761843364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,64,power_law_1.2,0.12311680316925049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,32,balanced,0.11009599765141805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,32,balanced,0.10661333799362183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,64,power_law_1.2,0.1490048050880432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,32,balanced,0.11003200213114421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,64,power_law_1.2,0.1643839955329895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,32,balanced,0.11492799719174702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,64,power_law_1.2,0.2146496057510376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,32,balanced,0.11524800459543864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,64,power_law_1.2,0.2560895919799805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,32,balanced,0.12024533748626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,64,power_law_1.2,0.36396160125732424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,32,balanced,0.1272266705830892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,64,power_law_1.2,0.4655168056488037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,32,balanced,0.13713066776593527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,64,power_law_1.2,0.6468224048614502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,64,power_law_1.2,0.8491904258728027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,32,balanced,0.15106667081514993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,64,power_law_1.2,1.0179519653320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,32,balanced,0.1655786633491516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,64,power_law_1.2,1.6507392883300782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,64,power_law_1.2,3.385689544677734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,32,balanced,0.19709332784016928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,32,balanced,0.051402668158213295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,32,balanced,0.22549333175023398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,32,balanced,0.05218133330345154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,32,balanced,0.05207466582457224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,32,balanced,0.05659733215967814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,32,balanced,0.06247999767462412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,32,balanced,0.07064533233642578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,32,balanced,0.08388266960779826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,32,balanced,0.2834986646970113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,32,balanced,0.08275199929873149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,32,balanced,0.0839413305123647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,32,balanced,0.08397333820660909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,32,balanced,0.08593599994977315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,32,balanced,0.08681600292523702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,32,balanced,0.08745599786440532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,32,balanced,0.3739039897918701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,32,balanced,0.09100799759229024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,32,balanced,0.09807999928792317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,32,balanced,0.09921600421269734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,32,balanced,0.1090826690196991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,32,balanced,0.11758400003115337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,32,balanced,0.1269546647866567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,32,balanced,0.15036799510320029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,32,balanced,0.4837546745936076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,32,balanced,0.05402666827042898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,32,balanced,0.17216533422470093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,32,balanced,0.054325332244237266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,32,balanced,0.055573334296544395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,32,balanced,0.2158986727396647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,32,balanced,0.06011199951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,32,balanced,0.07055999835332234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,32,balanced,0.07714666426181793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,32,balanced,0.09131733576456706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,32,balanced,0.09143466750780742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,32,balanced,0.2605440020561218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,32,balanced,0.09474666913350423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,32,balanced,0.09301333626111348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,32,balanced,0.09313066800435384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,32,balanced,0.09240532914797465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,32,balanced,0.09013332923253377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,32,balanced,0.6056693394978842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,32,balanced,0.09238400061925252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,32,balanced,0.34724799791971844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,32,balanced,0.09582400321960449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,32,power_law_1.01,0.09451519846916198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,32,balanced,0.09679466485977173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,32,balanced,0.10541333754857381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,32,power_law_1.01,0.10508160591125489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,32,balanced,0.11430399616559346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,32,power_law_1.01,0.10059520006179809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,32,power_law_1.01,0.07594239711761475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,32,balanced,0.12478933731714885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,32,balanced,0.449290672938029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,32,power_law_1.01,0.08253440260887146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,32,power_law_1.01,0.09078400135040283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,32,balanced,0.13831999897956848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,32,power_law_1.01,0.07152640223503112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,32,power_law_1.01,0.0906495988368988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,32,power_law_1.01,0.09715840220451355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,32,balanced,0.15453333655993143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,32,power_law_1.01,0.07835519909858704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,32,power_law_1.01,0.08046079874038696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,32,power_law_1.01,0.09934080243110657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,32,power_law_1.01,0.07725440263748169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,32,power_law_1.01,0.10016640424728393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,32,power_law_1.01,0.06301440000534057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,32,power_law_1.01,0.09364479780197144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,32,power_law_1.01,0.10011520385742187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,32,balanced,0.7550453344980875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,32,balanced,0.1896053353945414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,32,power_law_1.01,0.06727039813995361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,32,power_law_1.01,0.10520960092544555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,32,power_law_1.01,0.08496639728546143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,32,power_law_1.01,0.07015680074691773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,32,balanced,0.6244746843973795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,32,power_law_1.01,0.10690560340881347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,32,power_law_1.01,0.06048640012741089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,32,power_law_1.01,0.07329279780387879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,32,power_law_1.01,0.10670720338821411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,32,balanced,0.22317334016164145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,32,power_law_1.01,0.06547200083732604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,32,power_law_1.01,0.0759552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,32,power_law_1.01,0.11002240180969239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,32,power_law_1.01,0.07656959891319275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,32,power_law_1.01,0.11022720336914063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,32,power_law_1.01,0.07445759773254394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,32,power_law_1.01,0.11641600131988525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,32,power_law_1.01,0.07779840230941773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,32,power_law_1.2,0.09472640156745911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,32,power_law_1.01,0.07567999958992004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,32,power_law_1.01,0.12378239631652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,32,power_law_1.01,0.07881600260734559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,32,power_law_1.01,0.08394880294799804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,32,power_law_1.01,0.13356800079345704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,32,balanced,0.28784533341725665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,32,power_law_1.2,0.1049280047416687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,32,power_law_1.01,0.08327680230140685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,32,power_law_1.01,0.08213760256767273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,32,power_law_1.01,0.15633920431137086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,32,power_law_1.2,0.10138880014419556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,32,power_law_1.01,0.085343998670578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,32,power_law_1.01,0.17227519750595094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,32,power_law_1.01,0.08240640163421631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,32,power_law_1.2,0.07773439884185791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,32,power_law_1.01,0.08781440258026123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,32,power_law_1.01,0.22254080772399903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,32,power_law_1.01,0.08407040238380432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,32,power_law_1.2,0.08206080198287964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,32,power_law_1.01,0.09388800263404846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,32,power_law_1.01,0.2565119981765747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,32,power_law_1.01,0.08663039803504943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,32,balanced,0.8010186354319254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,32,power_law_1.2,0.08942720293998718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,32,power_law_1.01,0.0949184000492096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,32,power_law_1.01,0.08907520174980163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,32,power_law_1.2,0.09063040018081665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,32,balanced,0.3725653489430745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,32,power_law_1.01,0.10088959932327271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,32,power_law_1.01,0.3517184019088745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,32,power_law_1.01,0.09255040287971497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,32,power_law_1.2,0.09280639886856079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,32,power_law_1.01,0.11720319986343383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,32,power_law_1.01,0.09884799718856811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,32,power_law_1.2,0.09768959879875183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,32,power_law_1.01,0.44544639587402346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,32,power_law_1.01,0.1277567982673645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,32,power_law_1.2,0.0986303985118866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,32,power_law_1.01,0.09608960151672363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,32,power_law_1.01,0.154367995262146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,32,power_law_1.01,0.5843071937561035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,32,power_law_1.2,0.10206719636917114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,32,power_law_1.01,0.10166399478912354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,32,power_law_1.01,0.17845760583877562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,32,power_law_1.2,0.10187519788742065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,32,power_law_1.01,0.7612800121307373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,32,power_law_1.01,0.10869120359420777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,32,power_law_1.2,0.09884160161018371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,32,power_law_1.01,0.24623360633850097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,32,balanced,1.135146697362264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,32,power_law_1.01,0.9386048316955566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,32,power_law_1.01,0.12291200160980224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,32,power_law_1.2,0.10494719743728638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,32,power_law_1.01,0.2872960090637207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,32,power_law_1.01,0.14242559671401978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,32,balanced,0.5002506573994955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,32,power_law_1.2,0.11279360055923462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,32,power_law_1.01,0.3844160079956055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,32,power_law_1.01,1.5183679580688476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,32,power_law_1.01,0.1635967969894409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,32,power_law_1.2,0.11156480312347412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,32,power_law_1.01,0.4928895950317383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,32,power_law_1.01,0.20514559745788574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,32,power_law_1.2,0.1149183988571167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,32,power_law_1.01,2.88143367767334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,32,power_law_1.01,0.7242688179016114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,32,power_law_1.01,0.249183988571167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,32,power_law_1.2,0.1255552053451538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,32,power_law_1.01,0.942188835144043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,32,power_law_1.01,0.3324991941452026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,32,balanced,0.9901440143585205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,32,power_law_1.2,0.13598719835281373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,32,power_law_1.01,0.4272768020629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,32,power_law_1.01,1.1522624015808105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,32,power_law_1.2,0.16440320014953613
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,32,power_law_1.2,0.19619840383529663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,32,power_law_1.01,0.6086592197418212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,32,power_law_1.01,1.7571136474609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,32,power_law_1.2,0.2399359941482544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,32,power_law_1.01,0.7754816055297852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,32,power_law_1.2,0.26934399604797366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,32,power_law_1.01,3.5315391540527346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,32,power_law_1.01,0.9546175956726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,32,power_law_1.2,0.3803584098815918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,32,balanced,0.6464853286743164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,32,power_law_1.01,1.4786304473876952
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,32,power_law_1.2,0.5165952205657959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,32,power_law_1.2,0.6760000228881836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,32,power_law_1.01,2.9694143295288087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,32,power_law_1.2,0.893331241607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,32,power_law_1.2,1.0673151969909669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,32,power_law_1.2,1.6132991790771485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,32,power_law_1.2,3.4318401336669924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,32,balanced,0.8012479941050211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,32,balanced,1.5340159734090169
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,32,balanced,2.2012319564819336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,32,power_law_1.2,0.07552639842033386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,32,power_law_1.2,0.08430079817771911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,32,balanced,1.2287253538767497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,32,power_law_1.2,0.07515519857406616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,32,power_law_1.2,0.06488320231437683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,32,power_law_1.2,0.06769279837608337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,32,power_law_1.2,0.07352319955825806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,32,power_law_1.2,0.07175679802894593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,32,power_law_1.2,0.08163840174674988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,32,power_law_1.2,0.08015999794006348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,32,power_law_1.2,0.08328319787979126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,32,power_law_1.2,0.08134400248527526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,32,power_law_1.2,0.08392959833145142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,32,power_law_1.2,0.08592000007629394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,32,power_law_1.2,0.08613759875297547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,32,power_law_1.2,0.09648640155792236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,32,power_law_1.2,0.09559680223464966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,32,power_law_1.2,0.10532480478286743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,32,power_law_1.2,0.11746560335159302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,32,power_law_1.2,0.13029119968414307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,32,power_law_1.2,0.16017919778823853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,32,power_law_1.2,0.18680319786071778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,32,balanced,3.011199951171875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,32,power_law_1.2,0.2471872091293335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,32,power_law_1.2,0.3081471920013428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,32,power_law_1.2,0.4321023941040039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,32,power_law_1.2,0.5587264060974121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,32,power_law_1.2,0.7473408222198487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,32,power_law_1.2,1.0161279678344726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,32,power_law_1.2,1.2441535949707032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,32,power_law_1.2,1.9546112060546874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,32,power_law_1.2,3.7870079040527345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,32,balanced,2.4051574071248374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,32,power_law_1.2,0.07853440046310425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,32,power_law_1.2,0.09356160163879394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,32,power_law_1.2,0.08476799726486206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,32,power_law_1.2,0.06373760104179382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,32,power_law_1.2,0.06524159908294677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,32,power_law_1.2,0.07146880030632019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,32,power_law_1.2,0.07623040080070495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,32,power_law_1.2,0.08410879969596863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,32,power_law_1.2,0.08053759932518005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,32,power_law_1.2,0.08076800107955932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,32,power_law_1.2,0.08760319948196411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,32,power_law_1.2,0.08874239921569824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,32,power_law_1.2,0.0871616005897522
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,32,power_law_1.2,0.08664960265159607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,32,power_law_1.2,0.0946943998336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,32,power_law_1.2,0.0943231999874115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,32,power_law_1.2,0.1013375997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,32,power_law_1.2,0.11397119760513305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,32,power_law_1.2,0.12258559465408325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,32,power_law_1.2,0.15070719718933107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,32,power_law_1.2,0.17279360294342042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,32,power_law_1.2,0.21824638843536376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,32,power_law_1.2,0.2626944065093994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,32,power_law_1.2,0.36380159854888916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,32,power_law_1.2,0.4620672225952148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,32,power_law_1.2,0.6497151851654053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,32,power_law_1.2,0.9207167625427246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,32,power_law_1.2,1.1206015586853026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,32,power_law_1.2,1.6098304748535157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,32,power_law_1.2,3.679520034790039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,16,balanced,0.05180266499519348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,16,balanced,0.052485331892967224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,16,balanced,0.05203199883302053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,16,balanced,0.0558240016301473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,16,balanced,0.05857066810131073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,16,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,16,balanced,0.07085866729418437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,16,balanced,0.07222400108973186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,16,balanced,0.07252799967924754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,16,balanced,0.0721973329782486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,16,balanced,0.07457600037256877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,16,balanced,0.07454399764537811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,16,balanced,0.07657066484292348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,16,balanced,0.07918933530648549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,16,balanced,0.08284799754619598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,16,balanced,0.08442667126655579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,16,balanced,0.09165333708127339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,16,balanced,0.10334400335947673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,16,balanced,0.11310399572054546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,16,balanced,0.1341973344484965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,16,balanced,0.15014400084813437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,16,balanced,0.18621333440144858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,16,power_law_1.01,0.06401280164718628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,16,balanced,0.226474662621816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,16,power_law_1.01,0.07523840069770812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,16,power_law_1.01,0.06524159908294677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,16,power_law_1.01,0.0546176016330719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,16,power_law_1.01,0.05708799958229065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,16,balanced,0.29175466299057007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,16,power_law_1.01,0.06238719820976257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,16,power_law_1.01,0.06538239717483521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,16,power_law_1.01,0.06848639845848084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,16,power_law_1.01,0.06766719818115234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,16,power_law_1.01,0.06887680292129517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,16,power_law_1.01,0.0703935980796814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,16,balanced,0.37880531946818036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,16,power_law_1.01,0.07324159741401673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,16,power_law_1.01,0.07431679964065552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,16,power_law_1.01,0.07650560140609741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,16,power_law_1.01,0.08287360072135926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,16,power_law_1.01,0.08565760254859925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,16,power_law_1.01,0.09385600090026855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,16,power_law_1.01,0.10862079858779908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,16,power_law_1.01,0.12048640251159667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,16,balanced,0.5110400120417277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,16,power_law_1.01,0.1470720052719116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,16,power_law_1.01,0.17716480493545533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,16,power_law_1.01,0.21994240283966066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,16,power_law_1.01,0.26128640174865725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,16,power_law_1.01,0.33664000034332275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,16,power_law_1.01,0.44209918975830076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,16,power_law_1.01,0.6238656044006348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,16,power_law_1.01,0.7930496215820313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,16,power_law_1.01,0.9929087638854981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,16,power_law_1.01,1.4969216346740724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,16,balanced,0.6618560155232748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,16,power_law_1.01,3.082131195068359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,16,balanced,0.8204586505889893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,16,balanced,1.2483999729156494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,16,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,16,balanced,0.04387199878692627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,16,balanced,0.04390400151411692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,16,balanced,0.047914668917655945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,16,balanced,0.055904000997543335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,16,balanced,0.07380266487598419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,16,balanced,0.09231999516487122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,16,balanced,0.09120532870292664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,16,balanced,0.09132267038027446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,16,balanced,0.09118933478991191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,16,balanced,0.0943946639696757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,16,balanced,0.09512000282605489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,16,balanced,0.09603200356165568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,16,balanced,0.09743466973304749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,16,balanced,0.10316266616185506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,16,balanced,0.10333866874376933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,16,balanced,0.11121599872907002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,16,balanced,0.11648533741633098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,16,balanced,0.04572266836961111
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,16,balanced,0.12548800309499106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,16,balanced,0.04603200157483419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,16,balanced,0.04614933331807455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,16,balanced,0.143477330605189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,16,balanced,0.048112000028292336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,16,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,16,balanced,0.05916266640027364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,16,balanced,0.1555359959602356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,16,balanced,0.07687999804814656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,16,balanced,0.07433600227038066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,16,balanced,0.07597866654396057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,16,balanced,0.0766293356815974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,16,balanced,0.18498132626215616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,16,balanced,0.07836266855398814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,16,balanced,0.08043733239173889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,16,balanced,0.08281599978605907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,16,balanced,2.4429759979248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,16,balanced,0.08337066570917766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,16,balanced,0.2281386653582255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,16,balanced,0.08890666564305623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,16,balanced,0.09212799866994222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,16,power_law_1.01,0.0686079978942871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,16,balanced,0.10121066371599834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,16,balanced,0.11249066392580669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,16,balanced,0.2849973241488139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,16,power_law_1.01,0.07717120051383972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,16,balanced,0.12218667070070903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,16,power_law_1.01,0.09008640050888062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,16,power_law_1.01,0.05411199927330017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,16,balanced,0.14877333243687949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,16,power_law_1.01,0.05921279788017273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,16,power_law_1.01,0.06811519861221313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,16,balanced,0.1651946703592936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,16,balanced,0.3493653138478597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,16,power_law_1.01,0.07034879922866821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,16,power_law_1.01,0.08161280155181885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,16,power_law_1.01,0.08242560029029847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,16,balanced,0.21010667085647583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,16,power_law_1.01,0.08223360180854797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,16,power_law_1.01,0.08532480001449586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,16,power_law_1.01,0.08791679739952088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,16,balanced,0.2678613265355428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,16,power_law_1.01,0.08932480216026306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,16,power_law_1.01,0.08946560025215149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,16,balanced,0.47154664993286133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,16,power_law_1.01,0.09600639939308167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,16,power_law_1.01,0.10087679624557495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,16,power_law_1.01,0.10863360166549682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,16,balanced,0.3524159987767537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,16,power_law_1.01,0.12007679939270019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,16,power_law_1.01,0.1296704053878784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,16,power_law_1.01,0.1545024037361145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,16,power_law_1.01,0.1760640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,16,power_law_1.01,0.21369600296020508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,16,power_law_1.01,0.2695552110671997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,16,balanced,0.4482346773147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,16,power_law_1.01,0.34218881130218504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,16,power_law_1.01,0.44205441474914553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,16,balanced,0.5973333517710367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,16,power_law_1.2,0.06405119895935059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,16,power_law_1.01,0.5793280124664306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,16,power_law_1.2,0.0754047989845276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,16,power_law_1.01,0.7670400142669678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,16,power_law_1.2,0.05727360248565674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,16,power_law_1.2,0.05488640069961548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,16,power_law_1.01,0.998630428314209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,16,power_law_1.2,0.05735039710998535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,16,balanced,0.6360426743825277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,16,power_law_1.2,0.059910398721694944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,16,power_law_1.01,1.3464256286621095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,16,power_law_1.2,0.06231679916381836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,16,power_law_1.01,0.050963199138641356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,16,power_law_1.01,2.6922880172729493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,16,power_law_1.2,0.06796799898147583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,16,power_law_1.2,0.06863359808921814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,16,power_law_1.01,0.0686016023159027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,16,power_law_1.01,0.07153919935226441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,16,power_law_1.2,0.06983680129051209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,16,power_law_1.01,0.05055360198020935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,16,power_law_1.2,0.07190399765968322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,16,power_law_1.01,0.0529151976108551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,16,balanced,0.7543040116628011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,16,power_law_1.2,0.07189760208129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,16,power_law_1.01,0.05688959956169128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,16,power_law_1.2,0.0745855987071991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,16,power_law_1.01,0.06037759780883789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,16,power_law_1.01,0.06325759887695312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,16,power_law_1.2,0.07862399816513062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,16,power_law_1.01,0.0645632028579712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,16,power_law_1.2,0.08484479784965515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,16,balanced,0.8107626438140869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,16,power_law_1.01,0.06629120111465454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,16,power_law_1.01,0.07194240093231201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,16,power_law_1.2,0.08783360123634339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,16,power_law_1.01,0.07342079877853394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,16,power_law_1.2,0.09574400186538697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,16,power_law_1.2,0.06881279945373535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,16,power_law_1.2,0.10899840593338013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,16,power_law_1.01,0.07723519802093506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,16,power_law_1.01,0.08208000063896179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,16,power_law_1.2,0.12506239414215087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,16,power_law_1.2,0.07676159739494323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,16,power_law_1.2,0.15492479801177977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,16,power_law_1.01,0.08857600092887878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,16,power_law_1.2,0.06561279892921448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,16,power_law_1.2,0.18455040454864502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,16,power_law_1.01,0.09266560077667237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,16,power_law_1.2,0.05336959958076477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,16,power_law_1.2,0.2275968074798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,16,power_law_1.01,0.10280319452285766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,16,power_law_1.2,0.06165760159492493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,16,power_law_1.2,0.27451519966125487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,16,power_law_1.01,0.11558400392532349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,16,power_law_1.2,0.06630399823188782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,16,power_law_1.2,0.3841599941253662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,16,power_law_1.01,0.12306560277938842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,16,power_law_1.2,0.06871680021286011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,16,power_law_1.2,0.48378877639770507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,16,power_law_1.01,0.1550271987915039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,16,power_law_1.2,0.07950080037117005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,16,power_law_1.2,0.7115071773529053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,16,power_law_1.01,0.18187520503997803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,16,power_law_1.2,0.08128640055656433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,16,power_law_1.01,0.2346112012863159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,16,balanced,1.011013348897298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,16,power_law_1.2,0.9050751686096191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,16,power_law_1.2,0.08339200019836426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,16,power_law_1.01,0.30122239589691163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,16,power_law_1.2,0.0839680016040802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,16,power_law_1.2,1.165651226043701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,16,power_law_1.01,0.40938878059387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,16,power_law_1.2,0.08685439825057983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,16,balanced,1.0981653531392415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,16,power_law_1.2,1.8427263259887696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,16,power_law_1.01,0.5219583988189698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,16,power_law_1.2,0.08910719752311706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,16,power_law_1.01,0.756883192062378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,16,power_law_1.2,0.09299200177192687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,16,power_law_1.2,3.4635391235351562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,16,power_law_1.2,0.09921280145645142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,16,power_law_1.01,0.9596799850463867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,16,power_law_1.2,0.09940479993820191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,16,power_law_1.01,1.1601471900939941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,16,power_law_1.2,0.10744960308074951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,16,power_law_1.01,1.752364730834961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,16,power_law_1.2,0.12305279970169067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,16,power_law_1.2,0.13612799644470214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,16,power_law_1.01,3.6889278411865236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,16,power_law_1.2,0.15486719608306884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,16,power_law_1.2,0.18495359420776367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,16,power_law_1.2,0.22919039726257323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,16,power_law_1.2,0.2777024030685425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,16,power_law_1.2,0.36445438861846924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,16,power_law_1.2,0.46433281898498535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,16,power_law_1.2,0.6626431941986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,16,balanced,1.5421226819356282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,16,power_law_1.2,0.8380543708801269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,16,power_law_1.2,1.0249919891357422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,16,power_law_1.2,1.649216079711914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,16,power_law_1.2,3.1247488021850587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,16,power_law_1.2,0.05299199819564819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,16,power_law_1.2,0.07030400037765502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,16,power_law_1.2,0.051846402883529666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,16,power_law_1.2,0.04929920136928558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,16,power_law_1.2,0.053529602289199826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,16,power_law_1.2,0.05734400153160095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,16,power_law_1.2,0.05953279733657837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,16,power_law_1.2,0.0639743983745575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,16,balanced,2.1353440284729004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,16,power_law_1.2,0.07058560252189636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,16,power_law_1.2,0.06631680130958557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,16,power_law_1.2,0.07052800059318542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,16,power_law_1.2,0.07724800109863281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,16,power_law_1.2,0.0769536018371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,16,power_law_1.2,0.08296959996223449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,16,power_law_1.2,0.09043840169906617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,16,power_law_1.2,0.0911679983139038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,16,power_law_1.2,0.10428160429000854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,16,power_law_1.2,0.12588800191879274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,16,power_law_1.2,0.1302016019821167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,16,power_law_1.2,0.16487679481506348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,16,power_law_1.2,0.1888383984565735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,16,power_law_1.2,0.24276480674743653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,16,power_law_1.2,0.3385983943939209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,16,balanced,3.031386693318685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,16,power_law_1.2,0.43507838249206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,16,power_law_1.2,0.5516608238220215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,16,power_law_1.2,0.8531455993652344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,16,power_law_1.2,1.007487964630127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,16,power_law_1.2,1.3043647766113282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,16,power_law_1.2,2.0676799774169923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,16,power_law_1.2,4.283142471313477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,8,balanced,0.05402133365472158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,8,balanced,0.056346664826075234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,8,balanced,0.057946667075157166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,8,balanced,0.05996266504128774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,8,balanced,0.06651733318964641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,8,balanced,0.0798826664686203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,8,balanced,0.09079466263453166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,8,balanced,0.09125333031018575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,8,balanced,0.09277333815892537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,8,balanced,0.09336533149083455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,8,balanced,0.0941973328590393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,8,balanced,0.09711466232935588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,8,balanced,0.0971999963124593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,8,balanced,0.09905067086219788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,8,balanced,0.10697600245475769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,8,balanced,0.10705066720644633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,8,balanced,0.1186186671257019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,8,balanced,0.13049599528312683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,8,balanced,0.1413226624329885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,8,balanced,0.1667893330256144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,8,balanced,0.18288000424702963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,8,balanced,0.22511466344197592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,8,balanced,0.26878400643666583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,8,balanced,0.3611413240432739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,8,balanced,0.48605867226918537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,8,power_law_1.01,0.05692160129547119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,8,balanced,0.6602986653645834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,8,power_law_1.01,0.06385279893875122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,8,power_law_1.01,0.07349119782447815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,8,power_law_1.01,0.05889279842376709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,8,power_law_1.01,0.06283519864082336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,8,power_law_1.01,0.0701312005519867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,8,power_law_1.01,0.07483519911766053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,8,power_law_1.01,0.08035200238227844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,8,power_law_1.01,0.08375040292739869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,8,power_law_1.01,0.08372480273246766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,8,power_law_1.01,0.08481280207633972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,8,power_law_1.01,0.09052159786224365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,8,balanced,0.8571146329243978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,8,power_law_1.01,0.08946560025215149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,8,power_law_1.01,0.09296640157699584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,8,power_law_1.01,0.10123519897460938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,8,power_law_1.01,0.1054144024848938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,8,power_law_1.01,0.11548800468444824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,8,power_law_1.01,0.13537280559539794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,8,power_law_1.01,0.1515328049659729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,8,power_law_1.01,0.18446719646453857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,8,power_law_1.01,0.1924288034439087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,8,power_law_1.01,0.2508352041244507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,8,power_law_1.01,0.3127039909362793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,8,power_law_1.01,0.40767998695373536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,8,power_law_1.01,0.5431424140930176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,8,balanced,1.0597493648529053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,8,power_law_1.01,0.8032320022583008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,8,power_law_1.01,1.0538816452026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,8,power_law_1.01,1.2461440086364746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,8,balanced,0.04603200157483419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,8,power_law_1.01,1.9678911209106444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,8,balanced,0.04433600107828776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,8,power_law_1.01,3.763673782348633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,8,balanced,0.0480373352766037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,8,balanced,0.05187733471393585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,8,balanced,0.05955733358860016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,8,balanced,0.08031466603279114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,8,balanced,0.10288533568382263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,8,balanced,0.10006399949391682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,8,balanced,0.1016319990158081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,8,balanced,0.10194666186968486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,8,balanced,0.10318932930628459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,8,balanced,0.1034879982471466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,8,balanced,0.10530666510264079
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,8,balanced,0.10714133580525716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,8,balanced,0.11338133613268535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,8,balanced,0.11385599772135417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,8,balanced,0.12371733784675598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,8,balanced,0.13412266969680786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,8,balanced,1.6407893498738606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,8,balanced,0.14135467012723288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,8,balanced,0.16195199886957803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,8,balanced,0.1777013341585795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,8,balanced,0.21460266908009848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,8,balanced,0.2692213257153829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,8,balanced,0.33907198905944824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,8,balanced,0.43777068456013996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,8,balanced,0.5967093308766683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,8,balanced,0.045663997530937195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,8,balanced,0.047557334105173744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,8,balanced,0.047930667797724404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,8,balanced,0.050016000866889954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,8,balanced,0.05570666491985321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,8,balanced,0.06397866706053416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,8,balanced,0.08569600184758504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,8,balanced,0.08617066343625386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,8,balanced,0.08549867073694865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,8,balanced,0.08853866656621297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,8,balanced,0.08859200278917949
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,8,balanced,0.09155199925104777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,8,balanced,0.7524212996164957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,8,balanced,0.09297066926956177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,8,balanced,3.2043625513712564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,8,balanced,0.09301867087682088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,8,balanced,0.10318932930628459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,8,balanced,0.10451733072598775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,8,balanced,0.11475732922554016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,8,balanced,0.12838400403658548
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,8,balanced,0.1413279970486959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,8,balanced,0.16636266311009726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,8,balanced,0.19097065925598145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,8,balanced,0.24016533295313516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,8,balanced,0.9560320377349854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,8,power_law_1.01,0.05246080160140991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,8,balanced,0.3117226759592692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,8,power_law_1.01,0.05249919891357422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,8,power_law_1.01,0.07529600262641907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,8,power_law_1.01,0.05232639908790589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,8,power_law_1.01,0.05736960172653198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,8,balanced,0.4113653500874837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,8,power_law_1.01,0.06737279891967773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,8,power_law_1.01,0.07363839745521546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,8,power_law_1.01,0.08281599879264831
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,8,power_law_1.01,0.08328319787979126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,8,power_law_1.01,0.08601599931716919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,8,power_law_1.01,0.09019520282745361
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,8,balanced,0.5369439919789633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,8,power_law_1.01,0.09060479998588562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,8,power_law_1.01,0.09475200176239014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,8,power_law_1.01,0.09820160269737244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,8,power_law_1.01,0.10734080076217652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,8,power_law_1.01,0.10976639986038209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,8,power_law_1.01,0.11544320583343506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,8,power_law_1.01,0.12895359992980956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,8,balanced,1.4203680356343586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,8,power_law_1.01,0.14212479591369628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,8,balanced,0.7675093015034994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,8,power_law_1.01,0.16471680402755737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,8,power_law_1.01,0.049318400025367734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,8,power_law_1.01,0.19057919979095458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,8,power_law_1.01,0.2515199899673462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,8,power_law_1.01,0.05487359762191772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,8,power_law_1.01,0.2942080020904541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,8,power_law_1.01,0.06161919832229614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,8,power_law_1.01,0.40393600463867185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,8,power_law_1.01,0.05095679759979248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,8,power_law_1.01,0.5046207904815674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,8,power_law_1.01,0.055801600217819214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,8,power_law_1.01,0.7342400074005127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,8,power_law_1.01,0.062489598989486694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,8,power_law_1.01,0.9216383934020996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,8,power_law_1.01,0.06522240042686463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,8,power_law_1.01,1.1419455528259277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,8,power_law_1.01,0.07289599776268005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,8,power_law_1.01,0.07412480115890503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,8,balanced,0.9702933629353842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,8,power_law_1.01,1.6812736511230468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,8,power_law_1.01,0.07765120267868042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,8,power_law_1.01,0.07799040079116822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,8,power_law_1.01,3.428998565673828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,8,power_law_1.01,0.08343039751052857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,8,power_law_1.01,0.08370559811592101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,8,power_law_1.01,0.09256319999694824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,8,power_law_1.2,0.04989440143108368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,8,power_law_1.01,0.10190720558166504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,8,power_law_1.2,0.05288959741592407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,8,power_law_1.01,0.10472320318222046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,8,power_law_1.2,0.0525439977645874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,8,power_law_1.01,0.11089919805526734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,8,power_law_1.2,0.05367680191993714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,8,power_law_1.01,0.1314687967300415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,8,power_law_1.2,0.057555198669433594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,8,power_law_1.01,0.14312319755554198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,8,power_law_1.2,0.05466880202293396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,8,power_law_1.2,0.06371840238571166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,8,power_law_1.01,0.1765504002571106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,8,power_law_1.2,0.06968320012092591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,8,power_law_1.2,0.06764159798622131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,8,power_law_1.01,0.21122560501098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,8,balanced,1.2118293444315593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,8,power_law_1.2,0.060089600086212155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,8,power_law_1.2,0.07964159846305847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,8,power_law_1.01,0.26511359214782715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,8,power_law_1.2,0.05882880091667175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,8,power_law_1.2,0.08254719972610473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,8,power_law_1.2,0.06207360029220581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,8,power_law_1.2,0.08360959887504578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,8,power_law_1.01,0.3472640037536621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,8,power_law_1.2,0.07004799842834472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,8,power_law_1.2,0.08771839737892151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,8,power_law_1.01,0.4915264129638672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,8,power_law_1.2,0.07143679857254029
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,8,power_law_1.2,0.09065600037574768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,8,power_law_1.01,0.5997888088226319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,8,power_law_1.2,0.08122239708900451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,8,power_law_1.2,0.09459840059280396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,8,power_law_1.2,0.09739519953727722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,8,power_law_1.2,0.08260480165481568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,8,power_law_1.01,0.8923199653625489
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,8,power_law_1.2,0.10823040008544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,8,power_law_1.2,0.08135679960250855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,8,power_law_1.01,1.1053695678710938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,8,power_law_1.2,0.11190400123596192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,8,power_law_1.2,0.08458240032196045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,8,power_law_1.2,0.11887359619140625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,8,power_law_1.01,1.4074367523193358
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,8,power_law_1.2,0.08581119775772095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,8,power_law_1.2,0.1331455945968628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,8,power_law_1.2,0.08970879912376403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,8,power_law_1.01,2.0849983215332033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,8,power_law_1.2,0.14960000514984131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,8,power_law_1.2,0.09399679899215699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,8,power_law_1.2,0.17816319465637206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,8,power_law_1.2,0.10317440032958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,8,power_law_1.2,0.2118016004562378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,8,power_law_1.01,4.162188720703125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,8,power_law_1.2,0.10667519569396973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,8,power_law_1.2,0.2667455911636353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,8,balanced,2.7515999476114907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,8,power_law_1.2,0.3470911979675293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,8,power_law_1.2,0.11914240121841431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,8,power_law_1.2,0.14280320405960084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,8,power_law_1.2,0.42844161987304685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,8,power_law_1.2,0.1543552041053772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,8,power_law_1.2,0.5502848148345947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,8,balanced,1.8671573003133137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,8,power_law_1.2,0.1964416027069092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,8,power_law_1.2,0.8171008110046387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,8,power_law_1.2,0.20015358924865723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,8,power_law_1.2,1.0060480117797852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,8,power_law_1.2,0.2587584018707275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,8,power_law_1.2,0.31478400230407716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,8,power_law_1.2,1.1722304344177246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,8,power_law_1.2,0.04948480129241943
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,8,power_law_1.2,0.4204991817474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,8,power_law_1.2,1.9527168273925781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,8,power_law_1.2,0.05371519923210144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,8,power_law_1.2,0.5977407932281494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,8,power_law_1.2,0.04801279902458191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,8,power_law_1.2,0.9215167999267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,8,power_law_1.2,0.049420800805091855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,8,power_law_1.2,4.3884223937988285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,8,power_law_1.2,1.1546048164367675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,8,power_law_1.2,0.05249919891357422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,8,power_law_1.2,1.3234047889709473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,8,power_law_1.2,0.06104320287704468
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,8,power_law_1.2,2.2918912887573244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,8,power_law_1.2,0.06368640065193176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,8,power_law_1.2,0.07196800112724304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,8,power_law_1.2,4.29791374206543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,8,power_law_1.2,0.07345280051231384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,8,power_law_1.2,0.07416960000991821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,8,power_law_1.2,0.07712640166282654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,8,power_law_1.2,0.08039680123329163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,8,power_law_1.2,0.08386560082435608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,8,power_law_1.2,0.088128000497818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,8,power_law_1.2,0.10216959714889526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,8,power_law_1.2,0.10737279653549195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,8,power_law_1.2,0.11588480472564697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,8,power_law_1.2,0.13129600286483764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,8,power_law_1.2,0.146943998336792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,8,power_law_1.2,0.1856384038925171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,8,power_law_1.2,0.21299200057983397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,8,power_law_1.2,0.2908672094345093
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,8,power_law_1.2,0.3718400001525879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,8,power_law_1.2,0.5224319934844971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,8,power_law_1.2,0.647488021850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,8,power_law_1.2,0.9499711990356445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,8,balanced,3.647738774617513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,8,power_law_1.2,1.2108799934387207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,8,power_law_1.2,1.4454784393310547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,8,power_law_1.2,2.240787124633789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,8,power_law_1.2,4.529958343505859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,128,balanced,0.04055999964475632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,128,balanced,0.04159466673930486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,128,balanced,0.040565334260463715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,128,balanced,0.04301333427429199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,128,balanced,0.04316799839337667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,128,balanced,0.043653334180514015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,128,balanced,0.04381866753101349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,128,balanced,0.043791999419530235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,128,balanced,0.0435146689414978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,128,balanced,0.04383466641108195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,128,balanced,0.04548799991607666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,128,balanced,0.04525866607824961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,128,balanced,0.04650133351484934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,128,balanced,0.045935998360315956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,128,balanced,0.04765866696834564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,128,balanced,0.050000001986821495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,128,balanced,0.048250665267308555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,128,balanced,0.03558400024970373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,128,balanced,0.04997866849104563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,128,balanced,0.04997866849104563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,128,balanced,0.03750933210055033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,128,balanced,0.03547733277082443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,128,balanced,0.03953066716591517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,128,balanced,0.054042667150497437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,128,balanced,0.03756800045569738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,128,balanced,0.039887999494870506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,128,balanced,0.05624533196290334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,128,balanced,0.04055466751257578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,128,balanced,0.03957866628964742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,128,balanced,0.039893334110577904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,128,balanced,0.041690667470296226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,128,balanced,0.0598826656738917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,128,balanced,0.04185600082079569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,128,balanced,0.04358933369318644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,128,balanced,0.042768001556396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,128,balanced,0.04373333354791006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,128,balanced,0.06644799808661143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,128,balanced,0.04593066871166229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,128,balanced,0.046021332343419395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,128,balanced,0.04743466774622599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,128,balanced,0.050160000721613564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,128,balanced,0.08169066905975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,128,balanced,0.05380799869696299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,128,balanced,0.05632533133029938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,128,balanced,0.062309334675470986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,128,balanced,0.09442133704821269
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,128,balanced,0.07450133562088013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,128,balanced,0.09673066933949788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,128,balanced,0.11961066722869873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,128,balanced,0.1333013375600179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,128,balanced,0.16129600008328757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,128,balanced,0.14127467075983682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,128,balanced,0.2162826657295227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,128,balanced,0.16452266772588095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,128,balanced,0.27163734038670856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,128,balanced,0.330186665058136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,128,balanced,0.24194665749867758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,128,balanced,0.49639999866485596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,4,balanced,0.04849599798520406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,4,balanced,0.05008533100287119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,4,balanced,0.0537066658337911
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,4,balanced,0.055871998270352684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,4,balanced,0.06834133466084798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,4,balanced,0.09613333145777385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,4,balanced,0.11979200442632039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,4,balanced,0.12212799986203511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,4,balanced,0.123690664768219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,4,balanced,0.1227946678797404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,4,balanced,0.12230400244394939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,4,balanced,0.12613866726557413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,4,balanced,0.1260533332824707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,4,balanced,0.12402666608492534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,4,balanced,0.13127467036247253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,4,balanced,0.13195199767748514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,4,balanced,0.1409280002117157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,4,balanced,0.15388266245524088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,4,balanced,0.16710933049519858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,4,balanced,0.1922880013783773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,4,balanced,0.2163040041923523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,128,balanced,0.4267679850260417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,4,balanced,0.27267199754714966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,4,balanced,0.3452479839324951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,4,balanced,0.45165332158406574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,4,balanced,0.5876426696777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,4,balanced,0.04598399996757507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,4,balanced,0.04794133206208547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,128,balanced,0.93941330909729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,4,balanced,0.05162666738033295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,4,balanced,0.05377600093682607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,4,balanced,0.0621919979651769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,4,balanced,0.07842133442560832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,4,balanced,0.10520533720652263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,4,balanced,0.10113599896430969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,4,balanced,0.8532959620157877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,4,balanced,0.10130666693051656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,4,balanced,0.101200004418691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,4,balanced,0.10348266363143921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,4,balanced,0.10763200124104817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,4,balanced,0.10730133454004924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,4,balanced,0.11276800433794658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,4,balanced,0.11922132968902588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,4,balanced,0.12260799606641133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,4,balanced,0.13360533118247986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,4,balanced,0.15101866920789084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,4,balanced,1.0896586577097576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,4,balanced,0.1696000099182129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,4,balanced,0.20151466131210327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,4,balanced,0.2333013415336609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,4,balanced,0.29790933926900226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,4,power_law_1.01,0.047993600368499756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,4,balanced,0.38155198097229004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,4,power_law_1.01,0.050886398553848265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,4,power_law_1.01,0.06225280165672302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,4,power_law_1.2,0.04808320105075836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,4,balanced,1.361087958017985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,4,power_law_1.01,0.05491840243339539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,4,power_law_1.01,0.0643775999546051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,4,power_law_1.2,0.05061119794845581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,4,power_law_1.01,0.07103999853134155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,4,balanced,0.518885334332784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,4,power_law_1.2,0.05431039929389954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,4,power_law_1.01,0.08409600257873535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,4,power_law_1.2,0.055264002084732054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,4,power_law_1.01,0.09537919759750366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,4,power_law_1.2,0.06293759942054748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,4,power_law_1.01,0.09845119714736938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,4,power_law_1.2,0.07267839908599853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,4,power_law_1.01,0.09938560128211975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,4,power_law_1.2,0.07756800055503846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,4,power_law_1.01,0.1054144024848938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,4,balanced,0.7094986438751221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,4,power_law_1.01,0.04808320105075836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,4,power_law_1.01,0.11103999614715576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,4,power_law_1.2,0.09200639724731445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,4,power_law_1.01,0.11159679889678956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,4,power_law_1.2,0.0954688012599945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,4,power_law_1.01,0.053011202812194826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,4,power_law_1.01,0.11189759969711303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,4,power_law_1.2,0.09758080244064331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,4,power_law_1.01,0.05939199924468994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,4,power_law_1.01,0.1219648003578186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,4,power_law_1.2,0.10394879579544067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,4,power_law_1.01,0.052070397138595584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,4,power_law_1.01,0.12208000421524048
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,4,power_law_1.2,0.10930559635162354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,4,power_law_1.01,0.0604095995426178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,4,power_law_1.01,0.13224960565567018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,4,power_law_1.2,0.1093951940536499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,4,power_law_1.01,0.0700543999671936
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,4,power_law_1.01,0.15101439952850343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,4,power_law_1.2,0.11461759805679321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,4,power_law_1.01,0.07463039755821228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,4,power_law_1.01,0.16595840454101562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,4,power_law_1.01,0.08572160005569458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,4,power_law_1.2,0.11991039514541627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,4,balanced,1.009973367055257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,4,power_law_1.01,0.21404800415039063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,4,power_law_1.01,0.08744959831237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,4,power_law_1.2,0.1253376007080078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,4,power_law_1.01,0.23564159870147705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,4,power_law_1.01,0.09097599983215332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,4,power_law_1.2,0.13553279638290405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,4,power_law_1.01,0.31678080558776855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,4,power_law_1.01,0.09459840059280396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,4,power_law_1.2,0.14988160133361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,4,balanced,2.0282773971557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,4,power_law_1.01,0.3862783908843994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,4,power_law_1.01,0.10085760354995728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,4,power_law_1.2,0.1754431962966919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,4,power_law_1.01,0.5143167972564697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,4,power_law_1.01,0.10317440032958984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,4,power_law_1.2,0.04749439954757691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,4,power_law_1.2,0.2135103940963745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,4,power_law_1.01,0.11038719415664673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,4,power_law_1.01,0.6586048126220703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,4,power_law_1.2,0.25895678997039795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,4,power_law_1.2,0.05258240103721619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,4,power_law_1.01,0.12125439643859863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,4,power_law_1.01,0.9167296409606933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,4,power_law_1.2,0.31210238933563234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,4,power_law_1.2,0.050220799446105954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,4,power_law_1.01,0.12779519557952881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,4,power_law_1.2,0.39777920246124265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,4,power_law_1.01,1.2180480003356933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,4,power_law_1.2,0.05246080160140991
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,4,power_law_1.01,0.14058879613876343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,4,power_law_1.2,0.5384640216827392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,4,power_law_1.2,0.06074240207672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,4,power_law_1.01,1.567852783203125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,4,power_law_1.01,0.16692479848861694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,4,power_law_1.2,0.6731008052825928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,4,balanced,1.2985226313273113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,4,power_law_1.2,0.0682752013206482
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,4,power_law_1.01,2.381337547302246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,4,power_law_1.01,0.17307519912719727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,4,power_law_1.2,1.0025983810424806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,4,power_law_1.2,0.07260800004005433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,4,power_law_1.01,0.21807360649108887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,4,power_law_1.2,0.08266879916191101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,4,power_law_1.2,1.4057408332824708
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,4,power_law_1.01,4.680217742919922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,4,power_law_1.01,0.2587199926376343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,4,power_law_1.2,0.08598399758338929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,4,power_law_1.2,1.6965887069702148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,4,power_law_1.01,0.34810240268707277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,4,power_law_1.2,0.08703359961509705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,4,power_law_1.2,2.496051216125488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,4,power_law_1.01,0.43649921417236326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,4,power_law_1.2,0.09267200231552124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,4,power_law_1.01,0.5715263843536377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,4,power_law_1.2,0.09943040013313294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,4,power_law_1.2,5.012089538574219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,4,power_law_1.2,0.101254403591156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,4,power_law_1.2,0.10990079641342163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,4,power_law_1.01,0.7877503871917725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,4,power_law_1.2,0.12100479602813721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,4,power_law_1.01,1.1189375877380372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,4,power_law_1.2,0.12957439422607422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,4,power_law_1.2,0.14168959856033325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,4,power_law_1.01,1.4271615982055663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,4,power_law_1.2,0.16720000505447388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,4,balanced,1.621338685353597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,4,power_law_1.01,1.818822479248047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,4,power_law_1.2,0.1748095989227295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,4,power_law_1.2,0.21614720821380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,4,power_law_1.01,2.643187141418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,4,power_law_1.2,0.2638335943222046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,4,power_law_1.2,0.35006721019744874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,4,power_law_1.2,0.4411647796630859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,4,power_law_1.01,5.341696166992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,4,power_law_1.2,0.6272575855255127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,4,power_law_1.2,0.8339584350585938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,4,power_law_1.2,1.171776008605957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,4,power_law_1.2,1.5060864448547364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,4,power_law_1.2,1.9066560745239258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,4,power_law_1.2,2.8318143844604493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,4,power_law_1.2,5.73895034790039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,128,balanced,0.03811733424663544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,4,balanced,4.0173600514729815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,4,balanced,2.4775999387105307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,128,balanced,0.03788266579310099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,128,balanced,0.03758399933576584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,128,balanced,0.03997333347797394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,128,balanced,0.0383093332250913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,128,balanced,0.04155199974775314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,128,balanced,0.041936000188191734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,128,balanced,0.04010133445262909
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,128,balanced,0.04194133480389913
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,128,balanced,0.041637333730856575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,128,balanced,0.041637333730856575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,128,balanced,0.043807998299598694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,128,balanced,0.04204800228277842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,128,balanced,0.043706665436426796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,128,balanced,0.047877331574757896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,128,balanced,0.04584533472855886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,128,balanced,0.04701333244641622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,128,balanced,0.049973333875338234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,128,balanced,0.05223466455936432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,128,balanced,0.019194666296243668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,128,balanced,0.05433600147565206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,128,balanced,0.019845332950353622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,128,balanced,0.0205226664741834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,128,balanced,0.060229331254959106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,128,balanced,0.023455999791622162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,128,balanced,0.0379573330283165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,128,balanced,0.03133333226044973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,128,balanced,0.031717332700888314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,128,balanced,0.0680213322242101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,128,balanced,0.0317546675602595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,128,balanced,0.032469332218170166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,128,balanced,0.032357332607110344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,128,balanced,0.032613334556420646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,128,balanced,0.07868800063927968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,128,balanced,0.033733333150545754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,128,balanced,0.03356266766786575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,128,balanced,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,128,balanced,0.035162667433420815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,128,balanced,0.10127466917037964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,128,balanced,0.0364533339937528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,128,balanced,0.03576533248027166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,128,balanced,0.038906666139761605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,128,balanced,0.04164800047874451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,4,balanced,4.856890678405762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,128,balanced,0.04823466638724009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,128,balanced,0.11766933401425679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,128,balanced,0.05204799771308899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,128,balanced,0.06419733166694641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,128,balanced,0.15281599760055542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,128,balanced,0.08184533317883809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,128,balanced,0.10524266958236694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,128,balanced,0.19033066431681314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,128,balanced,0.1320480008920034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,128,balanced,0.18261865774790445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,128,balanced,0.22868265708287558
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,128,power_law_1.01,0.0383679986000061
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,128,power_law_1.01,0.03878400027751923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,128,power_law_1.01,0.03768320083618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,128,power_law_1.01,0.0406464010477066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,128,power_law_1.01,0.04017919898033142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,128,power_law_1.01,0.041382399201393125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,128,power_law_1.01,0.04152320027351379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,128,power_law_1.01,0.041503998637199405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,128,balanced,0.23281067609786987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,128,power_law_1.01,0.041459199786186215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,128,power_law_1.01,0.04136959910392761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,128,power_law_1.01,0.04227840006351471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,128,power_law_1.01,0.04385280013084412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,128,power_law_1.01,0.04305280148983002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,128,power_law_1.01,0.04359680116176605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,128,power_law_1.01,0.0450111985206604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,128,power_law_1.01,0.0459199994802475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,128,power_law_1.01,0.04730879962444305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,128,power_law_1.01,0.04935680031776428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,128,power_law_1.01,0.05278080105781555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,128,power_law_1.01,0.05947520136833191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,128,power_law_1.01,0.03441280126571655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,128,power_law_1.01,0.06636800169944763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,128,power_law_1.01,0.08124799728393554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,128,power_law_1.01,0.0341376006603241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,128,power_law_1.01,0.10072319507598877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,128,power_law_1.01,0.03505919873714447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,128,balanced,0.333077331384023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,128,power_law_1.01,0.1300927996635437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,128,power_law_1.01,0.03621760010719299
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,128,power_law_1.01,0.1668544054031372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,128,power_law_1.01,0.03731839954853058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,128,balanced,0.2847306728363037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,128,power_law_1.01,0.03826560080051422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,128,power_law_1.01,0.2413952112197876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,128,power_law_1.01,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,128,power_law_1.01,0.31157119274139405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,128,power_law_1.01,0.035622400045394895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,128,power_law_1.01,0.03809280097484589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,128,power_law_1.01,0.03861120045185089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,128,power_law_1.01,0.37004799842834474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,128,power_law_1.01,0.03860479891300202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,128,power_law_1.01,0.0353983998298645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,128,power_law_1.01,0.5803711891174317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,128,power_law_1.01,0.03979519903659821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,128,power_law_1.01,0.03611519932746887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,128,power_law_1.01,0.04014720022678375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,128,power_law_1.01,0.03712640106678009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,128,power_law_1.01,1.0803647994995118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,128,power_law_1.01,0.040889599919319154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,128,power_law_1.01,0.03847039937973022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,128,power_law_1.01,0.041465601325035094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,128,power_law_1.01,0.03875199854373932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,128,power_law_1.01,0.04447360038757324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,128,power_law_1.01,0.03914879858493805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,128,power_law_1.01,0.04474239945411682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,128,power_law_1.01,0.03857919871807099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,128,power_law_1.01,0.04648320078849792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,128,power_law_1.01,0.03903999924659729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,128,power_law_1.01,0.05015040040016174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,128,power_law_1.01,0.03930880129337311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,128,power_law_1.01,0.055315202474594115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,128,power_law_1.01,0.03932799994945526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,128,power_law_1.01,0.062412798404693604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,128,power_law_1.01,0.04028159976005554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,128,power_law_1.01,0.07255039811134338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,128,power_law_1.01,0.040678399801254275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,128,power_law_1.01,0.10010240077972413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,128,power_law_1.01,0.040992000699043275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,2,128,power_law_1.2,0.03939839899539947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,128,power_law_1.01,0.017612800002098083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,128,power_law_1.01,0.13308160305023192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,128,power_law_1.01,0.04366720020771027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,128,power_law_1.01,0.017529599368572235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,128,power_law_1.01,0.17045120000839234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,2,128,power_law_1.2,0.03899520039558411
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,128,power_law_1.01,0.044614401459693906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,128,power_law_1.01,0.018406400084495546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,2,128,power_law_1.2,0.0394239991903305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,128,power_law_1.01,0.21358079910278321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,128,power_law_1.01,0.04624640047550201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,128,power_law_1.01,0.02025599926710129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,2,128,power_law_1.2,0.04094719886779785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,128,power_law_1.01,0.05063679814338684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,128,power_law_1.01,0.3150592088699341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,128,power_law_1.01,0.03800320029258728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,2,128,power_law_1.2,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,128,balanced,0.4368586540222168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,128,power_law_1.01,0.029100799560546876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.01,0.05426560044288635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,128,power_law_1.01,0.40250239372253416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,128,power_law_1.01,0.029894399642944335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.01,0.06074240207672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,2,128,power_law_1.2,0.042156800627708435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,128,power_law_1.01,0.49080958366394045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.01,0.06913279891014099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,128,power_law_1.01,0.029286399483680725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,2,128,power_law_1.2,0.04262399971485138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.01,0.08691200017929077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,128,power_law_1.01,0.029721599817276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,128,power_law_1.01,0.7971327781677247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,2,128,power_law_1.2,0.04209280014038086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.01,0.1033087968826294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,128,power_law_1.01,0.02961280047893524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,128,power_law_1.01,0.02980479896068573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.01,0.1324288010597229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,2,128,power_law_1.2,0.04292480051517487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,128,power_law_1.01,0.030803200602531434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,2,128,power_law_1.2,0.042438399791717527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,128,power_law_1.01,1.7085760116577149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.01,0.1694208025932312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,128,power_law_1.01,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,2,128,power_law_1.2,0.04266240000724793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.01,0.23770880699157715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,128,power_law_1.01,0.031014400720596313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,2,128,power_law_1.2,0.042656001448631284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.01,0.32179200649261475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,128,power_law_1.01,0.03315840065479279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,2,128,power_law_1.2,0.04442239999771118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,128,power_law_1.01,0.41490559577941893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,128,power_law_1.01,0.03368319869041443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,2,128,power_law_1.2,0.04405759871006012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,128,power_law_1.01,0.03537920117378235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,2,128,power_law_1.2,0.04787839949131012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,128,power_law_1.01,0.5611711978912354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,128,power_law_1.01,0.03827199935913086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,2,128,power_law_1.2,0.04718720018863678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,128,power_law_1.01,1.2804800033569337
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.01,0.03991039991378784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,2,128,power_law_1.2,0.04896000027656555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.01,0.04712960124015808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,2,128,power_law_1.2,0.05148800015449524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,2,128,power_law_1.2,0.05980160236358643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.01,0.04976640045642853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,2,128,power_law_1.2,0.06601600050926208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.01,0.059008002281188965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,2,128,power_law_1.2,0.07864959836006165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.01,0.07259520292282104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,2,128,power_law_1.2,0.10835200548171997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.01,0.09887999892234803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,128,balanced,0.6170133352279663
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,2,128,power_law_1.2,0.12819199562072753
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.01,0.12282240390777588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,2,128,power_law_1.2,0.17868800163269044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.01,0.16714880466461182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,2,128,power_law_1.2,0.22716159820556642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.01,0.21936640739440919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,2,128,power_law_1.2,0.33873920440673827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,128,power_law_1.01,0.26740479469299316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,2,128,power_law_1.2,0.4395008087158203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,128,power_law_1.01,0.39800319671630857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,2,128,power_law_1.2,0.5404416084289551
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,2,128,power_law_1.2,0.844115161895752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,128,power_law_1.01,0.7973440170288086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,2,128,power_law_1.2,1.6441856384277345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,2,128,power_law_1.2,0.03372159898281098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,2,128,power_law_1.2,0.033497598767280576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,2,128,power_law_1.2,0.0349375993013382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,2,128,power_law_1.2,0.03600000143051148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,2,128,power_law_1.2,0.037011200189590455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,2,128,power_law_1.2,0.03788160085678101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,2,128,power_law_1.2,0.03829759955406189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,2,128,power_law_1.2,0.03790720105171204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,128,balanced,0.8445280392964681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,2,128,power_law_1.2,0.03826560080051422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,2,128,power_law_1.2,0.038361600041389464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,2,128,power_law_1.2,0.03973119854927063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,2,128,power_law_1.2,0.040275201201438904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,2,128,power_law_1.2,0.04078719913959503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,2,128,power_law_1.2,0.04167680144309997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,2,128,power_law_1.2,0.044614401459693906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,2,128,power_law_1.2,0.04479359984397888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,2,128,power_law_1.2,0.04730879962444305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,2,128,power_law_1.2,0.053958398103713986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,2,128,power_law_1.2,0.0586624026298523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,2,128,power_law_1.2,0.0677183985710144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,2,128,power_law_1.2,0.08634240031242371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,2,128,power_law_1.2,0.11887999773025512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,2,128,power_law_1.2,0.1476032018661499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,2,128,power_law_1.2,0.20449280738830566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,2,128,power_law_1.2,0.2563647985458374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,2,128,power_law_1.2,0.39130239486694335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,2,128,power_law_1.2,0.5128511905670166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,2,128,power_law_1.2,0.6311999797821045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,2,128,power_law_1.2,1.0187264442443849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,2,128,power_law_1.2,1.9946367263793945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,2,128,power_law_1.2,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,2,128,power_law_1.2,0.03480319976806641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,2,128,power_law_1.2,0.03444480001926422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,2,128,power_law_1.2,0.036646398901939395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,2,128,power_law_1.2,0.036646398901939395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,2,128,power_law_1.2,0.037836799025535585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,2,128,power_law_1.2,0.037427198886871335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,2,128,power_law_1.2,0.03792639970779419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,2,128,power_law_1.2,0.037939199805259706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,2,128,power_law_1.2,0.037920001149177554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,2,128,power_law_1.2,0.03847039937973022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,2,128,power_law_1.2,0.03932160139083862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,2,128,power_law_1.2,0.04033919870853424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,2,128,power_law_1.2,0.041126400232315063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,2,128,power_law_1.2,0.04414080083370209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,2,128,power_law_1.2,0.04440320134162903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,2,128,power_law_1.2,0.04711039960384369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,2,128,power_law_1.2,0.051481598615646364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.2,0.05661439895629883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.2,0.06738560199737549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.2,0.07719680070877075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.2,0.10055680274963379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.2,0.1272063970565796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.2,0.16387840509414672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.2,0.20476160049438477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.2,0.30052480697631834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.2,0.4042816162109375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,2,128,power_law_1.2,0.4990079879760742
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,2,128,power_law_1.2,0.7756415843963623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,2,128,power_law_1.2,0.017638400197029114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,2,128,power_law_1.2,1.508236789703369
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,2,128,power_law_1.2,0.017759999632835387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,2,128,power_law_1.2,0.018489600718021394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,2,128,power_law_1.2,0.020204800367355346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,2,128,power_law_1.2,0.038201600313186646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,2,128,power_law_1.2,0.029363200068473816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,2,128,power_law_1.2,0.029516801238059998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,2,128,power_law_1.2,0.029516801238059998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,2,128,power_law_1.2,0.029817599058151244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,2,128,power_law_1.2,0.029868799448013305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,2,128,power_law_1.2,0.029523199796676634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,2,128,power_law_1.2,0.030796799063682555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,2,128,power_law_1.2,0.030828800797462464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,2,128,power_law_1.2,0.03128319978713989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,2,128,power_law_1.2,0.03340800106525421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,2,128,power_law_1.2,0.03366400003433227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,2,128,power_law_1.2,0.03532159924507141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,2,128,power_law_1.2,0.03771519958972931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.2,0.04020479917526245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.2,0.04669440090656281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.2,0.04965760111808777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.2,0.06101760268211365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.2,0.07882879972457886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.2,0.10277119874954224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.2,0.12914559841156006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.2,0.1784832000732422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.2,0.23072640895843505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,2,128,power_law_1.2,0.2816319942474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,2,128,power_law_1.2,0.42831997871398925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,2,128,power_law_1.2,0.8411392211914063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,64,balanced,0.02102400114138921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,64,balanced,0.01897066707412402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,64,balanced,0.01956266661485036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,64,balanced,0.0189280000825723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,64,balanced,0.01932266727089882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,64,balanced,0.023381332556406658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,64,balanced,0.023071999351183575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,64,balanced,0.029120000700155895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,64,balanced,0.02943466603755951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,64,balanced,0.030106666187445324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,64,balanced,0.029813334345817566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,64,balanced,0.03156800071398417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,64,balanced,0.03196266790231069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,64,balanced,0.033530667424201965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,64,balanced,0.035375999907652535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,64,balanced,0.03339199970165888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,64,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,64,balanced,0.03537066777547201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,64,balanced,0.035589332381884255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,64,balanced,0.03376533339420954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,64,balanced,0.0402399996916453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,64,balanced,0.03357866654793421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,64,balanced,0.035717333356539406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,64,balanced,0.043525333205858864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,64,balanced,0.03649600098530451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,64,balanced,0.03748266647259394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,64,balanced,0.050437331199645996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,64,balanced,0.035973332822322845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,64,balanced,0.035717333356539406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,64,balanced,0.0377866675456365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,64,balanced,0.05586666862169901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,64,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,64,balanced,0.037589333951473236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,64,balanced,0.037717332442601524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,64,balanced,0.07003200054168701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,64,balanced,0.039706667264302574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,64,balanced,0.04487466812133789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,64,balanced,0.0417546679576238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,64,balanced,0.08982933561007182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,64,balanced,0.04385599990685781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,64,balanced,0.04577599962552389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,64,balanced,0.047685335079828896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,64,balanced,0.04967466493447622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,64,balanced,0.11344533165295918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,64,balanced,0.05398400127887726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,64,balanced,0.05823466678460439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,64,balanced,0.14455999930699667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,64,balanced,0.06710933148860931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,64,balanced,0.08687999844551086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,64,balanced,0.2000160018603007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,64,balanced,0.10027199983596802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,64,balanced,0.2588533361752828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,64,balanced,0.1285760005315145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,64,balanced,0.3134773373603821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,64,balanced,0.1597760021686554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,64,balanced,0.031685332457224526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,64,balanced,0.036373332142829895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,64,balanced,0.031248000760873158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,64,balanced,0.03329599897066752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,64,balanced,0.033546666304270424
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,64,power_law_1.01,0.03694080114364624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,64,balanced,0.03149333347876867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,64,balanced,0.033333333830038704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,64,power_law_1.01,0.035071998834609985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,64,balanced,0.033802665770053864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,64,power_law_1.01,0.0345984011888504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,64,balanced,0.03350399931271871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,64,power_law_1.01,0.03783040046691895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,64,balanced,0.033530667424201965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,64,power_law_1.01,0.0380160003900528
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,64,balanced,0.03364266703526179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,64,power_law_1.01,0.03948160111904144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,64,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,64,power_law_1.01,0.03943040072917938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,64,balanced,0.03533866753180822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,64,power_law_1.01,0.03976959884166718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,64,balanced,0.03534399966398875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,64,power_law_1.01,0.039878401160240176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,64,balanced,0.031712000568707786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,64,balanced,0.03967999915281931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,64,power_law_1.01,0.04035840034484863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,64,balanced,0.18970133860905966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,64,balanced,0.037392000357309975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,64,power_law_1.01,0.04152320027351379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,64,balanced,0.03338133295377096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,64,balanced,0.03181866556406021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,64,power_law_1.01,0.04131200015544891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,64,balanced,0.03940266619126002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,64,balanced,0.4888746738433838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,64,balanced,0.031317333380381264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,64,power_law_1.01,0.04196479916572571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,64,balanced,0.034341332813103996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,64,power_law_1.01,0.04264959990978241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,64,balanced,0.038058665891488395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,64,balanced,0.03461333364248276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,64,power_law_1.01,0.0456959992647171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,64,balanced,0.03989866624275843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,64,balanced,0.03530666728814443
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,64,power_law_1.01,0.046412798762321475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,64,balanced,0.03379733363787333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,64,power_law_1.01,0.04970879852771759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,64,balanced,0.04176533222198486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,64,balanced,0.03438399980465571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,64,power_law_1.01,0.0511680006980896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,64,balanced,0.03583999971548716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.01,0.05824000239372253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,64,balanced,0.04180799921353658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,64,balanced,0.0354720006386439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.01,0.06401280164718628
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,64,balanced,0.036805334190527596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.01,0.07125120162963867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,64,balanced,0.035616000493367515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,64,balanced,0.04760533571243286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.01,0.08988159894943237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,64,balanced,0.03773866593837738
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.01,0.10486400127410889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,64,balanced,0.04177066683769226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.01,0.14131200313568115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,64,balanced,0.03974399964014689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.01,0.1897279977798462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,64,balanced,0.0499946673711141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,64,power_law_1.01,0.033497598767280576
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,64,balanced,0.04187199970086416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.01,0.2501312017440796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.01,0.3307391881942749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,64,power_law_1.01,0.03534719944000244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,64,balanced,0.04365866879622141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,64,power_law_1.01,0.030079999566078187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,64,power_law_1.01,0.3764672040939331
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,64,balanced,0.04568000137805939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,64,power_law_1.01,0.030854400992393494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,64,balanced,0.06365866462389629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,64,power_law_1.01,0.5671103954315185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,64,power_law_1.01,0.030662399530410767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,64,power_law_1.01,0.031174400448799135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,64,balanced,0.04757866760094961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,64,power_law_1.01,0.01998720020055771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,64,power_law_1.01,1.3268223762512208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,64,power_law_1.01,0.031615999341011045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,64,power_law_1.01,0.03160319924354553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,64,power_law_1.01,0.01744000017642975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,64,balanced,0.26026666164398193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,64,power_law_1.01,0.031436800956726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,64,balanced,0.052416001756985985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,64,power_law_1.01,0.018035200238227845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,64,power_law_1.01,0.031353598833084105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,64,power_law_1.01,0.017574399709701538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,64,balanced,0.07025066514809926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,64,power_law_1.01,0.01764480024576187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,64,power_law_1.01,0.03256320059299469
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,64,power_law_1.01,0.020787200331687926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,64,power_law_1.01,0.03365119993686676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,64,balanced,0.0664106657107671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,64,power_law_1.01,0.02101760059595108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,64,power_law_1.01,0.03482879996299744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,64,power_law_1.01,0.029497599601745604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,64,power_law_1.01,0.034835198521614076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,64,power_law_1.01,0.029945600032806396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,64,power_law_1.01,0.030137598514556885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,64,power_law_1.01,0.03818239867687225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,64,power_law_1.01,0.02999039888381958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,64,power_law_1.01,0.03821440041065216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,64,balanced,0.08711999654769897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,64,power_law_1.01,0.03130879998207092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,64,power_law_1.01,0.04084480106830597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,64,power_law_1.01,0.03079040050506592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,64,power_law_1.01,0.044998401403427125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,64,power_law_1.01,0.03084160089492798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,64,power_law_1.01,0.047443199157714847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,64,balanced,0.08549333612124126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,64,power_law_1.01,0.0323199987411499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,64,power_law_1.01,0.030316799879074097
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,64,power_law_1.01,0.054630398750305176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,64,power_law_1.01,0.033395200967788696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,64,power_law_1.01,0.06242560148239136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,64,power_law_1.01,0.0304639995098114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,64,power_law_1.01,0.035667198896408084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,64,balanced,0.11591999729474385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,64,power_law_1.01,0.07761279940605163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,64,power_law_1.01,0.03845759928226471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,64,power_law_1.01,0.02942720055580139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,64,power_law_1.01,0.08871039748191833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.01,0.04180479943752289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,64,power_law_1.01,0.030048000812530517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,64,power_law_1.01,0.13000960350036622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.01,0.048102399706840514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,64,power_law_1.01,0.031071999669075014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.01,0.051737600564956666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,64,power_law_1.01,0.14722559452056885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,64,power_law_1.01,0.03194240033626557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.01,0.0627135992050171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.01,0.08101119995117187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,64,power_law_1.01,0.032051199674606325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,64,power_law_1.01,0.22973439693450928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.01,0.10657279491424561
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,64,power_law_1.01,0.032691198587417605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,64,balanced,0.14044800400733948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,64,power_law_1.01,0.24907519817352294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.01,0.12764159440994263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,64,power_law_1.01,0.03289600014686585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.01,0.17443840503692626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,64,power_law_1.01,0.033344000577926636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,64,power_law_1.01,0.2659392118453979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,64,balanced,0.9492692947387695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.01,0.22986240386962892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,64,balanced,0.10663466652234395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,64,power_law_1.01,0.03442560136318207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,64,power_law_1.01,0.4662335872650146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,64,power_law_1.01,0.2948863983154297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,64,power_law_1.01,0.03544960021972656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,64,power_law_1.01,0.03656319975852966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,64,power_law_1.01,0.4253695964813232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,64,power_law_1.01,0.9145088195800781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,64,power_law_1.01,0.036671999096870425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,64,power_law_1.01,0.8693696022033691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,64,power_law_1.01,0.04083200097084046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,64,power_law_1.01,0.04193280041217804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,64,power_law_1.01,0.04397439956665039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,64,balanced,0.19099734226862589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,64,power_law_1.01,0.04782080054283142
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,64,power_law_1.01,0.05190399885177612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,64,power_law_1.01,0.05926399827003479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,64,power_law_1.01,0.06840959787368775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,64,power_law_1.01,0.0975488007068634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,64,power_law_1.01,0.12520960569381714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,64,balanced,0.12313600381215413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,64,power_law_1.01,0.1635967969894409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,64,power_law_1.01,0.19811199903488158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,64,balanced,0.24399999777475992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,64,power_law_1.01,0.2852288007736206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,64,power_law_1.01,0.38000640869140623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,64,power_law_1.01,0.4505919933319092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,64,balanced,0.47551465034484863
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,64,power_law_1.01,0.7114240169525147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,64,power_law_1.01,1.337497615814209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,64,balanced,0.29574400186538696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,64,balanced,0.15799466768900552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,64,balanced,0.435754656791687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,64,balanced,0.272597332795461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,64,balanced,0.80349334081014
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,4,64,power_law_1.2,0.02990719974040985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,4,64,power_law_1.2,0.03415679931640625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,4,64,power_law_1.2,0.028044798970222475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,4,64,power_law_1.2,0.031251201033592226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,4,64,power_law_1.2,0.030348798632621764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,4,64,power_law_1.2,0.031167998909950256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,4,64,power_law_1.2,0.031865599751472476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,4,64,power_law_1.2,0.031142398715019226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,4,64,power_law_1.2,0.031673601269721983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,4,64,power_law_1.2,0.031763198971748355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,4,64,power_law_1.2,0.03284479975700379
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,4,64,power_law_1.2,0.035129600763320924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,4,64,power_law_1.2,0.03521920144557953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,4,64,power_law_1.2,0.03496319949626923
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,4,64,power_law_1.2,0.039110401272773744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,4,64,power_law_1.2,0.038867199420928956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,4,64,power_law_1.2,0.04289920032024384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,4,64,power_law_1.2,0.049619200825691226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,4,64,power_law_1.2,0.056364798545837404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,4,64,power_law_1.2,0.06317440271377564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,4,64,power_law_1.2,0.08065919876098633
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,4,64,power_law_1.2,0.10767359733581543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,4,64,power_law_1.2,0.13075200319290162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,4,64,power_law_1.2,0.03049600124359131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,4096,1536,8,128,4,64,power_law_1.2,0.037145599722862244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,4,64,power_law_1.2,0.184825599193573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,4,64,power_law_1.2,0.030086401104927062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,4096,1536,8,128,4,64,power_law_1.2,0.0351936012506485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,4,64,power_law_1.2,0.03023360073566437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,4096,1536,8,128,4,64,power_law_1.2,0.03563520014286041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,4,64,power_law_1.2,0.1991744041442871
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,4,64,power_law_1.2,0.03134079873561859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,4096,1536,8,128,4,64,power_law_1.2,0.03813759982585907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,4,64,power_law_1.2,0.0321152001619339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,4,64,power_law_1.2,0.38356480598449705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,4096,1536,8,128,4,64,power_law_1.2,0.038047999143600464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,4,64,power_law_1.2,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,4096,1536,8,128,4,64,power_law_1.2,0.03964160084724426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,4,64,power_law_1.2,0.43427839279174807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,4,64,power_law_1.2,0.032979199290275575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,4096,1536,8,128,4,64,power_law_1.2,0.039577600359916684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,4,64,power_law_1.2,0.033632001280784606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,4096,1536,8,128,4,64,power_law_1.2,0.04068480134010315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,4,64,power_law_1.2,0.408512020111084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,4,64,power_law_1.2,0.03427839875221252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1,4096,1536,8,128,4,64,power_law_1.2,0.01990399956703186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,4096,1536,8,128,4,64,power_law_1.2,0.03953279852867127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,4,64,power_law_1.2,0.03461120128631592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,4096,1536,8,128,4,64,power_law_1.2,0.0404992014169693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,4,64,power_law_1.2,0.9070015907287597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2,4096,1536,8,128,4,64,power_law_1.2,0.017606399953365326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,4,64,power_law_1.2,0.035519999265670774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,4096,1536,8,128,4,64,power_law_1.2,0.04104959964752197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,4,64,power_law_1.2,0.036294400691986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,4096,1536,8,128,4,64,power_law_1.2,0.04161919951438904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4,4096,1536,8,128,4,64,power_law_1.2,0.018303999304771425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,4,64,power_law_1.2,0.03728640079498291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,4,64,power_law_1.2,1.6430784225463868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,4096,1536,8,128,4,64,power_law_1.2,0.04257279932498932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,4,64,power_law_1.2,0.03815680146217346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8,4096,1536,8,128,4,64,power_law_1.2,0.0176256000995636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,4096,1536,8,128,4,64,power_law_1.2,0.042803201079368594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,4,64,power_law_1.2,0.04270080029964447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,4096,1536,8,128,4,64,power_law_1.2,0.04686079919338226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,4,64,power_law_1.2,0.0447488009929657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16,4096,1536,8,128,4,64,power_law_1.2,0.01788160055875778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,4096,1536,8,128,4,64,power_law_1.2,0.04767360091209412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,4,64,power_law_1.2,0.04824959933757782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,4096,1536,8,128,4,64,power_law_1.2,0.051052802801132204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32,4096,1536,8,128,4,64,power_law_1.2,0.021209600567817687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,4,64,power_law_1.2,0.05234559774398804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,48,4096,1536,8,128,4,64,power_law_1.2,0.020953600108623505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,4096,1536,8,128,4,64,power_law_1.2,0.05414400100708008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,64,4096,1536,8,128,4,64,power_law_1.2,0.029926401376724244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,4,64,power_law_1.2,0.053855997323989865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.2,0.058745598793029784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,80,4096,1536,8,128,4,64,power_law_1.2,0.02991360127925873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.2,0.07167360186576843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,4,64,power_law_1.2,0.063155198097229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,96,4096,1536,8,128,4,64,power_law_1.2,0.030297601222991945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.2,0.08562560081481933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,4,64,power_law_1.2,0.07429760098457336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,128,4096,1536,8,128,4,64,power_law_1.2,0.0302592009305954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.2,0.10606080293655396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,4,64,power_law_1.2,0.11314560174942016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,160,4096,1536,8,128,4,64,power_law_1.2,0.031404799222946166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.2,0.12503679990768432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.2,0.17228800058364868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,192,4096,1536,8,128,4,64,power_law_1.2,0.031436800956726074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,4,64,power_law_1.2,0.14986239671707152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.2,0.23335680961608887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,256,4096,1536,8,128,4,64,power_law_1.2,0.030636799335479737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,4,64,power_law_1.2,0.21298561096191407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.2,0.38290560245513916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,4,64,power_law_1.2,0.2506112098693848
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,320,4096,1536,8,128,4,64,power_law_1.2,0.03313280045986176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.2,0.4437568187713623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,4,64,power_law_1.2,0.42220158576965333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,384,4096,1536,8,128,4,64,power_law_1.2,0.03377920091152191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,4096,1536,8,128,4,64,power_law_1.2,0.6046336174011231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,4,64,power_law_1.2,0.5315775871276855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,512,4096,1536,8,128,4,64,power_law_1.2,0.03532159924507141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,768,4096,1536,8,128,4,64,power_law_1.2,0.04025599956512451
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,4,64,power_law_1.2,0.6605120182037354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,4096,1536,8,128,4,64,power_law_1.2,0.8409728050231934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.2,0.04323840141296387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,4,64,power_law_1.2,1.1295488357543946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.2,0.04902400076389313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,4096,1536,8,128,4,64,power_law_1.2,2.050860786437988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.2,0.05230720043182373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,4,64,power_law_1.2,2.3521856307983398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.2,0.06652160286903382
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.2,0.08600320219993592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.2,0.11052160263061524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.2,0.1421056032180786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.2,0.19508479833602904
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.2,0.25407359600067136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,20480,4096,1536,8,128,4,64,power_law_1.2,0.29804799556732176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,32768,4096,1536,8,128,4,64,power_law_1.2,0.47783679962158204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow_min_latency,nvfp4,65536,4096,1536,8,128,4,64,power_law_1.2,0.9360192298889161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,32,balanced,0.031258667508761086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,32,balanced,0.03030933439731598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,32,balanced,0.027232001225153606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,32,balanced,0.029637334247430164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,32,balanced,0.03127466638882955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,32,balanced,0.0314026673634847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,32,balanced,0.03145600110292435
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,32,balanced,0.03364799916744232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,32,balanced,0.033386667569478355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,32,balanced,0.03339199970165888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,32,balanced,0.03380800038576126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,32,balanced,0.029493334392706554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,32,balanced,0.03382399926582972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,32,balanced,0.035386666655540466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,32,balanced,0.03164800008138021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,32,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,32,balanced,0.02941333254178365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,32,balanced,0.039503999054431915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,32,balanced,0.02956799914439519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,32,balanced,0.03754666695992152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,32,balanced,0.03302400062481562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,32,balanced,0.03366400053103765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,32,balanced,0.03941866755485535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,32,balanced,0.03401066611210505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,32,balanced,0.03789866715669632
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,32,balanced,0.0337119996547699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,32,balanced,0.033626665671666466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,32,balanced,0.04146666576464971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,32,balanced,0.035717333356539406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,32,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,32,balanced,0.04338666796684265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,32,balanced,0.03628266602754593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,32,balanced,0.035674666364987694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,32,balanced,0.03759466608365377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,32,balanced,0.0439626673857371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,32,balanced,0.041759997606277466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,32,power_law_1.2,0.030956798791885377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,32,balanced,0.04052799940109253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,32,balanced,0.042133331298828125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,32,balanced,0.050901333491007485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,32,power_law_1.2,0.029075199365615846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,32,power_law_1.2,0.028479999303817748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,32,balanced,0.043925335009892784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,32,power_law_1.2,0.030272001028060914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,32,balanced,0.04774933556715647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,32,power_law_1.2,0.03128319978713989
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,32,power_law_1.2,0.03219200074672699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,32,balanced,0.054485330979029335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,32,power_law_1.2,0.03237760066986084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,32,balanced,0.050154666105906166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,32,power_law_1.2,0.03342080116271973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,32,power_law_1.2,0.033241599798202515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,32,balanced,0.056320001681645714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,32,power_law_1.2,0.03402239978313446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,32,power_law_1.2,0.03612799942493439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,32,power_law_1.2,0.036620798707008365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,32,balanced,0.07210666437943776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,32,power_law_1.2,0.03818239867687225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,32,balanced,0.07050666709740956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,32,power_law_1.2,0.038278400897979736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,32,power_law_1.2,0.04248960018157959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,32,power_law_1.2,0.04429439902305603
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,32,power_law_1.2,0.04547199904918671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,32,balanced,0.0922933320204417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,32,power_law_1.2,0.04915199875831604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,32,power_law_1.2,0.05761280059814453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,32,balanced,0.08726933598518372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,32,power_law_1.2,0.06496000289916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,32,power_law_1.2,0.09345279932022095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,32,power_law_1.2,0.1271232008934021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,32,power_law_1.2,0.14636800289154053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,32,balanced,0.12482133507728577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,32,power_law_1.2,0.2188096046447754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,32,power_law_1.2,0.27268478870391843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,32,power_law_1.2,0.5524479866027832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,32,power_law_1.2,0.5537471771240234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,32,power_law_1.01,0.03609600067138672
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,32,power_law_1.2,0.632755184173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,32,balanced,0.10841066638628642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,32,power_law_1.01,0.032025599479675294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,32,power_law_1.2,1.1839936256408692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,32,power_law_1.01,0.029382398724555968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,32,balanced,0.15941333770751953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,32,power_law_1.01,0.030656000971794127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,32,power_law_1.2,2.3035968780517577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,32,power_law_1.01,0.031116798520088196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,32,power_law_1.01,0.03258880078792572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,32,power_law_1.01,0.03247359991073608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,32,power_law_1.01,0.03244799971580505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,32,power_law_1.01,0.032364800572395325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,32,power_law_1.01,0.032518398761749265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,32,power_law_1.01,0.03322240114212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,32,power_law_1.01,0.03522560000419617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,32,power_law_1.01,0.03600000143051148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,32,power_law_1.01,0.03702400028705597
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,32,power_law_1.01,0.04019840061664581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,32,balanced,0.2155839999516805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,32,power_law_1.01,0.041407999396324155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,32,balanced,0.1313920021057129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,32,power_law_1.01,0.04131839871406555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,32,power_law_1.01,0.04428159892559051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,32,power_law_1.01,0.04833920001983642
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,32,power_law_1.01,0.0541375994682312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,32,power_law_1.01,0.06485120058059693
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,32,power_law_1.01,0.08677120208740234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,32,power_law_1.01,0.09388800263404846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,32,power_law_1.01,0.12584960460662842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,32,power_law_1.01,0.14977279901504517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,32,power_law_1.01,0.25443201065063475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,8,32,power_law_1.01,0.032876798510551454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,32,power_law_1.01,0.28702080249786377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,8,32,power_law_1.01,0.031667199730873105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,32,power_law_1.01,0.3151679992675781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,32,balanced,0.268832008043925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,8,32,power_law_1.01,0.031788799166679385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,32,power_law_1.01,0.5344448089599609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,8,32,power_law_1.01,0.030707201361656188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,8,32,power_law_1.01,0.031744000315666196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,32,power_law_1.01,0.9774911880493165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,8,32,power_law_1.01,0.033139199018478394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,32,balanced,0.14904533823331198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,8,32,power_law_1.01,0.03344640135765076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,8,32,power_law_1.01,0.03408640027046204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,8,32,power_law_1.01,0.034092798829078674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,8,32,power_law_1.01,0.035180801153182985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,8,32,power_law_1.01,0.035872000455856326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,8,32,power_law_1.01,0.03726080060005188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,8,32,power_law_1.01,0.03814400136470795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,8,32,power_law_1.01,0.03914240002632141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,8,32,power_law_1.01,0.04273279905319214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,8,32,power_law_1.01,0.04344319999217987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,8,32,power_law_1.01,0.04519039988517761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,8,32,power_law_1.01,0.050367999076843264
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,8,32,power_law_1.01,0.05496960282325745
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,32,balanced,0.3232586582501729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,8,32,power_law_1.01,0.0590399980545044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,8,32,power_law_1.01,0.076665598154068
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,8,32,power_law_1.01,0.09703680276870727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,8,32,power_law_1.01,0.135315203666687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,8,32,power_law_1.01,0.18145279884338378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,8,32,power_law_1.01,0.23084800243377684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,8,32,power_law_1.01,0.3410367965698242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,8,32,power_law_1.01,0.4356351852416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,8,32,power_law_1.01,0.530790376663208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,32,power_law_1.01,0.7444799900054931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,32,balanced,0.2047040065129598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,32,power_law_1.01,1.6654016494750976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,8,32,power_law_1.2,0.035123199224472046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,8,32,power_law_1.2,0.02993279993534088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,8,32,balanced,0.47808531920115155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,8,32,power_law_1.2,0.029254400730133058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,8,32,power_law_1.2,0.028723201155662535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,8,32,power_law_1.2,0.030828800797462464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,8,32,power_law_1.2,0.031430399417877196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,8,32,power_law_1.2,0.031795200705528257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,8,32,power_law_1.2,0.03193599879741669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,8,32,power_law_1.2,0.03216640055179596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,8,32,power_law_1.2,0.03283199965953827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,8,32,power_law_1.2,0.03454720079898834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,8,32,power_law_1.2,0.035129600763320924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,8,32,power_law_1.2,0.036313599348068236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,8,32,power_law_1.2,0.03671680092811584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,8,32,power_law_1.2,0.039852800965309146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,8,32,power_law_1.2,0.04252159893512726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,8,32,power_law_1.2,0.04191359877586365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,8,32,power_law_1.2,0.04609920084476471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,8,32,power_law_1.2,0.04979200065135956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,8,32,power_law_1.2,0.05605120062828064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,8,32,power_law_1.2,0.07258880138397217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,8,32,power_law_1.2,0.09353600144386291
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,8,32,power_law_1.2,0.11815680265426635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,8,32,power_law_1.2,0.16132479906082153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,8,32,power_law_1.2,0.19270399808883668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,8,32,power_law_1.2,0.281548810005188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,8,32,power_law_1.2,0.37991039752960204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,8,32,power_law_1.2,0.5094336032867431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,8,32,power_law_1.2,0.6460415840148925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,32,power_law_1.2,1.5209407806396484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,8,32,balanced,0.363045334815979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,16,balanced,0.029296000798543293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,16,balanced,0.030837332208951313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,16,balanced,0.029626667499542236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,16,balanced,0.031712000568707786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,16,balanced,0.03329599897066752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,16,balanced,0.033258666594823204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,16,balanced,0.03149333347876867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,16,balanced,0.033610666791598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,16,balanced,0.033301333586374916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,16,balanced,0.033615998923778534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,8,32,balanced,0.9032586415608724
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,16,balanced,0.0336053321758906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,16,balanced,0.03356266766786575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,16,balanced,0.03531199942032496
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,16,balanced,0.0355679988861084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,16,balanced,0.03765333443880081
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,16,balanced,0.03770133356253306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,16,balanced,0.04053866614898046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,16,balanced,0.04382933179537455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,16,balanced,0.04599999884764353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,16,balanced,0.04804266492525736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,16,balanced,0.052239999175071716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,16,balanced,0.06020266811052958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,16,balanced,0.07034666836261749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,16,power_law_1.01,0.03320960104465485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,16,power_law_1.01,0.028339201211929323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,16,balanced,0.02962133288383484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,16,power_law_1.01,0.027449598908424376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,16,balanced,0.032431999842325844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,16,power_law_1.01,0.03059839904308319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,16,balanced,0.09098133444786072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,16,balanced,0.03140799949566523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,16,power_law_1.01,0.03073279857635498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,16,balanced,0.03136533250411352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,16,power_law_1.01,0.03229439854621887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,16,balanced,0.033802665770053864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,16,power_law_1.01,0.03226880133152008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,16,balanced,0.035589332381884255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,16,power_law_1.01,0.03270399868488312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,16,balanced,0.03606933355331421
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,16,power_law_1.01,0.03333120048046112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,16,balanced,0.035802667339642845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,16,power_law_1.01,0.033504000306129454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,16,balanced,0.11181867122650146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,16,balanced,0.0359253336985906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,16,balanced,0.03766400118668874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,16,power_law_1.01,0.033843201398849485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,16,balanced,0.03770666569471359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,16,power_law_1.01,0.03497599959373474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,16,balanced,0.03818666686614355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,16,power_law_1.01,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,16,balanced,0.03791466603676478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,16,power_law_1.01,0.03760640025138855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,16,balanced,0.03961600114901861
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,16,power_law_1.01,0.04063999950885773
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,16,power_law_1.01,0.041171199083328246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,16,balanced,0.04282666742801666
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,16,power_law_1.01,0.044896000623703004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,16,balanced,0.04171733558177948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,16,power_law_1.01,0.047660800814628604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,16,balanced,0.04386133452256521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,16,power_law_1.01,0.05565440058708191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,16,balanced,0.14704533418019614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,16,power_law_1.01,0.03844479918479919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,16,power_law_1.01,0.059622400999069215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,16,balanced,0.04721599817276001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,16,power_law_1.01,0.03127039968967438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,16,power_law_1.01,0.07021440267562866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,16,power_law_1.01,0.029862400889396668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,16,balanced,0.05192000170548757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,16,power_law_1.01,0.09463679790496826
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,16,power_law_1.01,0.03160319924354553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,16,power_law_1.01,0.10985599756240845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,16,power_law_1.01,0.032652801275253295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,16,balanced,0.055919999877611794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,16,power_law_1.01,0.13968000411987305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,16,power_law_1.01,0.03490560054779053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,16,power_law_1.01,0.03482879996299744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,16,power_law_1.01,0.18195199966430664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,16,power_law_1.01,0.0354559987783432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,16,power_law_1.01,0.23834240436553955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,16,balanced,0.06299200157324474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,16,power_law_1.01,0.035769599676132205
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,16,power_law_1.01,0.2977920055389404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,16,power_law_1.01,0.036620798707008365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,16,power_law_1.01,0.41841278076171873
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,16,power_law_1.01,0.037273600697517395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,16,balanced,0.08231466511885326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,16,power_law_1.01,0.5083839893341064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,16,power_law_1.01,0.0387584000825882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,16,balanced,0.1728480060895284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,16,power_law_1.01,0.04056319892406464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,16,power_law_1.01,1.008403205871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,16,power_law_1.01,0.040608000755310056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,16,power_law_1.01,0.04504320025444031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,16,power_law_1.01,0.04807040095329285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,16,balanced,0.10689066847165425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,16,power_law_1.01,0.048019200563430786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,16,power_law_1.01,0.052960002422332765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,16,power_law_1.01,0.06156799793243408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,16,power_law_1.01,0.07778559923171997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,16,power_law_1.01,0.07604479789733887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,16,balanced,0.14563733339309692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,16,power_law_1.01,0.1081663966178894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,16,power_law_1.01,0.13286399841308594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,16,power_law_1.01,0.20126080513000488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,16,power_law_1.01,0.2471679925918579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,16,16,power_law_1.2,0.03405439853668213
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,16,power_law_1.01,0.3777087926864624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,16,balanced,0.21526400248209634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,16,16,power_law_1.2,0.028121599555015565
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,16,power_law_1.01,0.4750207901000977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,16,balanced,0.1870186726252238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,16,16,power_law_1.2,0.028870400786399842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,16,power_law_1.01,0.6139711856842041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,16,16,power_law_1.2,0.030457600951194763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,16,power_law_1.01,0.9134336471557617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,16,16,power_law_1.2,0.031744000315666196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,16,16,power_law_1.2,0.033036801218986514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,16,power_law_1.01,1.6823423385620118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,16,16,power_law_1.2,0.03332479894161224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,16,16,power_law_1.2,0.03296000063419342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,16,16,power_law_1.2,0.03343360126018524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,16,16,power_law_1.2,0.03342719972133636
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,16,16,power_law_1.2,0.035366401076316833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,16,16,power_law_1.2,0.035718399286270144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,16,balanced,0.25601067145665485
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,16,16,power_law_1.2,0.0357120007276535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,16,16,power_law_1.2,0.03820799887180328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,16,16,power_law_1.2,0.04183039963245392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,16,16,power_law_1.2,0.04180479943752289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,16,16,power_law_1.2,0.04669440090656281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,16,16,power_law_1.2,0.05235199928283692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,16,16,power_law_1.2,0.05678079724311828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,16,16,power_law_1.2,0.07242239713668823
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,16,16,power_law_1.2,0.08390399813652039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,16,16,power_law_1.2,0.10293760299682617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,16,16,power_law_1.2,0.1282240033149719
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,16,balanced,0.302730659643809
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,16,balanced,0.3145493268966675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,16,16,power_law_1.2,0.20558719635009765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,16,16,power_law_1.2,0.22124159336090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,16,16,power_law_1.2,0.33255040645599365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,16,16,power_law_1.2,0.4366335868835449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,16,16,power_law_1.2,0.5724927902221679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,16,16,power_law_1.2,0.9549951553344727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,16,power_law_1.2,1.8648576736450195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,16,16,power_law_1.2,0.03706879913806915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,16,balanced,0.38621334234873456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,16,16,power_law_1.2,0.030035200715065002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,16,16,power_law_1.2,0.028454399108886717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,16,16,power_law_1.2,0.03061760067939758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,16,16,power_law_1.2,0.03185279965400696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,16,16,power_law_1.2,0.03356159925460815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,16,16,power_law_1.2,0.034380799531936644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,16,16,power_law_1.2,0.0352512001991272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,16,16,power_law_1.2,0.03601279854774475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,16,16,power_law_1.2,0.0361407995223999
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,16,16,power_law_1.2,0.037747201323509214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,16,16,power_law_1.2,0.03871360123157501
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,16,16,power_law_1.2,0.040249601006507874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,16,16,power_law_1.2,0.04149119853973389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,16,16,power_law_1.2,0.04627839922904968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,16,16,power_law_1.2,0.048281601071357726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,16,16,power_law_1.2,0.04919680058956146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,16,16,power_law_1.2,0.05392000079154968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,16,16,power_law_1.2,0.06390399932861328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,16,16,power_law_1.2,0.07640960216522216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,16,16,power_law_1.2,0.09196799993515015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,16,16,power_law_1.2,0.13000960350036622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,16,16,power_law_1.2,0.1620736002922058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,16,16,power_law_1.2,0.22834560871124268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,16,16,power_law_1.2,0.30813438892364503
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,16,balanced,0.5822240114212036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,16,16,power_law_1.2,0.4588031768798828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,16,16,power_law_1.2,0.564793586730957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,16,16,power_law_1.2,0.7943679809570312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,16,16,power_law_1.2,1.2331199645996094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,16,16,balanced,0.5551413297653198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,16,power_law_1.2,2.4961856842041015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,16,16,balanced,1.1082346439361572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,8,balanced,0.029509333272775013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,8,balanced,0.02951466788848241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,8,balanced,0.02938133229811986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,8,balanced,0.029482667644818623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,8,balanced,0.03197333216667175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,8,balanced,0.03465600063403448
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,8,balanced,0.0339626669883728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,8,balanced,0.03334933271010717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,8,balanced,0.03365866591533025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,8,balanced,0.035274667044480644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,8,balanced,0.03543466577927271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,8,balanced,0.03561066587766012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,8,balanced,0.03551466763019562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,8,balanced,0.035749333600203194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,8,balanced,0.04155733436346054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,8,balanced,0.03963200002908707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,8,balanced,0.04370133578777313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,8,balanced,0.047728002071380615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,8,balanced,0.04943466683228811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,8,balanced,0.058245331048965454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,8,balanced,0.062314664324124656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,8,balanced,0.08006933331489563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,8,balanced,0.10262399911880493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,8,balanced,0.029648000995318096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,8,balanced,0.12956266601880392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,8,balanced,0.031471999982992806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,8,balanced,0.02962133288383484
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,8,balanced,0.033439998825391136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,8,balanced,0.03626133253177007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,8,balanced,0.037765334049860634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,8,balanced,0.0382080003619194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,8,balanced,0.037503999968369804
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,8,balanced,0.03984533250331879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,8,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,8,balanced,0.15615999698638916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,8,balanced,0.039546666045983635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,8,balanced,0.041589332123597465
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,8,balanced,0.041562666495641075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,8,balanced,0.04358933369318644
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,8,balanced,0.04594666759173075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,8,balanced,0.04789333542188009
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,8,balanced,0.04763199885686239
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,8,balanced,0.05212800204753876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,8,balanced,0.06031466523806254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,8,balanced,0.21890133619308472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,8,balanced,0.06849599877993266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,8,balanced,0.07425066828727722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,8,balanced,0.10538132985432942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,8,balanced,0.13794133067131042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,8,balanced,0.3300480047861735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,8,power_law_1.01,0.0358271986246109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,8,balanced,0.18836265802383423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,8,power_law_1.01,0.03020159900188446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,8,power_law_1.01,0.029651200771331786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,8,power_law_1.01,0.03025279939174652
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,8,power_law_1.01,0.03209599852561951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,8,power_law_1.01,0.0342848002910614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,8,power_law_1.01,0.034585601091384886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,8,power_law_1.01,0.03457919955253601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,8,balanced,0.23021332422892252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,8,power_law_1.01,0.03472639918327332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,8,power_law_1.01,0.0356799989938736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,8,power_law_1.01,0.03583360016345978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,8,power_law_1.01,0.03740800023078918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,8,power_law_1.01,0.03845759928226471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,8,power_law_1.01,0.03896960020065308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,8,power_law_1.01,0.04293760061264038
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,8,power_law_1.01,0.046054399013519286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,8,balanced,0.4166133403778076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,8,power_law_1.01,0.05051519870758057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,8,power_law_1.01,0.05312640070915222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,8,power_law_1.01,0.061900800466537474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,8,balanced,0.3302239974339803
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,8,power_law_1.01,0.07649919986724854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,8,power_law_1.01,0.0820032000541687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,8,power_law_1.01,0.10398720502853394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,8,power_law_1.01,0.13819520473480223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,8,power_law_1.01,0.17095680236816407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,8,power_law_1.01,0.2536128044128418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,8,power_law_1.01,0.26990718841552735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,8,power_law_1.01,0.5236159801483155
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,8,power_law_1.01,0.5552576065063477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,8,power_law_1.01,0.8648256301879883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,8,power_law_1.01,1.8558784484863282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,8,balanced,0.4207199811935425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,8,balanced,0.6197653214136759
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,4096,1536,8,128,32,8,power_law_1.2,0.03653120100498199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,8,power_law_1.01,0.03306879997253418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,4096,1536,8,128,32,8,power_law_1.2,0.03111039996147156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,4096,1536,8,128,32,8,power_law_1.2,0.030623999238014222
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,8,power_law_1.01,0.032518398761749265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,4096,1536,8,128,32,8,power_law_1.2,0.03075839877128601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,8,power_law_1.01,0.03166080117225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,4096,1536,8,128,32,8,power_law_1.2,0.03208959996700287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,8,power_law_1.01,0.03242239952087402
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,8,balanced,0.5185920000076294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,4096,1536,8,128,32,8,power_law_1.2,0.034483200311660765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,8,power_law_1.01,0.03369599878787995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,4096,1536,8,128,32,8,power_law_1.2,0.034944000840187076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,8,power_law_1.01,0.03689599931240082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,4096,1536,8,128,32,8,power_law_1.2,0.034796801209449765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,8,power_law_1.01,0.037747201323509214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,4096,1536,8,128,32,8,power_law_1.2,0.03603839874267578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,8,power_law_1.01,0.03798399865627289
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,4096,1536,8,128,32,8,power_law_1.2,0.0363072007894516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,8,power_law_1.01,0.038899201154708865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,4096,1536,8,128,32,8,power_law_1.2,0.0364544004201889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,8,power_law_1.01,0.03932160139083862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,4096,1536,8,128,32,8,power_law_1.2,0.03733760118484497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,8,power_law_1.01,0.04221439957618713
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,4096,1536,8,128,32,8,power_law_1.2,0.03893760144710541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,8,power_law_1.01,0.0430976003408432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,8,power_law_1.01,0.04419200122356415
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,4096,1536,8,128,32,8,power_law_1.2,0.039750400185585025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,8,power_law_1.01,0.04379520118236542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,4096,1536,8,128,32,8,power_law_1.2,0.0440447986125946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,8,power_law_1.01,0.048876801133155824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,4096,1536,8,128,32,8,power_law_1.2,0.046982398629188536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,8,power_law_1.01,0.04919680058956146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,4096,1536,8,128,32,8,power_law_1.2,0.05027840137481689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,8,power_law_1.01,0.053388798236846925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,4096,1536,8,128,32,8,power_law_1.2,0.05663359761238098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,8,power_law_1.01,0.061894398927688596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,4096,1536,8,128,32,8,power_law_1.2,0.06427519917488098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,8,power_law_1.01,0.06970239877700805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,4096,1536,8,128,32,8,power_law_1.2,0.0787392020225525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,8,power_law_1.01,0.0843775987625122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,4096,1536,8,128,32,8,power_law_1.2,0.08880000114440918
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,8,power_law_1.01,0.10006400346755981
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,4096,1536,8,128,32,8,power_law_1.2,0.11908479928970336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,8,power_law_1.01,0.13177599906921386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,4096,1536,8,128,32,8,power_law_1.2,0.1460736036300659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,8,power_law_1.01,0.17052799463272095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,8,power_law_1.01,0.23690240383148192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,4096,1536,8,128,32,8,power_law_1.2,0.2716480016708374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,8,power_law_1.01,0.3204927921295166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,4096,1536,8,128,32,8,power_law_1.2,0.29882240295410156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,8,power_law_1.01,0.4347008228302002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,4096,1536,8,128,32,8,power_law_1.2,0.3475008010864258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,8,power_law_1.01,0.5473087787628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,4096,1536,8,128,32,8,power_law_1.2,0.5721536159515381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,8,power_law_1.01,0.6750847816467285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,32,balanced,0.054010664423306785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,8,balanced,0.7912320295969645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,4096,1536,8,128,32,8,power_law_1.2,0.6892096042633057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,8,power_law_1.01,1.1733887672424317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,32,balanced,0.04363733530044556
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,4096,1536,8,128,32,8,power_law_1.2,1.2979328155517578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,32,balanced,0.04154666761557261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,8,power_law_1.01,2.282598304748535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,8,power_law_1.2,2.3560064315795897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,32,balanced,0.04384533564249674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,32,balanced,0.04762133459250132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,32,balanced,0.04780266682306925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,32,balanced,0.049925332268079124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,32,balanced,0.050053333242734276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,32,balanced,0.05051200091838837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,32,balanced,0.049866666396458946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,32,balanced,0.05082666873931885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,32,balanced,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,32,balanced,0.051776001850763954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,32,balanced,0.05394133428732554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,4096,1536,8,128,32,8,balanced,1.1723199685414631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,32,balanced,0.05605333546797434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,32,balanced,0.0582826683918635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,32,balanced,0.05819733440876007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,32,balanced,0.06615466872851054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,4096,1536,8,128,32,8,power_law_1.2,0.03501439988613129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,32,balanced,0.06618666648864746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,4096,1536,8,128,32,8,power_law_1.2,0.030777600407600404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,32,balanced,0.0783786674340566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,4096,1536,8,128,32,8,power_law_1.2,0.033036801218986514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,4096,1536,8,128,32,8,power_law_1.2,0.03218559920787811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,32,balanced,0.09211200475692749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,4096,1536,8,128,32,8,power_law_1.2,0.03356159925460815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,4096,1536,8,128,32,8,power_law_1.2,0.03683840036392212
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,4096,1536,8,128,32,8,power_law_1.2,0.03751679956912994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,32,balanced,0.11423466602961223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,4096,1536,8,128,32,8,power_law_1.2,0.03832319974899292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,32,balanced,0.04394666850566864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,4096,1536,8,128,32,8,power_law_1.2,0.03938559889793396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,4096,1536,8,128,32,8,power_law_1.2,0.03950079977512359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,32,balanced,0.04362666606903076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,4096,1536,8,128,32,8,power_law_1.2,0.041503998637199405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,32,balanced,0.04427733520666758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,32,balanced,0.1274133324623108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,4096,1536,8,128,32,8,power_law_1.2,0.0427839994430542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,32,balanced,0.045184001326560974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,32,balanced,0.04785066843032837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,4096,1536,8,128,32,8,power_law_1.2,0.04684160053730011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,32,balanced,0.0497920016447703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,4096,1536,8,128,32,8,power_law_1.2,0.04521600008010864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,32,balanced,0.049882665276527405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,32,balanced,0.05184000233809153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,4096,1536,8,128,32,8,power_law_1.2,0.05251200199127197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,32,balanced,0.05150400102138519
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,4096,1536,8,128,32,8,power_law_1.2,0.0525439977645874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,32,balanced,0.05179733534653982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,32,balanced,0.1678826610247294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,4096,1536,8,128,32,8,power_law_1.2,0.05470079779624939
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,32,balanced,0.05213866631189982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,32,balanced,0.05402133365472158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,8,balanced,1.542202631632487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,4096,1536,8,128,32,8,power_law_1.2,0.061868798732757566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,32,balanced,0.052890668312708534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,4096,1536,8,128,32,8,power_law_1.2,0.0720255970954895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,4096,1536,8,128,32,8,power_law_1.2,0.08698239922523499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,32,balanced,0.056608001391092934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,4096,1536,8,128,32,8,power_law_1.2,0.10329600572586059
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,32,balanced,0.0582239975531896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,4096,1536,8,128,32,8,power_law_1.2,0.14714239835739135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,32,balanced,0.0603413333495458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,4096,1536,8,128,32,8,power_law_1.2,0.1960319995880127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,32,balanced,0.18336000045140585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,32,balanced,0.06246933341026306
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,4096,1536,8,128,32,8,power_law_1.2,0.29265921115875243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,4096,1536,8,128,32,8,power_law_1.2,0.31648640632629393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,32,balanced,0.06611733138561249
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,4096,1536,8,128,32,8,power_law_1.2,0.548908805847168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,32,balanced,0.07027733325958252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,4096,1536,8,128,32,8,power_law_1.2,0.7098944187164307
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,32,balanced,0.09100799759229024
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,4096,1536,8,128,32,8,power_law_1.2,0.8481599807739257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,4096,1536,8,128,32,8,power_law_1.2,1.3606016159057617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,32,balanced,0.1074773371219635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,32,balanced,0.2563040057818095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,4096,1536,8,128,32,8,power_law_1.2,3.2759552001953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,32,balanced,0.1485919952392578
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,32,balanced,0.17409066359202066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,32,balanced,0.2327786684036255
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,32,balanced,0.3038453261057536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,32,balanced,0.2793546716372172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,32,power_law_1.01,0.06037120223045349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,32,power_law_1.01,0.04573439955711365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,32,power_law_1.01,0.04232319891452789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,32,power_law_1.01,0.045798400044441225
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,32,power_law_1.01,0.04702079892158508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,32,power_law_1.01,0.04891520142555237
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,32,power_law_1.01,0.04896000027656555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,32,balanced,0.35715198516845703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,32,power_law_1.01,0.05014399886131286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,32,balanced,0.39235198497772217
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,32,power_law_1.01,0.04955520033836365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,32,power_law_1.01,0.050271999835968015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,32,power_law_1.01,0.0514303982257843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,32,power_law_1.01,0.05264639854431152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,32,power_law_1.01,0.052748799324035645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,32,power_law_1.01,0.05426560044288635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,32,power_law_1.01,0.06304640173912049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,32,power_law_1.01,0.06353920102119445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,32,power_law_1.01,0.06908800005912781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,32,power_law_1.01,0.077183997631073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,32,power_law_1.01,0.08330879807472229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,32,power_law_1.01,0.10060800313949585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,32,balanced,0.49741868178049725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,32,power_law_1.01,0.11689599752426147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,32,power_law_1.01,0.14247679710388184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,32,power_law_1.01,0.16546560525894166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,32,power_law_1.01,0.21806719303131103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,32,power_law_1.01,0.2890048027038574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,32,power_law_1.01,0.38204159736633303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,32,power_law_1.01,0.4881408214569092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,32,power_law_1.01,0.5911231994628906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,32,balanced,0.536138653755188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,32,power_law_1.01,1.062547206878662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,32,power_law_1.01,2.0176767349243163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,32,balanced,0.6015733480453491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,32,power_law_1.01,0.04642559885978699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,8,32,power_law_1.2,0.055052798986434934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,32,power_law_1.01,0.04744960069656372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,32,power_law_1.01,0.04337919950485229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,8,32,power_law_1.2,0.04226559996604919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,32,power_law_1.01,0.04656639993190766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,8,32,power_law_1.2,0.042131200432777405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,32,power_law_1.01,0.04747520089149475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,8,32,power_law_1.2,0.0448063999414444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,32,power_law_1.01,0.04911360144615173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,8,32,power_law_1.2,0.04699519872665405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,32,power_law_1.01,0.05016319751739502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,8,32,power_law_1.2,0.04829440116882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,8,32,power_law_1.2,0.04853120148181915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,32,power_law_1.01,0.050937598943710326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,8,32,power_law_1.2,0.05002880096435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,32,power_law_1.01,0.05072640180587769
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,8,32,power_law_1.2,0.0500927984714508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,32,power_law_1.01,0.05195519924163818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,8,32,power_law_1.2,0.05006719827651977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,32,power_law_1.01,0.05355520248413086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,8,32,power_law_1.2,0.05115519762039185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,32,power_law_1.01,0.05454080104827881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,8,32,power_law_1.2,0.05359359979629517
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,32,balanced,0.9200479984283447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,32,power_law_1.01,0.055270397663116456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,8,32,power_law_1.2,0.05496320128440857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,8,32,power_law_1.2,0.04330880045890808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,32,power_law_1.01,0.05794559717178345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,8,32,power_law_1.2,0.05742719769477844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,32,power_law_1.01,0.06291199922561645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,8,32,power_law_1.2,0.06282879710197449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,8,32,power_law_1.2,0.04191359877586365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,32,power_law_1.01,0.06564480066299438
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,8,32,power_law_1.2,0.06670719981193543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,32,power_law_1.01,0.06655359864234925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,8,32,power_law_1.2,0.04031359851360321
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,8,32,power_law_1.2,0.07052800059318542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,32,power_law_1.01,0.07581440210342408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,8,32,power_law_1.2,0.08687360286712646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,8,32,power_law_1.2,0.04286719858646393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,32,power_law_1.01,0.08465279936790467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,8,32,power_law_1.2,0.09258880019187928
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,32,power_law_1.01,0.10828160047531128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,8,32,power_law_1.2,0.04439040124416351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,8,32,power_law_1.2,0.11284480094909669
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,8,32,power_law_1.2,0.046623998880386354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,32,power_law_1.01,0.13768320083618163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,8,32,power_law_1.2,0.13537919521331787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,8,32,power_law_1.2,0.04759680032730103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,32,power_law_1.01,0.17620480060577393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,8,32,power_law_1.2,0.15143040418624878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,8,32,power_law_1.2,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,32,power_law_1.01,0.21682560443878174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,8,32,power_law_1.2,0.048895999789237976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,8,32,power_law_1.2,0.2077888011932373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,32,power_law_1.01,0.291814398765564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,8,32,power_law_1.2,0.04967679977416992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,8,32,power_law_1.2,0.2634943962097168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,32,balanced,1.0173813501993816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,32,power_law_1.01,0.36636159420013426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,8,32,power_law_1.2,0.051545602083206174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,8,32,power_law_1.2,0.3547391891479492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,32,power_law_1.01,0.5449024200439453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,8,32,power_law_1.2,0.05284479856491089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,8,32,power_law_1.2,0.5012288093566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,32,power_law_1.01,0.6379263877868653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,8,32,power_law_1.2,0.05303040146827698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,8,32,power_law_1.2,0.05687680244445801
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,8,32,power_law_1.2,0.7008448123931885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,32,power_law_1.01,0.847321605682373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,8,32,power_law_1.2,0.06204800009727478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,8,32,power_law_1.2,0.8001791954040527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,32,power_law_1.01,1.2576000213623046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,8,32,power_law_1.2,0.06542720198631287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,8,32,power_law_1.2,0.06490240097045899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,8,32,power_law_1.2,1.4661888122558593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,8,32,power_law_1.2,0.07836160063743591
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,32,power_law_1.01,2.4817535400390627
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,8,32,power_law_1.2,0.08538240194320679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,8,32,power_law_1.2,2.7739200592041016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,8,32,power_law_1.2,0.11034879684448243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,8,32,power_law_1.2,0.13778560161590575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,8,32,power_law_1.2,0.184934401512146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,8,32,power_law_1.2,0.2370431900024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,8,32,power_law_1.2,0.32799999713897704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,8,32,power_law_1.2,0.424729585647583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,8,32,power_law_1.2,0.6357632160186768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,8,32,power_law_1.2,0.786137580871582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,8,32,power_law_1.2,1.0518207550048828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,8,32,power_law_1.2,1.6598527908325196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,32,power_law_1.2,3.1664127349853515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,8,32,balanced,1.7899999618530273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,16,balanced,0.041519999504089355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,16,balanced,0.04162666698296865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,16,balanced,0.04341333111127218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,16,balanced,0.04560000201066335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,16,balanced,0.04658666749795278
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,16,balanced,0.051039998730023704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,16,balanced,0.05204799771308899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,16,balanced,0.05178666611512502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,16,balanced,0.05198933184146881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,16,balanced,0.053344001372655235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,16,balanced,0.05198933184146881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,16,balanced,0.05299733579158783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,16,balanced,0.05392000079154968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,16,balanced,0.05605333546797434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,16,balanced,0.05929600199063619
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,16,balanced,0.0622026671965917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,16,balanced,0.06275199850400288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,16,balanced,0.07014933228492737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,16,balanced,0.07050666709740956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,16,balanced,0.0925600032011668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,16,balanced,0.04561600089073181
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,16,balanced,0.04563199977080027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,16,balanced,0.10338667035102844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,16,balanced,0.04544533292452494
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,16,balanced,0.04560000201066335
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,16,balanced,0.04962133367856344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,16,balanced,0.05192000170548757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,16,balanced,0.1330400009950002
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,16,balanced,0.05203733344872793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,16,balanced,0.053871999184290566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,16,balanced,0.05446400245030721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,16,balanced,0.05401599903901418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,16,balanced,0.144896000623703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,16,balanced,0.05585599939028422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,16,balanced,0.05602666735649109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,16,balanced,0.055829331278800964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,16,balanced,0.05807466804981232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,16,balanced,0.062261333068211876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,16,balanced,0.0639519989490509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,16,balanced,0.19274133443832397
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,16,balanced,0.0663679987192154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,16,balanced,0.07399466633796692
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,16,balanced,0.07653866708278656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,16,balanced,0.10033599535624187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,16,balanced,0.2225653330485026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,16,balanced,0.11788266897201538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,16,balanced,0.16867733001708984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,16,balanced,0.19134400288263956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,16,balanced,0.30396799246470135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,16,balanced,0.26332799593607586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,16,power_law_1.01,0.051123201847076416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,16,power_law_1.01,0.04778240025043488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,16,balanced,0.3141706585884094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,16,power_law_1.01,0.043993601202964784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,16,power_law_1.01,0.04471679925918579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,16,power_law_1.01,0.04642559885978699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,16,balanced,0.3742239872614543
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,16,power_law_1.01,0.04909439980983734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,16,power_law_1.01,0.05053439736366272
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,16,power_law_1.01,0.05137280225753784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,16,power_law_1.01,0.05235840082168579
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,16,power_law_1.01,0.05169919729232788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,16,power_law_1.01,0.053574401140213015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,16,power_law_1.01,0.05501440167427063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,16,power_law_1.01,0.05708799958229065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,16,balanced,0.4444426695505778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,16,power_law_1.01,0.06083199977874756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,16,power_law_1.01,0.0661952018737793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,16,power_law_1.01,0.06964480280876159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,16,power_law_1.01,0.07018240094184876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,16,power_law_1.01,0.08376960158348083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,16,power_law_1.01,0.08871039748191833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,16,power_law_1.01,0.10631680488586426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,16,power_law_1.01,0.13137919902801515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,16,power_law_1.01,0.16184959411621094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,16,balanced,0.4433279832204183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,16,power_law_1.01,0.18733439445495606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,16,power_law_1.01,0.2635584115982056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,16,power_law_1.01,0.32857599258422854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,16,balanced,0.5649386644363403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,16,power_law_1.01,0.4351039886474609
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,16,power_law_1.01,0.6050496101379395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,16,power_law_1.01,0.6644991874694824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,16,power_law_1.01,1.107487964630127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,16,power_law_1.01,2.105190467834473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,16,balanced,0.6963573296864828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,16,balanced,0.6669867038726807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,16,power_law_1.01,0.05175039768218994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,16,power_law_1.01,0.05080320239067078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,16,power_law_1.01,0.0441536009311676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,16,power_law_1.01,0.04588159918785095
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,16,power_law_1.01,0.046348801255226134
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,16,power_law_1.01,0.05002880096435547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,16,power_law_1.01,0.050323200225830075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,16,balanced,1.0561760266621907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,16,power_law_1.01,0.05173119902610779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,16,power_law_1.01,0.05198079943656921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,16,power_law_1.01,0.05358080267906189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,16,power_law_1.01,0.05482879877090454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,16,power_law_1.01,0.05618559718132019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,16,16,power_law_1.2,0.05237119793891907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,16,power_law_1.01,0.05733759999275208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,16,power_law_1.01,0.05834239721298218
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,16,16,power_law_1.2,0.04761599898338318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,16,power_law_1.01,0.06350719928741455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,16,16,power_law_1.2,0.04535680115222931
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,16,power_law_1.01,0.06587520241737366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,16,16,power_law_1.2,0.046623998880386354
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,16,power_law_1.01,0.06871039867401123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,16,16,power_law_1.2,0.04750719964504242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,16,power_law_1.01,0.08012160062789916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,16,16,power_law_1.2,0.05192959904670715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,16,power_law_1.01,0.08848639726638793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,16,16,power_law_1.2,0.05204480290412903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,16,power_law_1.01,0.12059520483016968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,16,16,power_law_1.2,0.05415679812431336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,16,power_law_1.01,0.1418239951133728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,16,16,power_law_1.2,0.05314559936523437
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,16,power_law_1.01,0.19127039909362792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,16,16,power_law_1.2,0.05397120118141174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,16,power_law_1.01,0.2421056032180786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,16,16,power_law_1.2,0.056729602813720706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,16,16,power_law_1.2,0.0523904025554657
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,16,16,power_law_1.2,0.05912960171699524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,16,power_law_1.01,0.30963840484619143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,16,16,power_law_1.2,0.06123520135879516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,16,power_law_1.01,0.4211584091186523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,16,16,power_law_1.2,0.06405760049819946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,16,16,power_law_1.2,0.04693120121955872
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,16,power_law_1.01,0.6327040195465088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,16,16,power_law_1.2,0.06981120109558106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,16,16,power_law_1.2,0.04549759924411774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,16,16,power_law_1.2,0.07623040080070495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,16,power_law_1.01,0.8097920417785645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,16,16,power_law_1.2,0.04622080028057098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,16,16,power_law_1.2,0.07888640165328979
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,16,16,power_law_1.2,0.04700160026550293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,16,power_law_1.01,0.9181119918823242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,16,16,power_law_1.2,0.0988864004611969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,16,16,power_law_1.2,0.05028480291366577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,16,16,power_law_1.2,0.0991104006767273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,16,power_law_1.01,1.4136575698852538
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,16,16,power_law_1.2,0.1214400053024292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,16,16,power_law_1.2,0.0514303982257843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,16,16,power_law_1.2,0.13575680255889894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,16,16,power_law_1.2,0.052799999713897705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,16,16,power_law_1.2,0.1691264033317566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,16,power_law_1.01,2.7968767166137694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,16,16,power_law_1.2,0.05368959903717041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,16,16,power_law_1.2,0.22140800952911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,16,16,power_law_1.2,0.05511680245399475
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,16,16,power_law_1.2,0.31931519508361816
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,16,16,power_law_1.2,0.05651199817657471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,16,balanced,1.2814826965332031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,16,16,power_law_1.2,0.3491328001022339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,16,16,power_law_1.2,0.05774719715118408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,16,16,power_law_1.2,0.583513593673706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,16,16,power_law_1.2,0.058880001306533813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,8,balanced,0.04160533348719279
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,16,16,power_law_1.2,0.8878399848937988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,16,16,power_law_1.2,0.060096001625061034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,16,16,power_law_1.2,0.06534399986267089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,8,balanced,0.04154133299986521
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,16,16,power_law_1.2,0.9252415657043457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,8,balanced,0.04526400069395701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,16,16,power_law_1.2,0.06874880194664001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,8,balanced,0.047637333472569786
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,16,16,power_law_1.2,0.07179520130157471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,16,16,power_law_1.2,1.9643135070800781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,8,balanced,0.051776001850763954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,8,balanced,0.056330665946006775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,16,16,power_law_1.2,0.08494719862937927
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,8,balanced,0.060090666015942894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,16,16,power_law_1.2,3.1634687423706054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,16,16,power_law_1.2,0.09547520279884339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,8,balanced,0.0569706658522288
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,8,balanced,0.057914664347966514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,16,16,power_law_1.2,0.12604160308837892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,8,balanced,0.05797866483529409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,16,16,power_law_1.2,0.16355839967727662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,8,balanced,0.06001066664854685
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,8,balanced,0.05796800057093302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,16,16,power_law_1.2,0.20179200172424316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,8,balanced,0.06069866816202799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,16,16,power_law_1.2,0.24641919136047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,8,balanced,0.06234666705131531
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,8,balanced,0.04359466830889384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,16,16,power_law_1.2,0.34226560592651367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,8,balanced,0.06619733572006226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,16,balanced,2.0665012995402017
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,16,16,power_law_1.2,0.44896640777587893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,8,balanced,0.04465066889921824
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,8,balanced,0.06818666557470958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,8,balanced,0.04566933214664459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,8,balanced,0.07028266787528992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,16,16,power_law_1.2,0.7288127899169922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,8,balanced,0.049957334995269775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,8,balanced,0.052282666166623436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,8,balanced,0.08085333307584126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,16,16,power_law_1.2,0.8762751579284668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,8,balanced,0.05589866638183594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,8,balanced,0.05820266902446747
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,8,balanced,0.057989334066708885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,16,16,power_law_1.2,1.1443519592285156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,8,balanced,0.08562133709589641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,8,balanced,0.05863999823729197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,8,balanced,0.06002133091290792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,16,16,power_law_1.2,1.909324836730957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,8,balanced,0.061386664708455406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,8,balanced,0.11142399907112122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,8,balanced,0.060234665870666504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,8,balanced,0.06337066491444905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,16,16,power_law_1.2,4.117491149902344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,8,balanced,0.06423466900984447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,8,balanced,0.12361600001653035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,8,balanced,0.06856533388296764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,8,balanced,0.16823466618855795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,8,balanced,0.07082133491834004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,8,balanced,0.07229866584142049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,8,balanced,0.0831573357184728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,8,balanced,0.18121065696080527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,8,balanced,0.08866666754086812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,8,balanced,0.12001066406567891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,8,balanced,0.14386666814486185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,8,balanced,0.2569706638654073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,8,balanced,0.19914666811625162
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,8,balanced,0.2983466585477193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,8,balanced,0.22482667366663614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,8,balanced,0.3162879943847656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,8,balanced,0.4039733409881592
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,8,balanced,0.3932960033416748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,8,power_law_1.01,0.04682239890098572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,8,power_law_1.01,0.050169599056243894
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,8,balanced,0.5042879978815714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,8,power_law_1.01,0.043635201454162595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,8,balanced,0.5509973367055258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,8,power_law_1.01,0.04752640128135681
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,8,power_law_1.01,0.050195199251174924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,8,power_law_1.01,0.05424000024795532
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,8,power_law_1.01,0.054816001653671266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,8,power_law_1.01,0.05692800283432007
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,8,power_law_1.01,0.057171201705932616
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,8,power_law_1.01,0.05783039927482605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,8,power_law_1.01,0.05891839861869812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,8,power_law_1.01,0.06119040250778198
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,8,balanced,0.7065227031707764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,8,power_law_1.01,0.06492159962654113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,8,power_law_1.01,0.06859520077705383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,8,balanced,0.6285386482874552
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,8,power_law_1.01,0.0742143988609314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,8,power_law_1.01,0.07689599990844727
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,8,power_law_1.01,0.07722880244255066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,8,power_law_1.01,0.08975359797477722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,8,power_law_1.01,0.10060160160064698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,8,power_law_1.01,0.12163840532302857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,8,power_law_1.01,0.15448319911956787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,8,power_law_1.01,0.19104000329971313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,8,power_law_1.01,0.23559041023254396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,8,power_law_1.01,0.3212480068206787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,8,power_law_1.01,0.3578304052352905
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,8,balanced,0.8694132963816324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,8,power_law_1.01,0.5678912162780761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,8,power_law_1.01,0.6918784141540527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,8,power_law_1.01,0.8768063545227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,8,power_law_1.01,1.2707072257995606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,8,power_law_1.01,2.6126144409179686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,8,balanced,0.9243413607279459
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,8,power_law_1.01,0.043161600828170776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,8,power_law_1.01,0.04689919948577881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,8,power_law_1.01,0.04081279933452606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,8,power_law_1.01,0.04516479969024658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,8,balanced,1.3345279693603516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,8,power_law_1.01,0.047200000286102294
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,8,power_law_1.01,0.052083200216293334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,8,power_law_1.01,0.05419520139694214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,8,power_law_1.01,0.05534719824790955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,8,power_law_1.01,0.0554751992225647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,8,power_law_1.01,0.056409597396850586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,8,power_law_1.01,0.058406400680541995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,8,power_law_1.01,0.05989760160446167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,8,power_law_1.01,0.06159999966621399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,8,power_law_1.01,0.06492159962654113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,8,power_law_1.01,0.07015039920806884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,8,power_law_1.01,0.07341439723968506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,8,power_law_1.01,0.07985280156135559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,8,power_law_1.01,0.09413120150566101
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,8,power_law_1.01,0.10327039957046509
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,6144,2560,8,160,32,8,power_law_1.2,0.0469184011220932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,6144,2560,8,160,32,8,power_law_1.2,0.04627200067043304
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,8,power_law_1.01,0.1380031943321228
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,8,power_law_1.01,0.16621439456939696
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,6144,2560,8,160,32,8,power_law_1.2,0.04242559969425201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,6144,2560,8,160,32,8,power_law_1.2,0.04243200123310089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,8,power_law_1.01,0.22364161014556885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,6144,2560,8,160,32,8,power_law_1.2,0.042534399032592776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,8,power_law_1.01,0.2927295923233032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,6144,2560,8,160,32,8,power_law_1.2,0.043654400110244754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,6144,2560,8,160,32,8,power_law_1.2,0.047328001260757445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,128,balanced,0.06669866542021434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,8,power_law_1.01,0.4012288093566895
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,6144,2560,8,160,32,8,power_law_1.2,0.04901759922504425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,6144,2560,8,160,32,8,power_law_1.2,0.047276800870895384
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,6144,2560,8,160,32,8,power_law_1.2,0.0495743989944458
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,6144,2560,8,160,32,8,power_law_1.2,0.05498239994049072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,128,balanced,0.06634133557478587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,8,power_law_1.01,0.514796781539917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,6144,2560,8,160,32,8,power_law_1.2,0.054604798555374146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,128,balanced,0.06635733445485432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,6144,2560,8,160,32,8,power_law_1.2,0.05425919890403748
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,6144,2560,8,160,32,8,power_law_1.2,0.055430400371551516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,128,balanced,0.06823466718196869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,8,power_law_1.01,0.6776768207550049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,6144,2560,8,160,32,8,power_law_1.2,0.05575039982795715
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,128,balanced,0.06845333178838094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,6144,2560,8,160,32,8,power_law_1.2,0.05713919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,128,balanced,0.0769760012626648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,6144,2560,8,160,32,8,power_law_1.2,0.05649920105934143
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,8,power_law_1.01,0.930515193939209
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,6144,2560,8,160,32,8,power_law_1.2,0.058355200290679934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,128,balanced,0.08955732981363933
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,6144,2560,8,160,32,8,power_law_1.2,0.05775359869003296
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,6144,2560,8,160,32,8,power_law_1.2,0.05917440056800842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,128,balanced,0.08896533648173015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,8,power_law_1.01,1.1610112190246582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,6144,2560,8,160,32,8,power_law_1.2,0.060057598352432254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,128,balanced,0.0963253378868103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,6144,2560,8,160,32,8,power_law_1.2,0.06259840130805969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,128,balanced,0.09089600046475728
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,6144,2560,8,160,32,8,power_law_1.2,0.06114559769630432
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,6144,2560,8,160,32,8,power_law_1.2,0.0633408010005951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,8,power_law_1.01,1.7447935104370118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,128,balanced,0.09132267038027446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,6144,2560,8,160,32,8,power_law_1.2,0.06638079881668091
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,6144,2560,8,160,32,8,power_law_1.2,0.06480000019073487
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,128,balanced,0.09913067022959392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,6144,2560,8,160,32,8,power_law_1.2,0.06813439726829529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,6144,2560,8,160,32,8,power_law_1.2,0.07065600156784058
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,8,power_law_1.01,3.2670654296875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,128,balanced,0.05611733098824819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,128,balanced,0.09700799981753032
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,8,balanced,1.7964000701904297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,6144,2560,8,160,32,8,power_law_1.2,0.0730239987373352
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,6144,2560,8,160,32,8,power_law_1.2,0.07605119943618774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,128,balanced,0.09477866689364116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,6144,2560,8,160,32,8,power_law_1.2,0.07685120105743408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,6144,2560,8,160,32,8,power_law_1.2,0.07978879809379577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,128,balanced,0.09886399904886882
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,128,balanced,0.056645333766937256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,6144,2560,8,160,32,8,power_law_1.2,0.08049280047416688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,6144,2560,8,160,32,8,power_law_1.2,0.08256639838218689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,128,balanced,0.0562720000743866
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,6144,2560,8,160,32,8,power_law_1.2,0.09806079864501953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,128,balanced,0.09905067086219788
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,128,balanced,0.05735999842484792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,6144,2560,8,160,32,8,power_law_1.2,0.11000959873199463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,6144,2560,8,160,32,8,power_law_1.2,0.09093120098114013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,128,balanced,0.06035199761390686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,128,balanced,0.10453333457310994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,6144,2560,8,160,32,8,power_law_1.2,0.13963520526885986
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,6144,2560,8,160,32,8,power_law_1.2,0.1028864026069641
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,128,balanced,0.06649066507816315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,8,balanced,2.606320063273112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,6144,2560,8,160,32,8,power_law_1.2,0.12683520317077637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,128,balanced,0.06862399975458781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,128,balanced,0.11541333794593811
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,6144,2560,8,160,32,8,power_law_1.2,0.17569279670715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,128,balanced,0.06900266806284587
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,6144,2560,8,160,32,8,power_law_1.2,0.16721919775009156
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,6144,2560,8,160,32,8,power_law_1.2,0.23811841011047363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,128,balanced,0.11716799934705098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,6144,2560,8,160,32,8,power_law_1.2,0.19945600032806396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,128,balanced,0.06924266616503398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,128,balanced,0.07241599758466084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,6144,2560,8,160,32,8,power_law_1.2,0.27536640167236326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,6144,2560,8,160,32,8,power_law_1.2,0.3017087936401367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,128,balanced,0.1318666636943817
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,128,balanced,0.07264000177383423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,6144,2560,8,160,32,8,power_law_1.2,0.36918399333953855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,6144,2560,8,160,32,8,power_law_1.2,0.4067647933959961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,128,balanced,0.07063999772071838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,6144,2560,8,160,32,8,power_law_1.2,0.4558527946472168
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,128,balanced,0.0745600014925003
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,6144,2560,8,160,32,8,power_law_1.2,0.4904831886291504
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,128,balanced,0.07675200204054515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,128,balanced,0.14314132928848267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,6144,2560,8,160,32,8,power_law_1.2,0.6082623958587646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,6144,2560,8,160,32,8,power_law_1.2,0.7337920188903808
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,128,balanced,0.08084266881148021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,6144,2560,8,160,32,8,power_law_1.2,0.8340543746948242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,6144,2560,8,160,32,8,power_law_1.2,1.044051170349121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,128,balanced,0.08229866623878479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,6144,2560,8,160,32,8,power_law_1.2,1.136774444580078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,6144,2560,8,160,32,8,power_law_1.2,1.3061823844909668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,128,balanced,0.0890933374563853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,128,balanced,0.17299199104309082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,6144,2560,8,160,32,8,power_law_1.2,1.4483136177062987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,128,balanced,0.09974933664004008
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,6144,2560,8,160,32,8,power_law_1.2,2.070969581604004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,128,balanced,0.1108746627966563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,128,balanced,0.2009226679801941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,6144,2560,8,160,32,8,power_law_1.2,3.4393600463867187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,6144,2560,8,160,32,8,power_law_1.2,3.6826942443847654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,128,balanced,0.13796266913414001
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,128,balanced,0.15896532932917276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,128,balanced,0.25269333521525067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,128,balanced,0.20126400391260782
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,128,balanced,0.24499199787775675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,128,balanced,0.3129226764043172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,128,balanced,0.3287786642710368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,128,balanced,0.4148586591084798
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,128,balanced,0.4166453282038371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,128,balanced,0.580618659655253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,128,balanced,0.5241066614786783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,128,balanced,0.0543146679798762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,128,balanced,0.7456426620483398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,128,balanced,0.055973331133524575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,128,balanced,0.05596266686916351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,128,balanced,0.05604266623655955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,128,balanced,0.05967999994754791
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,128,balanced,0.6567840178807577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,128,balanced,0.06451733410358429
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,128,balanced,0.06874666611353557
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,128,balanced,0.07032533486684163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,128,balanced,0.07022400200366974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,128,balanced,0.07028266787528992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,128,balanced,0.07622933387756348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,128,balanced,0.0726453314224879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,128,balanced,0.07054933408896129
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,128,balanced,0.0726986676454544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,128,balanced,0.08221866687138875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,128,balanced,0.0823520024617513
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,128,balanced,0.08646399776140849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,128,balanced,0.916869322458903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,128,balanced,0.09339200456937154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,128,balanced,0.10111467043558757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,128,balanced,0.11362666885058086
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,128,balanced,0.13777066270510355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,128,balanced,0.1657493313153585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,128,balanced,0.9816266695658366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,128,balanced,0.19980265696843466
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,128,balanced,0.25997332731882733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,128,balanced,1.4130880037943523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,128,balanced,0.33153067032496136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,128,power_law_1.01,0.08240000009536744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,128,power_law_1.01,0.08428159952163697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,128,power_law_1.01,0.08150399923324585
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,128,power_law_1.01,0.07749119997024537
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,128,power_law_1.01,0.08161280155181885
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,128,balanced,0.4535839955012004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,128,power_law_1.01,0.08131840229034423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,128,power_law_1.01,0.08972160220146179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,128,power_law_1.01,0.08786560297012329
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,128,power_law_1.01,0.0889087975025177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,128,power_law_1.01,0.08947839736938476
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,128,power_law_1.01,0.08970879912376403
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,128,power_law_1.01,0.0877120018005371
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,128,power_law_1.01,0.08871039748191833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,128,power_law_1.01,0.09317759871482849
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,128,power_law_1.01,0.06053760051727295
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,128,power_law_1.01,0.0942080020904541
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,128,power_law_1.01,0.058796799182891844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,128,power_law_1.01,0.09564160108566284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,128,power_law_1.01,0.05802239775657654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,128,balanced,0.5801920096079508
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,128,power_law_1.01,0.10327680110931396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,128,power_law_1.01,0.05864319801330566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,128,power_law_1.01,0.10794880390167236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,128,power_law_1.01,0.06142079830169678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,128,power_law_1.01,0.11756800413131714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,128,power_law_1.01,0.06287999749183655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,128,power_law_1.01,0.13430399894714357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,128,power_law_1.01,0.06311039924621582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,128,power_law_1.01,0.1504320025444031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,128,power_law_1.01,0.06445440053939819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,128,power_law_1.01,0.06530560255050659
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,128,power_law_1.01,0.18995840549468995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,128,power_law_1.01,0.06824960112571717
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,128,power_law_1.01,0.23243520259857178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,128,power_law_1.01,0.06730239987373351
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,128,power_law_1.01,0.06465920209884643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,128,power_law_1.01,0.06851840019226074
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,128,power_law_1.01,0.314137601852417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,128,balanced,1.907541275024414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,128,power_law_1.01,0.06270080208778381
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,128,power_law_1.01,0.06986240148544312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,128,power_law_1.01,0.4025792121887207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,128,power_law_1.01,0.06294400095939637
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,128,power_law_1.01,0.07099519968032837
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,128,power_law_1.01,0.5531968116760254
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,128,power_law_1.01,0.06398720145225525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,128,power_law_1.01,0.07154560089111328
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,128,power_law_1.01,0.7039167881011963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,128,power_law_1.01,0.0670527994632721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,128,power_law_1.01,0.07567999958992004
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,128,power_law_1.01,0.06693120002746582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,128,power_law_1.01,0.9063679695129394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,128,power_law_1.01,0.08085119724273682
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,128,balanced,0.714021364847819
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,128,power_law_1.01,0.06653439998626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,128,power_law_1.01,0.08539519906044006
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,128,power_law_1.01,1.3168448448181151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,128,power_law_1.01,0.06912000179290771
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,128,power_law_1.01,0.09735680222511292
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,128,power_law_1.01,0.06924160122871399
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,128,power_law_1.01,2.577280044555664
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,128,power_law_1.01,0.11335680484771729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,128,power_law_1.01,0.07189760208129883
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,128,power_law_1.01,0.13815040588378907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,128,power_law_1.01,0.07153919935226441
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,128,power_law_1.01,0.15930880308151246
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,128,power_law_1.01,0.07027199864387512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,2,128,power_law_1.2,0.06478719711303711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,128,power_law_1.01,0.2149440050125122
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,2,128,power_law_1.2,0.0844543993473053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,128,power_law_1.01,0.07159039974212647
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,128,power_law_1.01,0.26438400745391843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,128,power_law_1.01,0.07603840231895446
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,128,balanced,2.7774505615234375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,2,128,power_law_1.2,0.06371840238571166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,2,128,power_law_1.2,0.08558719754219055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,128,power_law_1.01,0.3587968111038208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,128,power_law_1.01,0.07567359805107117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,2,128,power_law_1.2,0.062105602025985716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,2,128,power_law_1.2,0.08001279830932617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,128,power_law_1.01,0.46526079177856444
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,2,128,power_law_1.2,0.07688959836959838
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,128,power_law_1.01,0.6596223831176757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,2,128,power_law_1.2,0.06101760268211365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,128,power_law_1.01,0.08325759768486023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,2,128,power_law_1.2,0.084307199716568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,2,128,power_law_1.2,0.06401919722557067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,128,power_law_1.01,0.8622079849243164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,128,power_law_1.01,0.08887680172920227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,2,128,power_law_1.2,0.0839680016040802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,2,128,power_law_1.2,0.06883199810981751
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,2,128,power_law_1.2,0.0876800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,128,power_law_1.01,0.10255359411239624
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,128,power_law_1.01,1.0705663681030273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,2,128,power_law_1.2,0.0642304003238678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,2,128,power_law_1.2,0.0875711977481842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,128,power_law_1.01,0.12356480360031127
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,128,power_law_1.01,1.712224006652832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,2,128,power_law_1.2,0.06858239769935608
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,2,128,power_law_1.2,0.09004799723625183
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,128,power_law_1.01,0.14147839546203614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,2,128,power_law_1.2,0.07017599940299987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,128,balanced,1.0986773173014324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,128,power_law_1.01,3.3857921600341796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,2,128,power_law_1.2,0.08711680173873901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,128,power_law_1.01,0.1841599941253662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,2,128,power_law_1.2,0.09060479998588562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,2,128,power_law_1.2,0.06877440214157104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,128,power_law_1.01,0.21819519996643066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,2,128,power_law_1.2,0.07061120271682739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,2,128,power_law_1.2,0.09226880073547364
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,128,power_law_1.01,0.29575679302215574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,2,128,power_law_1.2,0.06961920261383056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,128,power_law_1.01,0.3816704034805298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,2,128,power_law_1.2,0.08826239705085755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,128,power_law_1.01,0.5335040092468262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,2,128,power_law_1.2,0.07008000016212464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,2,128,power_law_1.2,0.09067519903182983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,128,power_law_1.01,0.6875072002410889
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,2,128,power_law_1.2,0.07269759774208069
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,2,128,power_law_1.2,0.09662079811096191
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,128,power_law_1.01,0.8444160461425781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,2,128,power_law_1.2,0.07758079767227173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,2,128,power_law_1.2,0.09991040229797363
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,128,power_law_1.01,1.3402175903320312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,2,128,power_law_1.2,0.081004798412323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,2,128,power_law_1.2,0.1003648042678833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,128,power_law_1.01,2.646976089477539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,2,128,power_law_1.2,0.08464000225067139
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,2,128,power_law_1.2,0.10382720232009887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,2,128,power_law_1.2,0.11041280031204223
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,2,128,power_law_1.2,0.1156991958618164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,2,128,power_law_1.2,0.14320000410079955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,2,128,power_law_1.2,0.11905280351638795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,2,128,power_law_1.2,0.17222399711608888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,2,128,power_law_1.2,0.1394368052482605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,2,128,power_law_1.2,0.2251904010772705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,2,128,power_law_1.2,0.28092160224914553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,2,128,power_law_1.2,0.16488319635391235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,2,128,power_law_1.2,0.39647998809814455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,2,128,power_law_1.2,0.21256959438323975
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,2,128,power_law_1.2,0.5049536228179932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,2,128,power_law_1.2,0.24765439033508302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,2,128,power_law_1.2,0.7473279953002929
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,2,128,power_law_1.2,0.35004799365997313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,2,128,power_law_1.2,0.9634880065917969
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,2,128,power_law_1.2,0.42533121109008787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,2,128,power_law_1.2,1.1721920013427733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,2,128,power_law_1.2,0.6773695945739746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,2,128,power_law_1.2,0.8800064086914062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,2,128,power_law_1.2,1.8921920776367187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,2,128,power_law_1.2,1.0519040107727051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,2,128,power_law_1.2,3.845151901245117
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,2,128,power_law_1.2,1.6719551086425781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,2,128,power_law_1.2,3.8793983459472656
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,128,balanced,2.15174929300944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,2,128,power_law_1.2,0.06577280163764954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,2,128,power_law_1.2,0.06488959789276123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,2,128,power_law_1.2,0.06388480067253113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,2,128,power_law_1.2,0.06225280165672302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,2,128,power_law_1.2,0.06289280056953431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,2,128,power_law_1.2,0.0664255976676941
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,2,128,power_law_1.2,0.06595839858055115
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,2,128,power_law_1.2,0.06981120109558106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,2,128,power_law_1.2,0.06908159852027893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,2,128,power_law_1.2,0.06602879762649536
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,2,128,power_law_1.2,0.06970880031585694
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,64,balanced,0.058149332801500954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,2,128,power_law_1.2,0.07184000015258789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,2,128,power_law_1.2,0.07061120271682739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,64,balanced,0.05805333455403646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,64,balanced,0.058431997895240784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,2,128,power_law_1.2,0.07244160175323486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,64,balanced,0.06038933495680491
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,2,128,power_law_1.2,0.07645440101623535
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,64,balanced,0.064410666624705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,64,balanced,0.07600533465544383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,2,128,power_law_1.2,0.07804800271987915
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,64,balanced,0.08731733759244283
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,64,balanced,0.08714133501052856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,2,128,power_law_1.2,0.08250880241394043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,64,balanced,0.08657067020734151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,64,balanced,0.0539680023988088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,64,balanced,0.08653866251309712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,2,128,power_law_1.2,0.09393280148506164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,64,balanced,0.08927999933560689
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,2,128,power_law_1.2,0.10531840324401856
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,64,balanced,0.05394133428732554
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,64,balanced,0.09488000472386678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,2,128,power_law_1.2,0.13160320520401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,64,balanced,0.05418133238951365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,64,balanced,0.09089066584904988
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,2,128,power_law_1.2,0.14840320348739625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,64,balanced,0.055946667989095054
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,64,balanced,0.08921600381533305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,2,128,power_law_1.2,0.19397759437561035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,64,balanced,0.06155733267466227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,64,balanced,0.09613333145777385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,2,128,power_law_1.2,0.23648641109466553
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,64,balanced,0.06341866652170818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,64,balanced,0.09699199597040813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,64,balanced,0.06660800178845723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,2,128,power_law_1.2,0.3212928056716919
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,64,balanced,0.06790400048096974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,64,balanced,0.10136000315348308
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,2,128,power_law_1.2,0.41733760833740235
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,64,balanced,0.06846400101979573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,64,balanced,0.10811733206113179
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,2,128,power_law_1.2,0.6157440185546875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,64,balanced,0.07046933472156525
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,64,balanced,0.07019733389218648
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,64,balanced,0.11624000469843547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,2,128,power_law_1.2,0.7881408214569092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,64,balanced,0.05003733436266581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,64,balanced,0.07254933317502339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,64,balanced,0.07451733450094859
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,2,128,power_law_1.2,0.9824511528015136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,64,balanced,0.052015999952952065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,64,balanced,0.13029332955678305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,64,balanced,0.07525866727034251
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,64,balanced,0.05123200019200643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,64,balanced,0.051957334081331887
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,2,128,power_law_1.2,1.5758015632629394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,64,balanced,0.07865599791208903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,64,balanced,0.05993066728115082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,64,balanced,0.14284267028172812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,64,balanced,0.0806933343410492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,64,balanced,0.062080000837643944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,64,balanced,0.06902933120727539
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,64,balanced,0.08725866675376892
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,64,balanced,0.07162666817506154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,2,128,power_law_1.2,3.09737606048584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,64,balanced,0.09911466638247173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,64,balanced,0.0703306645154953
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,64,balanced,0.17334934075673422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,64,balanced,0.07163733243942261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,64,balanced,0.11136000355084737
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,64,balanced,0.07439466814200084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,64,balanced,0.07237333556016286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,64,balanced,0.13900799552599588
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,64,balanced,0.07268266876538594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,64,balanced,0.20118399461110434
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,64,balanced,0.075013334552447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,64,balanced,0.160261332988739
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,64,balanced,0.07918933530648549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,64,balanced,0.07876800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,64,power_law_1.01,0.08736000061035157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,64,balanced,0.08664000034332275
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,64,balanced,0.20432533820470175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,64,balanced,0.2558133403460185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,64,power_law_1.01,0.08488960266113281
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,64,balanced,0.09193600217501323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,64,power_law_1.01,0.0812287986278534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,64,balanced,0.2473706603050232
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,64,balanced,0.10135466853777568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,64,power_law_1.01,0.07776640057563781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,64,power_law_1.01,0.07720320224761963
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,64,balanced,0.12132799625396729
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,64,power_law_1.01,0.08520960211753845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,64,balanced,0.3158186674118042
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,64,power_law_1.01,0.08042880296707153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,64,balanced,0.13371733824412027
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,64,balanced,0.33155733346939087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,64,power_law_1.01,0.08487039804458618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,64,power_law_1.01,0.0844543993473053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,64,balanced,0.1695093313852946
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,64,power_law_1.01,0.08599680066108703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,64,power_law_1.01,0.08498560190200806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,64,power_law_1.01,0.08818560242652893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,64,power_law_1.01,0.08765439987182617
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,64,balanced,0.4203786849975586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,64,balanced,0.20227199792861938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,64,power_law_1.01,0.08799999952316284
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,64,balanced,0.42239999771118164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,64,power_law_1.01,0.09342079758644103
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,64,power_law_1.01,0.09535359740257263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,64,power_law_1.01,0.10135040283203126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,64,power_law_1.01,0.06813439726829529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,64,power_law_1.01,0.10804480314254761
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,64,balanced,0.26518932978312176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,64,power_law_1.01,0.1173248052597046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,64,power_law_1.01,0.06481919884681701
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,64,power_law_1.01,0.1334720015525818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,64,balanced,0.589845339457194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,64,power_law_1.01,0.062438398599624634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,64,power_law_1.01,0.1513983964920044
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,64,power_law_1.01,0.058873599767684935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,64,power_law_1.01,0.18316160440444945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,64,balanced,0.3397333224614461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,64,power_law_1.01,0.0597055971622467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,64,power_law_1.01,0.21556479930877687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,64,power_law_1.01,0.0661184012889862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,64,balanced,0.533461332321167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,64,power_law_1.01,0.06439679861068726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,64,power_law_1.01,0.2977855920791626
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,64,power_law_1.01,0.067603200674057
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,64,power_law_1.01,0.37319040298461914
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,64,power_law_1.01,0.07391999959945679
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,64,power_law_1.01,0.527558422088623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,4,64,power_law_1.2,0.08735359907150268
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,64,power_law_1.01,0.0676479995250702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,64,power_law_1.01,0.6878208160400391
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,64,power_law_1.01,0.06794880032539367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,4,64,power_law_1.2,0.08465279936790467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,64,balanced,0.46504000822703045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,64,balanced,0.7568106651306152
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,64,power_law_1.01,0.06770560145378113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,64,power_law_1.01,0.8102272033691407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,4,64,power_law_1.2,0.081740802526474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,64,power_law_1.01,0.06908159852027893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,64,power_law_1.01,0.06862720251083373
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,64,power_law_1.01,1.3471743583679199
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,64,power_law_1.01,0.06947839856147767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,4,64,power_law_1.2,0.07432320117950439
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,64,power_law_1.01,0.06036480069160462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,64,balanced,0.6693066755930582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,64,power_law_1.01,2.5848255157470703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,64,power_law_1.01,0.0706816017627716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,4,64,power_law_1.2,0.07640960216522216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,64,power_law_1.01,0.06304640173912049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,64,power_law_1.01,0.07237120270729065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,4,64,power_law_1.2,0.08106880187988282
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,64,power_law_1.01,0.0688704013824463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,64,power_law_1.01,0.07871360182762147
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,4,64,power_law_1.2,0.08121600151062011
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,64,power_law_1.01,0.06814079880714416
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,64,power_law_1.01,0.08174719810485839
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,4,64,power_law_1.2,0.08436480164527893
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,64,power_law_1.01,0.07010560035705567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,64,power_law_1.01,0.08825600147247314
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,4,64,power_law_1.2,0.08522239923477173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,64,power_law_1.01,0.06991360187530518
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,64,balanced,0.5942773421605428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,64,power_law_1.01,0.09994239807128906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,4,64,power_law_1.2,0.08608639836311341
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,64,power_law_1.01,0.0706496000289917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,4,64,power_law_1.2,0.08587520122528076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,64,power_law_1.01,0.11564160585403442
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,64,power_law_1.01,0.07167360186576843
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,64,power_law_1.01,0.14030079841613768
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,4,64,power_law_1.2,0.08837760090827942
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,64,balanced,0.9344639778137207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,64,power_law_1.01,0.07359359860420227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,64,power_law_1.01,0.16550400257110595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,4,64,power_law_1.2,0.08823680281639099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,64,power_law_1.01,0.2129215955734253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,64,power_law_1.01,0.07068799734115601
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,4,64,power_law_1.2,0.09050880074501037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,64,power_law_1.01,0.26440958976745604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,64,power_law_1.01,0.07471359968185425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,4,64,power_law_1.2,0.09693440198898315
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,64,power_law_1.01,0.3631743907928467
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,64,power_law_1.01,0.07875840067863464
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,4,64,power_law_1.2,0.09357439875602722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,4,64,power_law_1.2,0.10081919431686401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,64,power_law_1.01,0.0785152018070221
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,64,power_law_1.01,0.46212477684020997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,4,64,power_law_1.2,0.10928640365600586
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,64,power_law_1.01,0.08620160222053527
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,64,power_law_1.01,0.6655424118041993
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,4,64,power_law_1.2,0.11710079908370971
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,64,power_law_1.01,0.09155200123786926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,64,balanced,0.7342186768849691
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,64,power_law_1.01,0.85098876953125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,4,64,power_law_1.2,0.13899519443511962
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,64,power_law_1.01,0.10153599977493286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,4,64,power_law_1.2,0.1575103998184204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,64,power_law_1.01,0.12511359453201293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,64,balanced,1.0002453327178955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,64,power_law_1.01,1.0547904014587401
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,64,power_law_1.01,0.14612480401992797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,4,64,power_law_1.2,0.19525760412216187
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,64,power_law_1.01,0.18995200395584105
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,4,64,power_law_1.2,0.2426687955856323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,64,power_law_1.01,1.693779182434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,64,power_law_1.01,0.22184960842132567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,4,64,power_law_1.2,0.3155711889266968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,64,power_law_1.01,3.3246593475341797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,64,power_law_1.01,0.3036288022994995
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,4,64,power_law_1.2,0.40974721908569334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,64,power_law_1.01,0.38204801082611084
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,4,64,power_law_1.2,0.578931188583374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,64,power_law_1.01,0.5308800220489502
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,4,64,power_law_1.2,0.8270784378051758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,64,power_law_1.01,0.6959551811218262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,4,64,power_law_1.2,0.9469056129455566
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,64,balanced,1.4394292831420898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,64,power_law_1.01,0.8500991821289062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,4,64,power_law_1.2,1.550387191772461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,64,power_law_1.01,1.3669376373291016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,64,power_law_1.2,3.4672382354736326
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,64,power_law_1.01,2.6811647415161133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,64,balanced,1.126032034556071
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,4,64,power_law_1.2,0.06780160069465638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,4,64,power_law_1.2,0.06502400040626526
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,4,64,power_law_1.2,0.062438398599624634
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,4,64,balanced,1.9540640513102214
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,4,64,power_law_1.2,0.05902079939842224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,4,64,power_law_1.2,0.06204800009727478
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,4,64,power_law_1.2,0.0637440025806427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,4,64,power_law_1.2,0.06488320231437683
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,4,64,power_law_1.2,0.06593919992446899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,64,balanced,2.836810747782389
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,4,64,power_law_1.2,0.067084801197052
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,4,64,power_law_1.2,0.0668287992477417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,4,64,power_law_1.2,0.06788480281829834
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,4,64,power_law_1.2,0.07047039866447449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,4,64,power_law_1.2,0.06903679966926575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,64,balanced,2.2169920603434243
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,4,64,power_law_1.2,0.07296640276908875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,4,64,power_law_1.2,0.07850880026817322
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,4,64,power_law_1.2,0.08225280046463013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,4,64,power_law_1.2,0.08644480109214783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,4,64,power_law_1.2,0.10424319505691529
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,4,64,power_law_1.2,0.11749759912490845
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,4,64,power_law_1.2,0.1425536036491394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,4,64,power_law_1.2,0.16953599452972412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,4,64,power_law_1.2,0.22558720111846925
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,4,64,power_law_1.2,0.2731519937515259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,4,64,power_law_1.2,0.37664639949798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,4,64,power_law_1.2,0.4924352169036865
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,4,64,power_law_1.2,0.6982783794403076
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,4,64,power_law_1.2,0.935206413269043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,4,64,power_law_1.2,1.1854207992553711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,4,64,power_law_1.2,1.8719039916992188
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,4,64,power_law_1.2,3.7749439239501954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,4,64,power_law_1.2,0.07482240200042725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,4,64,power_law_1.2,0.06718720197677612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,4,64,power_law_1.2,0.06911360025405884
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,4,64,power_law_1.2,0.05921279788017273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,4,64,power_law_1.2,0.05983999967575073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,4,64,power_law_1.2,0.06496000289916992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,4,64,power_law_1.2,0.06743040084838867
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,4,64,power_law_1.2,0.0686847984790802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,4,64,power_law_1.2,0.06821759939193725
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,4,64,power_law_1.2,0.06986240148544312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,4,64,power_law_1.2,0.07111039757728577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,4,64,power_law_1.2,0.07114239931106567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,4,64,power_law_1.2,0.07063680291175842
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,4,64,power_law_1.2,0.0736191987991333
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,4,64,power_law_1.2,0.07757440209388733
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,4,64,power_law_1.2,0.08000640273094177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,4,64,power_law_1.2,0.08357120156288148
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,4,64,power_law_1.2,0.0944703996181488
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,4,64,power_law_1.2,0.10718719959259033
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,4,64,power_law_1.2,0.1336127996444702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,4,64,power_law_1.2,0.14947839975357055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,4,64,power_law_1.2,0.19667199850082398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,4,64,power_law_1.2,0.22892799377441406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,32,balanced,0.05974400043487549
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,4,64,power_law_1.2,0.3162175893783569
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,4,64,power_law_1.2,0.40860800743103026
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,32,balanced,0.0606879989306132
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,32,balanced,0.060047999024391174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,4,64,power_law_1.2,0.615180778503418
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,32,balanced,0.06412266691525777
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,32,balanced,0.06653333206971486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,4,64,power_law_1.2,0.7445375919342041
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,32,balanced,0.080485333998998
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,4,64,power_law_1.2,0.9500991821289062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,32,balanced,0.09136000275611877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,32,balanced,0.090938667456309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,4,64,power_law_1.2,1.5640704154968261
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,32,balanced,0.09139733513196309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,32,balanced,0.09106133381525676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,32,balanced,0.09290666381518047
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,4,64,power_law_1.2,3.1459072113037108
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,32,balanced,0.09498666723569234
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,32,balanced,0.09355733791987102
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,32,balanced,0.09174399574597676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,32,balanced,0.10057600339253743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,32,balanced,0.09949333469072978
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,32,balanced,0.05021866659323374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,32,balanced,0.10520533720652263
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,32,balanced,0.052144000927607216
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,32,balanced,0.11386133233706157
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,32,balanced,0.05175999800364176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,32,balanced,0.05204799771308899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,32,balanced,0.05570133527119955
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,32,balanced,0.11941867073376973
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,32,balanced,0.060346667965253196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,32,balanced,0.05186666548252106
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,32,balanced,0.05220800141493479
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,32,balanced,0.06427200138568878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,32,balanced,0.05733866492907206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,32,balanced,0.134661336739858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,32,balanced,0.062234664956728615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,32,balanced,0.0705973356962204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,32,balanced,0.06762133538722992
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,32,balanced,0.0714026689529419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,32,balanced,0.07277866701285045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,32,balanced,0.14882133404413858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,32,balanced,0.07681599756081899
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,32,balanced,0.0707946668068568
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,32,balanced,0.07715199887752533
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,32,balanced,0.0706879993279775
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,32,balanced,0.0745119998852412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,32,balanced,0.07237333556016286
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,32,balanced,0.0765066643555959
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,32,balanced,0.07190399865309398
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,32,balanced,0.18061333894729614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,32,balanced,0.07876800000667572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,32,balanced,0.07246933380762736
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,32,balanced,0.0767680009206136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,32,balanced,0.07642666498819987
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,32,balanced,0.08115733166535695
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,32,balanced,0.08077333370844524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,32,balanced,0.08286400139331818
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,32,balanced,0.20871466398239136
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,32,balanced,0.08493333061536153
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,32,balanced,0.08542933066685994
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,32,balanced,0.0909493366877238
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,32,balanced,0.09457066655158997
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,32,balanced,0.10377599795659383
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,32,balanced,0.09911466638247173
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,32,balanced,0.11560533444086711
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,32,balanced,0.10849600036938985
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,32,balanced,0.2646613319714864
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,32,balanced,0.14297599593798319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,32,balanced,0.12735999623934427
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,32,balanced,0.1655786633491516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,32,balanced,0.1442346672217051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,32,balanced,0.3323040008544922
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,32,balanced,0.2102186679840088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,32,balanced,0.18086934089660645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,32,power_law_1.01,0.06892160177230836
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,32,power_law_1.01,0.08076159954071045
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,32,power_law_1.01,0.08977280259132385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,32,power_law_1.01,0.07218559980392455
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,32,power_law_1.01,0.08652160167694092
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,32,power_law_1.01,0.06856319904327393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,32,balanced,0.21433067321777344
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,32,balanced,0.2547733386357625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,32,power_law_1.01,0.07025920152664185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,32,power_law_1.01,0.05983999967575073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,32,power_law_1.01,0.07432960271835327
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,32,power_law_1.01,0.061990398168563846
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,32,power_law_1.01,0.07786880135536194
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,32,power_law_1.01,0.06584320068359376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,32,power_law_1.01,0.07994239926338195
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,32,balanced,0.4456373453140259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,32,power_law_1.01,0.08336640000343323
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,32,power_law_1.01,0.06541439890861511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,32,power_law_1.01,0.08425599932670594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,32,power_law_1.01,0.06902400255203248
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,32,balanced,0.2794026732444763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,32,balanced,0.33949867884318036
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,32,power_law_1.01,0.08543999791145325
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,32,power_law_1.01,0.07148159742355346
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,32,power_law_1.01,0.08815360069274902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,32,power_law_1.01,0.0732159972190857
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,32,power_law_1.01,0.08863999843597412
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,32,power_law_1.01,0.07312639951705932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,32,power_law_1.01,0.06914560198783874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,32,power_law_1.01,0.0907584011554718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,32,power_law_1.01,0.07381759881973267
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,32,power_law_1.01,0.0918720006942749
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,32,power_law_1.01,0.07475200295448303
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,32,power_law_1.01,0.07429119944572449
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,32,power_law_1.01,0.0976639986038208
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,32,power_law_1.01,0.07659519910812378
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,32,power_law_1.01,0.07081599831581116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,32,balanced,0.3612906535466512
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,32,balanced,0.43317333857218426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,32,power_law_1.01,0.08330240249633789
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,32,power_law_1.01,0.09726719856262207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,32,power_law_1.01,0.0574783980846405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,32,power_law_1.01,0.0846015989780426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,32,balanced,0.5630346536636353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,32,power_law_1.01,0.10271999835968018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,32,power_law_1.01,0.0926912009716034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,32,power_law_1.01,0.06045439839363098
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,32,power_law_1.01,0.10990719795227051
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,32,power_law_1.01,0.10408960580825806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,32,power_law_1.01,0.066348797082901
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,32,power_law_1.01,0.11832959651947021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,32,power_law_1.01,0.12138880491256714
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,32,power_law_1.01,0.0683456003665924
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,32,power_law_1.01,0.13852159976959227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,32,power_law_1.01,0.14865920543670655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,32,power_law_1.01,0.06937599778175355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,32,power_law_1.01,0.17488640546798706
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,32,power_law_1.01,0.15551359653472902
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,32,balanced,0.606277346611023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,32,balanced,0.4919999837875366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,32,power_law_1.01,0.07272319793701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,32,power_law_1.01,0.223142409324646
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,32,power_law_1.01,0.192684805393219
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,32,power_law_1.01,0.0713919997215271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,32,power_law_1.01,0.2734463930130005
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,32,power_law_1.01,0.22042241096496581
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,32,power_law_1.01,0.07434239983558655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,32,power_law_1.01,0.38035199642181394
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,32,power_law_1.01,0.30140159130096433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,32,power_law_1.01,0.0748416006565094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,32,power_law_1.01,0.48555521965026854
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,32,power_law_1.01,0.37664639949798584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,32,power_law_1.01,0.07564160227775574
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,32,power_law_1.01,0.699455976486206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,32,power_law_1.01,0.07588480114936828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,32,power_law_1.01,0.5119359970092774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,32,power_law_1.01,0.870304012298584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,32,power_law_1.01,0.08414720296859741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,32,power_law_1.01,0.6977983951568604
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,32,power_law_1.01,0.0834559977054596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,32,power_law_1.01,1.1049280166625977
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,32,balanced,0.7047466437021891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,32,power_law_1.01,0.8335103988647461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,32,power_law_1.01,0.0904640018939972
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,32,power_law_1.01,1.7013696670532226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,32,power_law_1.01,0.09963520169258118
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,32,power_law_1.01,1.3315072059631348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,32,power_law_1.01,0.11011840105056762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,32,balanced,0.7824479738871256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,32,power_law_1.01,3.3856510162353515
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,32,balanced,0.6322400172551473
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,32,power_law_1.01,2.6073984146118163
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,32,power_law_1.01,0.13173760175704957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,32,power_law_1.01,0.1520640015602112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,32,power_law_1.01,0.19251199960708618
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,32,power_law_1.01,0.23505918979644774
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,8,32,power_law_1.2,0.08217599987983704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,32,power_law_1.01,0.31629440784454343
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,8,32,power_law_1.2,0.08983039855957031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,32,power_law_1.01,0.3972223997116089
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,8,32,power_law_1.2,0.08217599987983704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,32,power_law_1.01,0.5667263984680175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,8,32,power_law_1.2,0.06948480010032654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,32,power_law_1.01,0.7341824054718018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,8,32,power_law_1.2,0.07598080039024353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,32,power_law_1.01,0.8911808013916016
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,8,32,power_law_1.2,0.07961599826812744
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,8,32,power_law_1.2,0.07870720028877258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,32,power_law_1.01,1.4405247688293457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,8,32,power_law_1.2,0.08503040075302123
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,32,balanced,0.7756746610005697
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,32,balanced,0.9689973195393881
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,8,32,power_law_1.2,0.08588799834251404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,32,power_law_1.01,2.8376768112182615
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,8,32,power_law_1.2,0.08439679741859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,32,balanced,1.0551520188649495
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,8,32,power_law_1.2,0.08691200017929077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,8,32,power_law_1.2,0.08842880129814149
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,8,32,power_law_1.2,0.08891519904136658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,8,32,power_law_1.2,0.09018880128860474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,8,32,power_law_1.2,0.09852160215377807
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,8,32,power_law_1.2,0.09942399859428405
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,8,32,power_law_1.2,0.10101759433746338
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,8,32,power_law_1.2,0.11203839778900146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,8,32,power_law_1.2,0.11920640468597413
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,8,32,power_law_1.2,0.14207359552383422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,8,32,power_law_1.2,0.16099200248718262
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,8,32,power_law_1.2,0.20241920948028563
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,8,32,power_law_1.2,0.23188478946685792
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,8,32,power_law_1.2,0.3498816013336182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,8,32,power_law_1.2,0.41660161018371583
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,8,32,power_law_1.2,0.5899007797241211
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,8,32,power_law_1.2,0.06652799844741822
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,8,32,power_law_1.2,0.7255167961120605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,32,balanced,1.1982666651407878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,32,balanced,1.4999359448750813
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,8,32,power_law_1.2,0.06890239715576171
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,8,32,power_law_1.2,0.06211199760437012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,8,32,power_law_1.2,0.920736026763916
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,8,32,power_law_1.2,0.05743359923362732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,8,32,power_law_1.2,0.058796799182891844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,8,32,power_law_1.2,1.4192831993103028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,8,32,power_law_1.2,0.06261119842529297
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,8,32,power_law_1.2,0.06576640009880066
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,32,power_law_1.2,2.9302080154418944
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,8,32,power_law_1.2,0.06769919991493226
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,8,32,power_law_1.2,0.06839039921760559
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,8,32,power_law_1.2,0.06991999745368957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,8,32,power_law_1.2,0.07090560197830201
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,8,32,power_law_1.2,0.07281280159950257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,8,32,power_law_1.2,0.07344639897346497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,8,32,power_law_1.2,0.0760703980922699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,8,32,power_law_1.2,0.08212479948997498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,8,32,power_law_1.2,0.08414720296859741
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,8,32,power_law_1.2,0.091430401802063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,8,32,power_law_1.2,0.10918400287628174
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,8,32,power_law_1.2,0.12058240175247192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,8,32,balanced,2.0675786336263022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,8,32,power_law_1.2,0.15229439735412598
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,8,32,power_law_1.2,0.17571200132369996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,8,32,power_law_1.2,0.22873599529266359
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,8,32,power_law_1.2,0.27740800380706787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,8,32,power_law_1.2,0.39163520336151125
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,8,32,power_law_1.2,0.5120960235595703
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,8,32,power_law_1.2,0.7216767787933349
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,8,32,power_law_1.2,0.9660991668701172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,8,32,power_law_1.2,1.196723175048828
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,8,32,power_law_1.2,1.9021696090698241
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,32,power_law_1.2,3.7227264404296876
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,32,balanced,2.3531626065572104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,8,32,balanced,2.9537814458211265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,8,32,power_law_1.2,0.06947199702262878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,8,32,power_law_1.2,0.07489280104637146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,8,32,power_law_1.2,0.06291840076446534
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,8,32,power_law_1.2,0.05804160237312317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,8,32,power_law_1.2,0.05983359813690185
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,8,32,power_law_1.2,0.06533120274543762
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,8,32,power_law_1.2,0.06319360136985779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,8,32,power_law_1.2,0.07066879868507385
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,8,32,power_law_1.2,0.0699455976486206
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,8,32,power_law_1.2,0.06869120001792908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,8,32,power_law_1.2,0.07244160175323486
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,8,32,power_law_1.2,0.0717248022556305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,8,32,power_law_1.2,0.07620480060577392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,8,32,power_law_1.2,0.07518720030784606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,8,32,power_law_1.2,0.0834879994392395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,8,32,power_law_1.2,0.0846015989780426
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,8,32,power_law_1.2,0.09077759981155395
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,8,32,power_law_1.2,0.10238720178604126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,8,32,power_law_1.2,0.11003520488739013
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,8,32,power_law_1.2,0.13578879833221436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,8,32,power_law_1.2,0.16145919561386107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,8,32,power_law_1.2,0.20330240726470947
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,8,32,power_law_1.2,0.2457535982131958
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,8,32,power_law_1.2,0.3414144039154053
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,8,32,power_law_1.2,0.42658557891845705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,8,32,power_law_1.2,0.6158207893371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,8,32,power_law_1.2,0.821446418762207
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,8,32,power_law_1.2,1.032915210723877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,8,32,power_law_1.2,1.6939968109130858
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,8,32,power_law_1.2,3.292512130737305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,16,balanced,0.05385066568851471
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,16,balanced,0.05440000196297964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,16,power_law_1.01,0.054079997539520266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,16,balanced,0.053743998209635414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,16,balanced,0.05797866483529409
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,16,balanced,0.06018133461475372
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,16,power_law_1.01,0.0644864022731781
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,16,balanced,0.06614399949709575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,16,balanced,0.07212799787521362
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,16,power_law_1.01,0.07401599884033203
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,16,balanced,0.07250133156776428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,16,power_law_1.01,0.053907197713851926
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,16,balanced,0.07250133156776428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,16,balanced,0.072543998559316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,16,power_law_1.01,0.057183998823165896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,16,balanced,0.07468266785144806
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,16,power_law_1.01,0.06234880089759827
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,16,balanced,0.07530133426189423
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,16,power_law_1.01,0.06287360191345215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,16,balanced,0.07660266757011414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,16,balanced,0.0783733328183492
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,16,power_law_1.01,0.06584320068359376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,16,balanced,0.08277866741021474
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,16,power_law_1.01,0.06670079827308655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,16,balanced,0.08284266789754231
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,16,power_law_1.01,0.06843519806861878
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,16,balanced,0.08970133463541667
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,16,power_law_1.01,0.06951040029525757
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,16,balanced,0.09813867012659709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,16,power_law_1.01,0.07156479954719544
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,16,power_law_1.01,0.07307519912719726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,16,balanced,0.10565333565076192
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,16,power_law_1.01,0.07431039810180665
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,16,power_law_1.01,0.08010240197181702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,16,balanced,0.12930132945378622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,16,power_law_1.01,0.08213760256767273
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,16,power_law_1.01,0.08911359906196595
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,16,balanced,0.14672533671061197
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,16,power_law_1.01,0.10513919591903687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,16,power_law_1.01,0.11780480146408082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,16,balanced,0.18353599309921265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,16,power_law_1.01,0.13904000520706178
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,16,power_law_1.01,0.1714560031890869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,16,power_law_1.01,0.22140800952911377
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,16,balanced,0.2246453364690145
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,16,power_law_1.01,0.25729920864105227
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,16,power_law_1.01,0.3494271993637085
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,16,power_law_1.01,0.4393919944763184
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,16,power_law_1.01,0.6218688011169433
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,16,balanced,0.28862400849660236
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,16,power_law_1.01,0.8180416107177735
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,16,power_law_1.01,1.006668758392334
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,16,power_law_1.01,1.4946751594543457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,16,power_law_1.01,2.96563835144043
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,16,balanced,0.3742773135503133
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,16,balanced,0.5067466497421265
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,16,balanced,0.6562186479568481
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,16,balanced,0.8092959721883138
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,16,balanced,0.05598933498064677
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,16,balanced,0.04837333162625631
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,16,balanced,0.0479360024134318
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,16,balanced,0.0496373325586319
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,16,balanced,0.05182399849096934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,16,balanced,0.062037333846092224
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,16,balanced,0.07228266696135204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,16,balanced,0.07638933261235555
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,16,balanced,0.07526933153470357
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,16,balanced,0.07592533528804779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,16,balanced,0.07660266757011414
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,16,balanced,0.08035199840863545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,16,balanced,0.07899733384450276
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,16,balanced,0.08113066852092743
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,16,balanced,0.08619200189908345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,16,balanced,0.08692266543706258
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,16,balanced,0.09288533528645833
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,16,balanced,0.043653334180514015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,16,balanced,0.10131733616193135
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,16,balanced,0.04393066465854645
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,16,balanced,0.04589866598447164
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,16,balanced,0.1067573328812917
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,16,balanced,0.04595733185609182
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,16,balanced,0.049695998430252075
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,16,balanced,0.12204800049463908
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,16,balanced,0.05455466608206431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,16,balanced,0.05924266576766968
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,16,balanced,1.2450026671091716
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,16,balanced,0.059845333298047386
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,16,balanced,0.13649066289265951
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,16,balanced,0.060362666845321655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,16,balanced,0.06191466748714447
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,16,balanced,0.062362665931383766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,16,balanced,0.16262400150299072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,16,balanced,0.06437333424886067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,16,balanced,0.067071999112765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,16,balanced,0.06613866488138835
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,16,balanced,0.07258666555086772
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,16,balanced,0.20708266894022623
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,16,balanced,0.07464533547560374
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,16,balanced,0.0827893316745758
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,16,balanced,0.0957973301410675
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,16,balanced,0.11111467083295186
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,16,balanced,0.26639999945958454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,16,balanced,0.13501333196957907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,16,balanced,0.15636799732844034
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,16,balanced,0.1995519995689392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,16,balanced,0.3309920032819112
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,16,balanced,0.2552586595217387
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,16,balanced,0.3410559892654419
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,16,balanced,0.4647626479466756
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,16,balanced,0.4323466618855794
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,16,power_law_1.01,0.05936639904975891
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,16,power_law_1.01,0.07329279780387879
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,16,power_law_1.01,0.06938239932060242
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,16,power_law_1.01,0.05180799961090088
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,16,power_law_1.01,0.05444480180740356
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,16,balanced,0.577733318010966
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,16,power_law_1.01,0.06074240207672119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,16,power_law_1.01,0.061939197778701785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,16,balanced,0.6186399857203165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,16,power_law_1.01,0.0663424015045166
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,16,balanced,2.4371360143025718
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,16,power_law_1.01,0.06637439727783204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,16,power_law_1.01,0.0695039987564087
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,16,power_law_1.01,0.07007359862327575
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,16,power_law_1.01,0.07206400036811829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,16,power_law_1.01,0.07206400036811829
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,16,power_law_1.01,0.07683200240135193
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,16,power_law_1.01,0.08366720080375671
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,16,power_law_1.01,0.08570880293846131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,16,power_law_1.01,0.095551997423172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,16,balanced,0.794981320699056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,16,power_law_1.01,0.10368000268936158
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,16,balanced,0.7171253363291422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,16,power_law_1.01,0.10581120252609252
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1,7168,2048,8,384,16,16,power_law_1.2,0.05400320291519165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,16,power_law_1.01,0.12906240224838256
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,16,power_law_1.01,0.15483520030975342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2,7168,2048,8,384,16,16,power_law_1.2,0.0644927978515625
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,16,16,power_law_1.2,0.05779839754104614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,16,power_law_1.01,0.1895359992980957
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4,7168,2048,8,384,16,16,power_law_1.2,0.06295040249824524
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,16,16,power_law_1.2,0.05550720095634461
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,16,power_law_1.01,0.24058239459991454
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8,7168,2048,8,384,16,16,power_law_1.2,0.05441280007362366
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,16,16,power_law_1.2,0.059724801778793336
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,16,power_law_1.01,0.3170047998428345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16,7168,2048,8,384,16,16,power_law_1.2,0.05624960064888
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,16,16,power_law_1.2,0.048979198932647704
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,16,power_law_1.01,0.3929856061935425
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32,7168,2048,8,384,16,16,power_law_1.2,0.06030719876289368
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,16,power_law_1.01,0.04979200065135956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,16,16,power_law_1.2,0.053932797908782956
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,48,7168,2048,8,384,16,16,power_law_1.2,0.06287360191345215
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,16,power_law_1.01,0.5652544021606445
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,16,16,power_law_1.2,0.057004797458648684
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,64,7168,2048,8,384,16,16,power_law_1.2,0.0662335991859436
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,16,power_law_1.01,0.05948160290718078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,16,balanced,0.9880692958831787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,16,power_law_1.01,0.7011775970458984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,16,16,power_law_1.2,0.058745598793029784
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,80,7168,2048,8,384,16,16,power_law_1.2,0.06657919883728028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,16,power_law_1.01,0.06386560201644897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,16,16,power_law_1.2,0.06351360082626342
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,16,power_law_1.01,0.8577088356018067
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,96,7168,2048,8,384,16,16,power_law_1.2,0.06734079718589783
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,16,power_law_1.01,0.04815999865531921
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,16,16,power_law_1.2,0.06333439946174621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,128,7168,2048,8,384,16,16,power_law_1.2,0.0690496027469635
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,16,power_law_1.01,0.049055999517440795
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,16,power_law_1.01,1.2745535850524903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,16,16,power_law_1.2,0.06408960223197938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,160,7168,2048,8,384,16,16,power_law_1.2,0.07226240038871765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,16,power_law_1.01,0.05371519923210144
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,16,16,power_law_1.2,0.0661184012889862
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,192,7168,2048,8,384,16,16,power_law_1.2,0.07228800058364868
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,16,power_law_1.01,0.054118400812149046
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,16,power_law_1.01,2.6581823348999025
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,16,16,power_law_1.2,0.07114239931106567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,16,power_law_1.01,0.05925760269165039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,256,7168,2048,8,384,16,16,power_law_1.2,0.07420160174369812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,16,16,power_law_1.2,0.07095680236816407
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,16,16,power_law_1.2,0.07416960000991821
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,16,power_law_1.01,0.05932160019874573
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,320,7168,2048,8,384,16,16,power_law_1.2,0.08048639893531799
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,16,balanced,1.0719359715779622
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,16,16,power_law_1.2,0.08179200291633607
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,16,power_law_1.01,0.05989760160446167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,384,7168,2048,8,384,16,16,power_law_1.2,0.08309119939804077
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,16,16,power_law_1.2,0.08503680229187012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,512,7168,2048,8,384,16,16,power_law_1.2,0.09301120042800903
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,16,power_law_1.01,0.06238719820976257
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,16,16,power_law_1.2,0.09232640266418457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,16,power_law_1.01,0.06472960114479065
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,768,7168,2048,8,384,16,16,power_law_1.2,0.10662399530410767
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,16,16,power_law_1.2,0.10618879795074462
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,16,power_law_1.01,0.06663680076599121
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1024,7168,2048,8,384,16,16,power_law_1.2,0.11845120191574096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,16,16,power_law_1.2,0.11594239473342896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,16,power_law_1.01,0.06966400146484375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,16,16,power_law_1.2,0.13666559457778932
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,1536,7168,2048,8,384,16,16,power_law_1.2,0.1455615997314453
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,16,power_law_1.01,0.0769536018371582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,16,16,power_law_1.2,0.1598912000656128
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,2048,7168,2048,8,384,16,16,power_law_1.2,0.17496320009231567
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,16,balanced,1.5160853068033855
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,16,power_law_1.01,0.08183040022850037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,16,16,power_law_1.2,0.2005824089050293
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,3072,7168,2048,8,384,16,16,power_law_1.2,0.22731521129608154
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,16,power_law_1.01,0.09017599821090698
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,16,16,power_law_1.2,0.25223679542541505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,4096,7168,2048,8,384,16,16,power_law_1.2,0.26604158878326417
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,16,power_law_1.01,0.10858880281448365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,16,16,power_law_1.2,0.34444799423217776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,6144,7168,2048,8,384,16,16,power_law_1.2,0.3638144016265869
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,16,power_law_1.01,0.11727999448776245
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,16,16,power_law_1.2,0.42412800788879396
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,8192,7168,2048,8,384,16,16,power_law_1.2,0.4796735763549805
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,16,power_law_1.01,0.1487104058265686
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,16,16,power_law_1.2,0.6095680236816406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,12288,7168,2048,8,384,16,16,power_law_1.2,0.6871615886688233
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,16,power_law_1.01,0.17368320226669312
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,16384,7168,2048,8,384,16,16,power_law_1.2,0.8539456367492676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,16,16,power_law_1.2,0.871513557434082
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,16,power_law_1.01,0.2199552059173584
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,20480,7168,2048,8,384,16,16,power_law_1.2,1.0910079956054688
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,16,16,power_law_1.2,0.9477760314941406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,16,power_law_1.01,0.2869247913360596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,16,power_law_1.01,0.3949824094772339
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,16,16,power_law_1.2,1.4668160438537599
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,32768,7168,2048,8,384,16,16,power_law_1.2,1.6681600570678712
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,16,power_law_1.01,0.5018752098083497
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,16,power_law_1.2,3.482175827026367
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,16,power_law_1.01,0.7192575931549072
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,nvfp4,65536,7168,2048,8,384,16,16,power_law_1.2,3.3712383270263673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,16,power_law_1.01,0.9265151977539062
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,16,power_law_1.01,1.1251328468322754
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,16,power_law_1.01,1.7847423553466797
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,16,power_law_1.01,3.529497528076172
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,16,16,balanced,2.087381362915039
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,16,balanced,2.988346735636393
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,16,16,power_law_1.2,0.049030399322509764
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,16,16,power_law_1.2,0.057715201377868654
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,16,16,power_law_1.2,0.050944000482559204
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,16,16,power_law_1.2,0.04829440116882324
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,16,16,power_law_1.2,0.04843519926071167
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,16,16,power_law_1.2,0.05189120173454285
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,16,16,power_law_1.2,0.053504002094268796
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,16,16,power_law_1.2,0.05622400045394897
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,16,16,power_law_1.2,0.05872640013694763
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,16,16,power_law_1.2,0.057574397325515746
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,16,16,power_law_1.2,0.06113280057907104
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,16,16,power_law_1.2,0.06454399824142457
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,16,16,power_law_1.2,0.06686080098152161
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,16,16,power_law_1.2,0.06918399930000305
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,16,16,power_law_1.2,0.0762880027294159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,16,16,power_law_1.2,0.0840831995010376
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,16,16,power_law_1.2,0.0910975992679596
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,16,16,power_law_1.2,0.11153279542922974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,16,16,power_law_1.2,0.12132480144500732
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,16,16,power_law_1.2,0.1530943989753723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,16,16,power_law_1.2,0.1798848032951355
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,16,16,power_law_1.2,0.23073279857635498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,16,16,power_law_1.2,0.3011264085769653
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,16,16,power_law_1.2,0.4264768123626709
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,16,16,power_law_1.2,0.5269760131835938
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,16,16,power_law_1.2,0.8403264045715332
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,16,16,power_law_1.2,0.978816032409668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,16,16,power_law_1.2,1.2838272094726562
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,16,16,power_law_1.2,1.9491199493408202
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,16,16,power_law_1.2,4.075711822509765
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,8,balanced,0.04980266590913137
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,8,balanced,0.04762666424115499
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,8,balanced,0.0498879998922348
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,8,balanced,0.049914668003718056
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,8,balanced,0.05579733351866404
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,8,balanced,0.07042133311430614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,8,balanced,0.08272000153859456
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,8,balanced,0.08543466528256734
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,8,balanced,0.08443733056386311
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,8,balanced,0.08685866991678874
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,8,balanced,0.08646933237711589
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,8,balanced,0.0890826682249705
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,8,balanced,0.0890880028406779
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,8,balanced,0.0918293297290802
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,8,balanced,0.09633066256841023
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,8,balanced,0.09983999530474345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,8,balanced,0.10450133681297302
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,8,balanced,0.11482666929562886
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,8,balanced,0.12473066647847493
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,8,balanced,0.14205333590507507
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,8,balanced,0.1592693328857422
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,8,balanced,0.19377599159876505
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,8,balanced,0.24888533353805542
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,8,balanced,0.3196000059445699
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,8,balanced,0.4151253302892049
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,8,balanced,0.5899306535720825
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,8,balanced,0.043653334180514015
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,8,balanced,0.04599999884764353
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,8,balanced,0.0480373352766037
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,8,balanced,0.04975999891757965
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,8,balanced,0.05417066812515259
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,8,balanced,0.06405866642793019
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,8,balanced,0.0708426684141159
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,8,balanced,0.07192533214886983
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,8,balanced,0.07248533268769582
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,8,balanced,0.07436266541481018
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,8,balanced,0.07459199925263722
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,8,balanced,0.07666133344173431
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,8,balanced,0.07842666904131572
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,8,balanced,0.7395466963450114
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,8,balanced,0.08086400230725606
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,8,balanced,0.08805333574612935
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,8,balanced,0.09059199690818787
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,8,power_law_1.01,0.05170559883117676
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,8,balanced,0.09704533219337463
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,8,power_law_1.01,0.05080320239067078
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,8,balanced,0.11620799700419109
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,8,power_law_1.01,0.06017919778823853
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,8,balanced,0.13074666261672974
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,8,power_law_1.01,0.05018240213394165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,8,power_law_1.01,0.05597440004348755
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,8,balanced,0.15686399737993875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,8,power_law_1.01,0.06156799793243408
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,8,power_law_1.01,0.06462079882621766
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,8,balanced,0.18069332838058472
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,8,power_law_1.01,0.07245439887046815
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,8,power_law_1.01,0.07185279726982116
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,8,power_law_1.01,0.07368959784507752
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,8,balanced,0.22817599773406982
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,8,balanced,0.9321706295013428
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,8,power_law_1.01,0.07736319899559022
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,8,power_law_1.01,0.08253440260887146
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1,7168,2048,8,384,32,8,power_law_1.2,0.04838399887084961
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,8,power_law_1.01,0.07978879809379577
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,8,balanced,0.2977866729100545
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,8,power_law_1.01,0.08576639890670776
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2,7168,2048,8,384,32,8,power_law_1.2,0.052095997333526614
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4,7168,2048,8,384,32,8,power_law_1.2,0.05065600275993347
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,8,power_law_1.01,0.09208959937095643
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8,7168,2048,8,384,32,8,power_law_1.2,0.04809600114822388
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,8,power_law_1.01,0.09493759870529175
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16,7168,2048,8,384,32,8,power_law_1.2,0.05451520085334778
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,8,power_law_1.01,0.10494719743728638
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,8,power_law_1.01,0.04717440009117126
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,8,balanced,0.39740268389383954
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32,7168,2048,8,384,32,8,power_law_1.2,0.06122879981994629
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,8,power_law_1.01,0.11747200489044189
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,48,7168,2048,8,384,32,8,power_law_1.2,0.06287999749183655
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,8,power_law_1.01,0.12652159929275514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,64,7168,2048,8,384,32,8,power_law_1.2,0.06863999962806702
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,8,power_law_1.01,0.15343999862670898
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,8,power_law_1.01,0.05318400263786316
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,80,7168,2048,8,384,32,8,power_law_1.2,0.07175679802894593
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,8,power_law_1.01,0.1782464027404785
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,8,power_law_1.01,0.06123520135879516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,96,7168,2048,8,384,32,8,power_law_1.2,0.07132800221443177
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,8,power_law_1.01,0.23432960510253906
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,8,power_law_1.01,0.0493120014667511
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,128,7168,2048,8,384,32,8,power_law_1.2,0.07466880083084107
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,8,balanced,0.5199093421300253
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,8,power_law_1.01,0.05241600275039673
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,160,7168,2048,8,384,32,8,power_law_1.2,0.0790336012840271
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,8,power_law_1.01,0.29049599170684814
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,8,power_law_1.01,0.0584447979927063
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,192,7168,2048,8,384,32,8,power_law_1.2,0.07934079766273498
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,8,power_law_1.01,0.3831423997879028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1,7168,2048,8,384,32,8,power_law_1.2,0.04855040013790131
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,256,7168,2048,8,384,32,8,power_law_1.2,0.0818880021572113
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,8,power_law_1.01,0.05983999967575073
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,8,power_law_1.01,0.4928575992584229
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,8,power_law_1.01,0.06656000018119812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,320,7168,2048,8,384,32,8,power_law_1.2,0.09438080191612244
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,8,balanced,1.3931466738382976
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2,7168,2048,8,384,32,8,power_law_1.2,0.052883201837539674
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,8,power_law_1.01,0.6889920234680176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,8,power_law_1.01,0.06647040247917176
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,384,7168,2048,8,384,32,8,power_law_1.2,0.09571840167045594
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4,7168,2048,8,384,32,8,power_law_1.2,0.05258880257606506
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,8,power_law_1.01,0.9045184135437012
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,8,balanced,0.745797316233317
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,8,power_law_1.01,0.06841599941253662
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,512,7168,2048,8,384,32,8,power_law_1.2,0.1041983962059021
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8,7168,2048,8,384,32,8,power_law_1.2,0.048089599609375
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,8,power_law_1.01,0.0702015995979309
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,8,power_law_1.01,1.1453503608703612
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,768,7168,2048,8,384,32,8,power_law_1.2,0.11841919422149658
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16,7168,2048,8,384,32,8,power_law_1.2,0.05187199711799621
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,8,power_law_1.01,0.07388799786567687
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1024,7168,2048,8,384,32,8,power_law_1.2,0.1288383960723877
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,8,power_law_1.01,1.6740928649902345
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32,7168,2048,8,384,32,8,power_law_1.2,0.058156800270080564
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,8,power_law_1.01,0.07677440047264099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,1536,7168,2048,8,384,32,8,power_law_1.2,0.16151039600372313
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,48,7168,2048,8,384,32,8,power_law_1.2,0.0595583975315094
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,8,power_law_1.01,0.0825984001159668
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,2048,7168,2048,8,384,32,8,power_law_1.2,0.18788479566574096
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,8,power_law_1.01,3.355712127685547
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,64,7168,2048,8,384,32,8,power_law_1.2,0.06520320177078247
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,8,power_law_1.01,0.09025920033454896
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,3072,7168,2048,8,384,32,8,power_law_1.2,0.23454079627990723
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,80,7168,2048,8,384,32,8,power_law_1.2,0.06689280271530151
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,8,power_law_1.01,0.09685119986534119
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,4096,7168,2048,8,384,32,8,power_law_1.2,0.30737919807434083
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,96,7168,2048,8,384,32,8,power_law_1.2,0.06913279891014099
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,8,power_law_1.01,0.10885759592056274
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,6144,7168,2048,8,384,32,8,power_law_1.2,0.4060031890869141
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,128,7168,2048,8,384,32,8,power_law_1.2,0.07059839963912964
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,8,power_law_1.01,0.12549760341644287
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,8192,7168,2048,8,384,32,8,power_law_1.2,0.5241024017333984
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,160,7168,2048,8,384,32,8,power_law_1.2,0.07516800165176392
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,8,power_law_1.01,0.1356287956237793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,8,balanced,0.9578506946563721
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,192,7168,2048,8,384,32,8,power_law_1.2,0.07802240252494812
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,12288,7168,2048,8,384,32,8,power_law_1.2,0.8057215690612793
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,8,power_law_1.01,0.17253119945526124
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,256,7168,2048,8,384,32,8,power_law_1.2,0.08275200128555298
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,8,power_law_1.01,0.19684480428695678
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,16384,7168,2048,8,384,32,8,power_law_1.2,0.9829504013061523
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,320,7168,2048,8,384,32,8,power_law_1.2,0.09110400080680847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,8,power_law_1.01,0.2607232093811035
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,384,7168,2048,8,384,32,8,power_law_1.2,0.09816319942474365
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,20480,7168,2048,8,384,32,8,power_law_1.2,1.1414591789245605
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,8,power_law_1.01,0.3408512115478516
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,512,7168,2048,8,384,32,8,power_law_1.2,0.10810879468917847
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,8,power_law_1.01,0.45808000564575196
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,32768,7168,2048,8,384,32,8,power_law_1.2,1.8830272674560546
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,768,7168,2048,8,384,32,8,power_law_1.2,0.13241599798202514
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1024,7168,2048,8,384,32,8,power_law_1.2,0.13626240491867064
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,8,power_law_1.01,0.5875328063964844
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,8,power_law_1.2,3.838739013671875
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,1536,7168,2048,8,384,32,8,power_law_1.2,0.17655680179595948
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,8,power_law_1.01,0.8393152236938477
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,2048,7168,2048,8,384,32,8,power_law_1.2,0.2056960105895996
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,8,power_law_1.01,1.0963583946228028
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,3072,7168,2048,8,384,32,8,power_law_1.2,0.2753727912902832
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,8,power_law_1.01,1.3983039855957031
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,4096,7168,2048,8,384,32,8,power_law_1.2,0.34753921031951907
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,8,power_law_1.01,2.138707160949707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,6144,7168,2048,8,384,32,8,power_law_1.2,0.5019264221191406
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,8,balanced,1.191210667292277
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,8192,7168,2048,8,384,32,8,power_law_1.2,0.6645567893981934
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,8,power_law_1.01,4.090009689331055
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,12288,7168,2048,8,384,32,8,power_law_1.2,0.9154687881469726
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,16384,7168,2048,8,384,32,8,power_law_1.2,1.2794176101684571
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,20480,7168,2048,8,384,32,8,power_law_1.2,1.5399423599243165
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,8,power_law_1.2,2.447539138793945
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,8,power_law_1.2,4.539532852172852
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,float16,65536,7168,2048,8,384,32,8,balanced,2.7125492095947266
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,32768,7168,2048,8,384,32,8,balanced,1.836949348449707
TRTLLM,1.2.0rc5,NVIDIA B200,moe,moe_torch_flow,fp8,65536,7168,2048,8,384,32,8,balanced,3.6208054224650064
