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
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,1,power_law_1.01,0.21703744888305665
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,1,power_law_1.01,0.2861427116394043
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,1,power_law_1.01,0.4914316940307617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,1,power_law_1.01,0.3305759811401367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,1,power_law_1.01,0.6043558502197266
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,1,power_law_1.01,0.6101414489746093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,1,power_law_1.01,0.6649286651611328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,1,power_law_1.01,0.6058278274536133
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,1,power_law_1.01,0.7968339538574218
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,1,power_law_1.01,0.8707276916503905
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,1,power_law_1.01,1.0390592193603516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,1,power_law_1.01,1.227026596069336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,1,power_law_1.01,1.61727294921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,1,power_law_1.01,2.0438911437988283
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,1,power_law_1.01,2.8835546874999998
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,1,power_law_1.01,0.6006739044189453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,1,power_law_1.01,0.6134630584716797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,1,power_law_1.01,0.6250368118286133
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,1,power_law_1.01,0.6437875366210937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,1,power_law_1.01,0.7126943969726562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,1,power_law_1.01,3.7233184814453124
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,1,power_law_1.01,5.55913818359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,1,power_law_1.01,7.400570678710937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,1,power_law_1.01,0.11799807548522949
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,1,power_law_1.01,0.15329279899597167
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,1,power_law_1.01,0.17833023071289061
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,1,power_law_1.01,11.87507080078125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,1,power_law_1.01,0.2645350456237793
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,1,power_law_1.01,16.408613281250002
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,1,power_law_1.01,0.27948799133300783
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,1,power_law_1.01,0.29039936065673827
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,1,power_law_1.01,0.29849279403686524
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,1,power_law_1.01,0.3007814407348633
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,1,power_law_1.01,0.2886553573608398
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,1,power_law_1.01,0.3068332862854004
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,1,power_law_1.01,0.3170041656494141
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,1,power_law_1.01,0.35914817810058597
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,1,power_law_1.01,0.3783001708984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,1,power_law_1.01,0.4004134368896485
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,1,power_law_1.01,0.28713279724121094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,1,power_law_1.01,0.4330624008178711
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,1,power_law_1.01,0.28098880767822265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,1,power_law_1.01,0.5307091140747071
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,1,power_law_1.01,0.6651641845703125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,1,power_law_1.01,0.8877536010742186
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,1,power_law_1.01,1.1390918731689452
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,1,power_law_1.01,1.6192825317382813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,1,power_law_1.01,2.123063049316406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,1,power_law_1.01,3.0566778564453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,1,power_law_1.01,4.093453979492187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,1,power_law_1.01,26.101496582031253
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,1,power_law_1.01,6.064210205078124
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,1,power_law_1.01,8.081775512695312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,1,1,power_law_1.01,10.011478881835938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,1,power_law_1.01,36.920522460937505
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,1,1,power_law_1.01,47.227680664062504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,1,power_law_1.01,0.05912255764007568
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,1,power_law_1.01,0.15757887840270995
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,1,power_law_1.01,0.16114559173583984
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,1,power_law_1.01,0.16299392700195312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,1,power_law_1.01,0.07976831912994384
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,1,power_law_1.01,0.09880767822265625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,1,power_law_1.01,0.1555846405029297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,1,power_law_1.01,0.1591500759124756
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,1,power_law_1.01,0.15762816429138185
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,1,power_law_1.01,0.15841407775878907
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,1,power_law_1.01,0.15553279876708986
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,1,power_law_1.01,0.1617593574523926
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,1,power_law_1.01,0.1701100730895996
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,1,power_law_1.01,0.18760000228881837
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,1,power_law_1.01,0.19574272155761718
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,1,power_law_1.01,0.20680639266967776
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,1,power_law_1.01,0.23539648056030274
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,1,power_law_1.01,0.3013555145263672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.01,0.3556256103515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.01,0.4744697570800781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.01,0.5862764739990235
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,1,power_law_1.2,0.23353343963623047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,1,power_law_1.2,0.27742015838623046
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,1,power_law_1.2,0.31474111557006834
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,1,power_law_1.2,0.5017113494873047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,1,power_law_1.2,0.5837804794311523
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,1,power_law_1.2,0.614192008972168
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.01,1.1270841979980468
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.01,0.879402847290039
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,1,power_law_1.2,0.6164934539794922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.01,1.6215621948242187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,1,power_law_1.2,0.6187148666381835
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,1,power_law_1.2,0.63102783203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,1,power_law_1.2,0.6462886047363281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,1,power_law_1.2,0.6896460723876954
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,1,power_law_1.2,0.6926847839355469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,1,power_law_1.2,0.7040518188476563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,1,power_law_1.2,0.8029209899902344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,1,power_law_1.2,0.8520089721679687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,1,power_law_1.2,1.0750137329101563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,1,power_law_1.2,1.2707430267333986
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,1,power_law_1.2,1.6159155273437498
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,1,power_law_1.2,2.070655975341797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,1,power_law_1.2,2.8865274047851566
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,1,power_law_1.2,3.781269836425781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.01,2.1892262268066407
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,1,power_law_1.2,5.555742797851563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,1,power_law_1.2,7.550615844726562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.01,3.250808410644531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,1,power_law_1.2,0.13292351722717285
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,1,power_law_1.2,0.14626943588256835
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,1,power_law_1.2,0.1696326446533203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,1,power_law_1.2,0.26339136123657225
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,1,power_law_1.2,0.2738035202026367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,1,power_law_1.2,0.28941247940063475
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,1,power_law_1.2,0.28945663452148435
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,1,power_law_1.2,0.28800191879272463
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,1,power_law_1.2,0.29992448806762695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,1,power_law_1.2,0.3036838340759277
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,1,power_law_1.2,0.3053625679016113
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,1,power_law_1.2,0.3175391960144043
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,1,power_law_1.2,0.31373695373535154
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,1,power_law_1.2,12.286893310546876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,1,power_law_1.2,0.359857292175293
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,1,power_law_1.2,0.396129264831543
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,1,power_law_1.2,0.41623744964599607
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,1,power_law_1.2,0.4385676956176757
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,1,power_law_1.2,0.5378662490844727
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,1,power_law_1.2,0.6578623962402343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,1,power_law_1.2,0.8826918029785157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,1,power_law_1.2,1.1360467529296874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,1,power_law_1.2,1.6493125915527345
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,1,power_law_1.2,16.799666748046874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,1,power_law_1.2,2.103223114013672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.01,4.268653564453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,1,power_law_1.2,3.1023373413085937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,1,power_law_1.2,4.113463745117188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,1,power_law_1.2,0.06427199840545654
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,1,power_law_1.2,0.07395455837249756
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,1,power_law_1.2,0.09002431869506836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,1,power_law_1.2,0.15452351570129394
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,1,power_law_1.2,0.15192511558532715
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,1,power_law_1.2,0.15824383735656739
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,1,power_law_1.2,6.041503295898438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,1,power_law_1.2,0.15919103622436523
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,1,power_law_1.2,0.15913472175598145
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,1,power_law_1.2,0.16044544219970702
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,1,power_law_1.2,0.16259519577026366
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,1,power_law_1.2,0.16332479476928713
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,1,power_law_1.2,0.1659609603881836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,1,power_law_1.2,0.18829248428344728
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,1,power_law_1.2,26.6979248046875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,1,power_law_1.2,8.057903442382813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,1,1,power_law_1.2,10.004721069335938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,1,power_law_1.2,0.31711616516113283
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.01,5.3096435546875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.2,0.5224364852905273
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,1,power_law_1.2,37.4897509765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.2,0.8828505706787109
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,1,1,power_law_1.2,47.2976318359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,1,power_law_1.2,0.1692723274230957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,1,power_law_1.2,0.21687360763549807
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,1,power_law_1.2,0.2003718376159668
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,1,power_law_1.2,0.23635839462280273
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.2,0.3586284637451172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,1,power_law_1.01,0.3129836845397949
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,1,power_law_1.01,0.4186924743652344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,1,power_law_1.01,0.5075711822509765
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,1,power_law_1.01,0.8136326599121094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,1,power_law_1.01,0.9302297973632813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,1,power_law_1.01,0.945568618774414
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,1,power_law_1.01,0.9506476593017578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.2,0.6098054504394531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,1,power_law_1.01,0.9560332489013671
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,1,power_law_1.01,0.9720722961425782
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,1,power_law_1.01,1.00461181640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,1,power_law_1.01,1.1106227111816405
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,1,power_law_1.01,1.2340089416503905
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,1,power_law_1.01,1.3064472961425782
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.2,1.142863998413086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,1,power_law_1.01,1.461523895263672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,1,power_law_1.01,1.5833229064941405
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,1,power_law_1.01,1.6672288513183595
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,1,power_law_1.01,2.080124816894531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,1,power_law_1.01,2.832024841308594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,1,power_law_1.01,3.4981729125976564
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,1,power_law_1.01,4.944936218261718
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,1,power_law_1.01,6.353118896484375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,1,power_law_1.01,9.327166748046874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.2,1.6525184631347656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,1,power_law_1.01,12.652592773437501
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.2,2.2446585083007813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.2,3.3150463867187496
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,1,power_law_1.01,20.006346435546874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,1,power_law_1.01,0.1604595184326172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,1,power_law_1.01,0.21427135467529296
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,1,power_law_1.01,0.2619462394714355
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.2,4.3562060546875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,1,power_law_1.01,0.43182209014892575
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,1,power_law_1.01,0.4621043014526367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,1,power_law_1.01,0.4684307098388672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,1,power_law_1.01,0.4760531234741211
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,1,power_law_1.01,0.4659078216552734
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,1,power_law_1.01,0.4699635314941406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,1,power_law_1.01,0.4821209716796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,1,power_law_1.01,0.4877555084228516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,1,power_law_1.01,0.4972748947143555
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.2,5.4401007080078125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,1,power_law_1.01,0.5032704162597657
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,1,power_law_1.01,0.5334259033203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,1,power_law_1.01,0.5500691223144532
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,1,power_law_1.01,0.572619514465332
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,1,power_law_1.01,0.64853759765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,1,power_law_1.01,0.7940934753417969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,1,power_law_1.01,1.0019884490966797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,1,power_law_1.01,1.3158131408691407
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,1,power_law_1.01,1.6653587341308593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,1,power_law_1.01,2.4753369140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,1,power_law_1.01,3.0980679321289064
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,1,power_law_1.01,4.5107751464843755
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,1,power_law_1.01,28.072446289062498
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,1,power_law_1.01,5.926356201171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,1,power_law_1.01,8.726641235351561
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,1,power_law_1.01,11.64810791015625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,1,1,power_law_1.01,14.6006201171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,1,power_law_1.01,44.8533984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,1,power_law_1.01,63.4055908203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,1,power_law_1.01,0.08050944328308104
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,1,power_law_1.01,0.24739967346191408
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,1,power_law_1.01,0.2478175926208496
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,1,power_law_1.01,0.1210048007965088
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,1,power_law_1.01,0.2530624008178711
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,1,power_law_1.01,0.25380096435546873
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,1,power_law_1.01,0.15070464134216308
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,1,power_law_1.01,0.24479360580444337
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,1,power_law_1.01,0.24688640594482422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,1,power_law_1.01,0.2494495964050293
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,1,1,power_law_1.01,83.18716796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,1,power_law_1.01,0.24897792816162106
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,1,power_law_1.01,0.2663360023498535
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,1,power_law_1.01,0.2543948745727539
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,1,power_law_1.01,0.30410495758056644
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,1,power_law_1.01,0.31658559799194336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,1,power_law_1.01,0.3391775894165039
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,1,power_law_1.01,0.5157772827148437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,1,power_law_1.01,0.3868217468261719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.01,0.6216972732543946
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.01,0.8080703735351562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.01,1.0227903747558593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,1,power_law_1.2,0.3559756851196289
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,1,power_law_1.2,0.3883846282958984
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,1,power_law_1.2,0.4845836639404297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,1,power_law_1.2,0.8149657440185546
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,1,power_law_1.2,0.9062509155273437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.01,1.3800653076171874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,1,power_law_1.2,0.9354099273681641
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,1,power_law_1.2,0.9646176147460939
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,1,power_law_1.2,0.9467040252685546
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,1,power_law_1.2,0.9982444763183593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,1,power_law_1.2,1.0187910461425782
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,1,power_law_1.2,1.1548851013183594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,1,power_law_1.2,1.2517183685302735
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.01,1.8656927490234374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,1,power_law_1.2,1.3085939025878905
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,1,power_law_1.2,1.4576095581054687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,1,power_law_1.2,1.5343423461914063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,1,power_law_1.2,1.7897824096679689
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,1,power_law_1.2,2.0586866760253906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,1,power_law_1.2,2.833775939941406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,1,power_law_1.2,3.4643975830078126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.01,2.7604345703125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,1,power_law_1.2,4.883283081054687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.01,3.6397824096679683
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,1,power_law_1.2,6.3771557617187495
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.01,5.388457641601563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,1,power_law_1.2,9.29958984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,1,power_law_1.2,12.692103271484374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,1,power_law_1.2,0.16113216400146485
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,1,power_law_1.2,0.19879104614257812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,1,power_law_1.2,0.23899904251098633
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,1,power_law_1.2,0.43104896545410154
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,1,power_law_1.2,0.44825984954833986
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,1,power_law_1.2,0.4661356735229492
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,1,power_law_1.2,0.46753982543945316
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,1,power_law_1.2,0.4689510345458984
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,1,power_law_1.2,0.4787039947509766
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,1,power_law_1.2,0.48496768951416014
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,1,power_law_1.2,0.4868934249877929
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,1,power_law_1.2,0.49802814483642577
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,1,power_law_1.2,0.4966239929199219
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,1,power_law_1.2,0.5318790435791015
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,1,power_law_1.2,20.312810058593747
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,1,power_law_1.2,0.5530956649780274
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,1,power_law_1.2,0.5862406539916992
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,1,power_law_1.2,0.6791289520263672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,1,power_law_1.2,0.8349478149414062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,1,power_law_1.2,1.0024249267578125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,1,power_law_1.2,1.3381727600097657
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,1,power_law_1.2,1.7049005126953127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.01,7.177188720703126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,1,power_law_1.2,2.362244415283203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,1,power_law_1.2,3.1438885498046876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,1,power_law_1.2,28.559768066406253
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,1,power_law_1.2,4.435881652832031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,1,power_law_1.2,5.885494384765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,1,power_law_1.2,0.08126655578613282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,1,power_law_1.2,0.10901951789855957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,1,power_law_1.2,0.14078080177307128
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,1,power_law_1.2,0.24547456741333007
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,1,power_law_1.2,0.24142143249511722
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,1,power_law_1.2,0.24977855682373046
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,1,power_law_1.2,0.25100671768188476
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,1,power_law_1.2,0.24752639770507812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,1,power_law_1.2,0.25237823486328126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,1,power_law_1.2,8.781618041992187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,1,power_law_1.2,0.2540204811096191
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,1,power_law_1.2,0.2619923210144043
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.01,8.733826293945313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,1,power_law_1.2,0.32355457305908203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,1,power_law_1.2,11.74797119140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,1,1,power_law_1.2,14.549069824218751
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,1,power_law_1.2,45.534565429687504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,1,power_law_1.2,64.5156494140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,1,power_law_1.2,0.2591276741027832
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,1,power_law_1.2,0.2720115280151367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,1,power_law_1.2,0.3247891235351562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,1,1,power_law_1.2,84.07786132812501
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,1,power_law_1.2,0.3534457778930664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,1,power_law_1.2,0.3989766311645508
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,1,power_law_1.2,0.510081901550293
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.2,0.6131462478637696
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.2,0.8000390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.2,1.0277005004882813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.2,1.4353535461425782
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.2,1.9112229919433594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.2,2.7600491333007815
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,1,power_law_1.01,0.175098876953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,1,power_law_1.01,0.3198784065246582
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,1,power_law_1.01,0.3568032073974609
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,1,power_law_1.01,0.7392614746093751
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.2,3.637784423828125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.2,5.34316650390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,1,power_law_1.01,2.5215008544921877
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,1,power_law_1.01,2.6311614990234373
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,1,power_law_1.01,2.745836181640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,1,power_law_1.01,1.2207341003417969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,1,power_law_1.01,1.6930706787109375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,1,power_law_1.01,2.8017401123046874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,1,power_law_1.01,3.02258544921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,1,power_law_1.01,3.225355529785156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,1,power_law_1.01,3.4176101684570312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,1,power_law_1.01,3.773502807617187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,1,power_law_1.01,3.545264587402344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,1,power_law_1.01,3.4691058349609376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,1,power_law_1.01,3.762244567871094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,1,power_law_1.01,4.479508972167968
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,1,power_law_1.01,4.122852478027344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,1,power_law_1.01,5.295780639648437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,1,power_law_1.01,6.0510406494140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,1,power_law_1.01,7.667765502929687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.2,7.187093505859375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,1,power_law_1.01,9.406217041015626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,1,power_law_1.01,12.7198486328125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,1,power_law_1.01,15.87328125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,1,power_law_1.01,0.17928127288818357
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,1,power_law_1.01,0.23508031845092772
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,1,power_law_1.01,0.26870336532592776
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,1,power_law_1.01,0.9005401611328125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,1,power_law_1.01,0.44622718811035156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,1,power_law_1.01,1.2673497772216797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,1,power_law_1.01,1.3327352905273437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.2,8.877879638671875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,1,power_law_1.01,1.3647091674804688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,1,power_law_1.01,22.3804931640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,1,power_law_1.01,1.4196287536621095
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,1,power_law_1.01,1.4971974182128907
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,1,power_law_1.01,1.5218258666992188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,1,power_law_1.01,1.5698060607910156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,1,power_law_1.01,0.6792928314208984
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,1,power_law_1.01,1.8397068786621094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,1,power_law_1.01,1.612872314453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,1,power_law_1.01,28.8489501953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,1,power_law_1.01,1.7922540283203126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,1,power_law_1.01,1.8817369079589845
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,1,power_law_1.01,1.9510540771484375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,1,power_law_1.01,2.2552088928222656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,1,power_law_1.01,2.054502410888672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,1,power_law_1.01,2.951243591308594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,1,power_law_1.01,3.401506042480469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,1,power_law_1.01,2.5057279968261716
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,1,1,power_law_1.01,35.26718017578125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,1,power_law_1.01,4.287310791015625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,1,power_law_1.01,5.237761230468751
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,1,power_law_1.01,7.17143798828125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,1,power_law_1.01,9.053586425781251
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,1,1,power_law_1.01,10.892145996093749
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,1,power_law_1.01,0.08875840187072755
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,1,power_law_1.01,0.12165184020996093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,1,power_law_1.01,0.1344108772277832
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,1,power_law_1.01,0.81959228515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,1,power_law_1.01,0.37997310638427734
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,1,power_law_1.01,0.8389395141601563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,1,power_law_1.01,0.8842662048339843
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,1,power_law_1.01,0.2376972770690918
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,1,power_law_1.01,0.9155142211914062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,1,power_law_1.01,0.5183065414428711
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,1,power_law_1.01,0.9363795471191405
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,1,power_law_1.01,0.9735628509521485
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,1,power_law_1.01,1.0047309112548828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,1,power_law_1.01,0.7551910400390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,1,power_law_1.01,1.0404351806640624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,1,power_law_1.01,1.0690675354003907
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,1,power_law_1.01,0.7876000213623047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.01,1.257017593383789
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.01,1.4636915588378907
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,1,power_law_1.01,1.1723661041259765
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.01,2.0605068969726563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.01,2.4642373657226564
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.01,1.6094406127929688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.01,3.2590579223632816
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,1,power_law_1.2,0.34307136535644533
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,1,power_law_1.2,0.6633689880371094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,1,power_law_1.2,1.0579443359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,1,power_law_1.2,0.17777408599853514
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,1,power_law_1.2,1.5069357299804689
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.01,4.091351623535156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,1,power_law_1.2,2.415515594482422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,1,power_law_1.2,2.694261169433594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,1,power_law_1.2,0.29889535903930664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,1,power_law_1.2,2.6015713500976565
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,1,power_law_1.2,2.7958081054687502
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,1,power_law_1.2,3.0568862915039063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,1,power_law_1.2,3.2826989746093753
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,1,power_law_1.2,3.494388427734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,1,power_law_1.2,3.930753173828125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,1,power_law_1.2,3.6123123168945312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,1,power_law_1.2,3.7274777221679685
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,1,power_law_1.2,3.9267398071289064
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,1,power_law_1.2,4.382020568847656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,1,power_law_1.2,5.6523590087890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,1,power_law_1.2,4.760389709472657
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,1,power_law_1.2,6.469424438476563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,1,power_law_1.2,8.096573486328126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,1,power_law_1.2,9.7989208984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.01,5.724470825195313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.01,7.565343627929687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,1,power_law_1.2,12.83334716796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,1,power_law_1.2,0.22793983459472655
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,1,1,power_law_1.01,9.185422973632813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,1,power_law_1.2,0.4134745788574219
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,1,power_law_1.2,0.17707775115966798
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,1,power_law_1.2,0.8045452880859376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,1,power_law_1.2,16.143687744140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,1,power_law_1.2,1.2322144317626953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,1,power_law_1.2,1.2958303833007814
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,1,power_law_1.2,1.3628218078613281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,1,power_law_1.2,0.2504857635498047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,1,power_law_1.2,1.411751708984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,1,power_law_1.2,1.470436553955078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,1,power_law_1.2,1.5554489135742187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,1,power_law_1.2,0.5992268753051758
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,1,power_law_1.2,1.550320587158203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,1,power_law_1.2,1.6662200927734374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,1,power_law_1.2,22.526960449218752
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,1,power_law_1.2,1.8640473937988282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,1,power_law_1.2,1.8405958557128905
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,1,power_law_1.2,1.9540818786621095
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,1,power_law_1.2,2.0781292724609375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,1,power_law_1.2,2.145857238769531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,1,power_law_1.2,2.6115142822265627
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,1,power_law_1.2,3.1087832641601563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,1,power_law_1.2,3.607781066894531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,1,power_law_1.2,2.4058374023437503
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,1,power_law_1.2,4.466482543945313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,1,power_law_1.2,29.243388671875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,1,power_law_1.2,0.08847104072570801
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,1,power_law_1.2,5.34742919921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,1,power_law_1.2,0.12034111976623536
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,1,power_law_1.2,0.12034111976623536
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,1,power_law_1.2,0.2185420799255371
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,1,power_law_1.2,0.3333734512329102
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,1,power_law_1.2,0.46293888092041013
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,1,power_law_1.2,0.7269734191894531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,1,power_law_1.2,0.7785132598876954
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,1,power_law_1.2,0.8085273742675781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,1,power_law_1.2,7.212085571289063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,1,power_law_1.2,0.8385062408447265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,1,power_law_1.2,9.05861083984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,1,power_law_1.2,0.8902777862548827
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,1,power_law_1.2,0.9224845123291014
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,1,power_law_1.2,0.949920654296875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,1,1,power_law_1.2,11.001258544921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,1,power_law_1.2,1.0100077056884766
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,1,1,power_law_1.2,35.724052734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,1,power_law_1.2,1.0463635253906252
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,1,power_law_1.2,1.09804931640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,1,power_law_1.2,1.2802079772949218
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.2,1.408025665283203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.2,1.676958770751953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.2,1.9237414550781249
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,1,power_law_1.2,1.1463449859619141
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.2,2.5000108337402347
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.2,3.0494439697265627
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,1,power_law_1.01,0.03967616081237793
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.2,4.311902160644531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,1,power_law_1.01,0.05011519908905029
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,1,power_law_1.01,0.2102118492126465
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,1,power_law_1.01,0.20205312728881836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,1,power_law_1.01,0.21654144287109375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,1,power_law_1.01,0.220248966217041
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,1,power_law_1.01,0.2369638442993164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,1,power_law_1.01,0.24379968643188477
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,1,power_law_1.01,0.24287744522094729
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,1,power_law_1.01,0.2556531143188477
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,1,power_law_1.01,0.27250431060791014
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,1,power_law_1.01,0.2777689552307129
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,1,power_law_1.01,0.18456064224243166
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,1,power_law_1.01,0.2896620750427246
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,1,power_law_1.01,0.3247328186035156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,1,power_law_1.01,0.3558611297607422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,1,power_law_1.01,0.43700542449951174
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,1,power_law_1.01,0.06862592220306396
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,1,power_law_1.01,0.5245753479003906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,1,power_law_1.01,0.6963308715820313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,1,power_law_1.01,0.893421401977539
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,1,power_law_1.01,1.2657542419433594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,1,power_law_1.01,1.6902957153320313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,1,power_law_1.01,2.3973933410644532
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,1,power_law_1.01,0.048764801025390624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,1,power_law_1.01,3.1928793334960934
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,1,power_law_1.01,0.06445184230804443
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,1,1,power_law_1.01,3.9401037597656248
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,1,power_law_1.01,0.07916096210479737
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,1,power_law_1.01,0.09608063697814942
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,1,power_law_1.01,0.14448320388793945
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,1,power_law_1.01,0.044742398262023926
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,1,power_law_1.01,0.14747455596923828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,1,power_law_1.01,0.15157759666442872
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,1,power_law_1.01,0.1522214412689209
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,1,power_law_1.01,0.1734489631652832
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,1,power_law_1.01,0.1306668758392334
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,1,power_law_1.01,0.15451775550842287
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,1,power_law_1.01,0.17887487411499023
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,1,power_law_1.01,0.1727686309814453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,1,power_law_1.01,0.18141504287719729
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,1,power_law_1.01,0.1871820831298828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,1,power_law_1.01,0.20023679733276367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,1,power_law_1.01,0.20934783935546877
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,1,power_law_1.01,0.22763904571533203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,1,power_law_1.01,0.26330047607421875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,1,power_law_1.01,0.30985088348388673
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,1,power_law_1.01,0.38831104278564454
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,1,power_law_1.01,0.46679233551025395
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,1,power_law_1.01,0.6345619201660156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,1,power_law_1.01,0.8056697845458984
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,1,power_law_1.01,1.1421580505371094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,1,power_law_1.01,1.4712249755859375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,1,1,power_law_1.01,1.8043898010253905
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.2,5.645908203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,1,power_law_1.01,0.03329920053482056
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,1,power_law_1.01,0.03987391948699951
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,1,power_law_1.01,0.07884607791900636
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,1,power_law_1.01,0.08408639907836915
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,1,power_law_1.01,0.045148801803588864
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,1,power_law_1.01,0.086943359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,1,power_law_1.01,0.0889958381652832
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,1,power_law_1.01,0.0519871997833252
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,1,power_law_1.01,0.0928390407562256
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,1,power_law_1.01,0.08950207710266114
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,1,power_law_1.01,0.09493375778198242
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,1,power_law_1.01,0.09608063697814942
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,1,power_law_1.01,0.10152576446533204
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,1,power_law_1.01,0.059425277709960936
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,1,power_law_1.01,0.10542271614074708
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,1,power_law_1.01,0.10919551849365235
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,1,power_law_1.01,0.11840767860412597
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,1,power_law_1.01,0.13680319786071776
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,1,power_law_1.01,0.1557171154022217
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,1,power_law_1.01,0.19875263214111327
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,1,power_law_1.01,0.4316006469726562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,1,power_law_1.01,0.24481664657592775
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,1,power_law_1.01,0.33643775939941406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,1,power_law_1.01,0.630995216369629
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,1,power_law_1.2,0.04677887916564941
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,1,power_law_1.2,0.1215385627746582
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,1,power_law_1.01,0.8346444702148437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,1,power_law_1.2,0.18822463989257812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,1,power_law_1.2,0.2043609619140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,1,power_law_1.2,0.21233919143676755
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,1,power_law_1.2,0.22201791763305664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,1,power_law_1.2,0.22265024185180665
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,1,power_law_1.2,0.24004543304443357
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,1,power_law_1.2,0.24319616317749024
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,1,power_law_1.2,0.24726655960083005
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,1,power_law_1.2,0.036980481147766114
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,1,power_law_1.2,0.2648409652709961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,1,power_law_1.2,0.27744384765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,1,power_law_1.01,1.2490521240234376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,1,1,power_law_1.01,2.099517364501953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,1,power_law_1.2,0.33000511169433594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,1,power_law_1.2,0.290948486328125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,1,power_law_1.2,0.36274112701416017
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,1,power_law_1.2,0.44630401611328124
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,1,power_law_1.01,1.6769381713867186
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,1,power_law_1.2,0.08545408248901368
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,1,power_law_1.2,0.060155520439147944
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,1,power_law_1.2,0.5270739364624023
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,1,power_law_1.2,0.27606336593627934
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,1,power_law_1.2,0.710606689453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,1,power_law_1.2,0.9111264038085938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,1,power_law_1.2,0.06024640083312989
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,1,power_law_1.2,1.292941436767578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,1,power_law_1.2,0.07377855777740479
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,1,power_law_1.2,0.09305600166320802
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,1,power_law_1.2,0.1337062358856201
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,1,power_law_1.2,0.14637248039245604
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,1,power_law_1.2,0.14784832000732423
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,1,power_law_1.2,0.15369600296020508
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,1,power_law_1.2,0.04337152004241943
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,1,power_law_1.2,0.15095104217529295
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,1,power_law_1.2,3.1930859375000002
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,1,power_law_1.2,0.15561599731445314
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,1,power_law_1.2,0.1759129524230957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,1,power_law_1.2,0.1730009651184082
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,1,power_law_1.2,0.1839878463745117
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,1,power_law_1.2,0.1912665557861328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,1,1,power_law_1.2,3.947726135253906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,1,power_law_1.2,0.194072322845459
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,1,power_law_1.2,2.4110520935058597
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,1,power_law_1.2,0.04867584228515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,1,power_law_1.2,0.2006912040710449
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,1,power_law_1.2,0.21308095932006835
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,1,power_law_1.2,0.23220672607421874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,1,power_law_1.2,0.31290239334106446
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,1,power_law_1.2,0.47776447296142577
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,1,power_law_1.2,0.6431046295166015
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,1,power_law_1.2,0.8114342498779298
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,1,power_law_1.2,1.7017190551757815
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,1,power_law_1.2,1.146341781616211
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,1,power_law_1.2,0.03447616100311279
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,1,power_law_1.2,0.039918720722198486
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,1,power_law_1.2,1.4766029357910155
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,1,power_law_1.2,0.04973696231842041
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,1,power_law_1.2,0.05668992042541504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,1,power_law_1.2,0.07914559841156006
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,1,1,power_law_1.2,1.814780731201172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,1,power_law_1.2,0.27285247802734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,1,power_law_1.2,0.38619903564453123
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,1,power_law_1.2,0.09043968200683594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,1,power_law_1.2,0.0879321575164795
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,1,power_law_1.2,0.0900607967376709
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,1,power_law_1.2,0.08555520057678223
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,1,power_law_1.2,0.09294719696044922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,1,power_law_1.2,0.1120582389831543
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,1,power_law_1.2,0.10624256134033203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,1,power_law_1.2,0.042768640518188475
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,1,power_law_1.2,0.09809663772583008
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,1,power_law_1.2,0.11608192443847656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,1,power_law_1.2,0.1563596820831299
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,1,power_law_1.2,0.12933312416076662
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,1,power_law_1.2,0.09932991981506348
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,1,power_law_1.2,0.18522432327270508
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,1,power_law_1.2,0.2500339126586914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,1,power_law_1.2,0.3153145599365234
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,1,power_law_1.2,0.5971295928955078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,1,power_law_1.2,0.8757241821289062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,1,power_law_1.01,0.1032953643798828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,1,power_law_1.2,1.1526348876953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,1,power_law_1.01,0.19959808349609376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,1,power_law_1.2,0.45444606781005864
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,1,power_law_1.01,0.3028767967224121
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,1,power_law_1.01,0.39506622314453127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,1,power_law_1.01,0.644027557373047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,1,power_law_1.01,0.7299884796142578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,1,power_law_1.01,0.7537554931640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,1,power_law_1.01,0.7712185668945313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,1,power_law_1.2,2.2644192504882814
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,1,power_law_1.01,0.8317005157470703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,1,power_law_1.01,0.8666783905029296
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,1,power_law_1.2,1.7087373352050783
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,1,power_law_1.01,0.8729036712646485
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,1,power_law_1.01,0.7040428924560547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,1,power_law_1.01,0.9021785736083985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,1,power_law_1.01,0.934037094116211
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,1,power_law_1.01,0.9513465881347656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,1,power_law_1.01,0.0877228832244873
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,1,power_law_1.01,1.0045708465576173
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,1,power_law_1.01,1.1689810943603516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,1,power_law_1.01,1.2972479248046875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,1,power_law_1.01,1.630530548095703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,1,1,power_law_1.2,2.830929870605469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,1,power_law_1.01,2.0100775146484375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,1,power_law_1.01,2.6909939575195314
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,1,power_law_1.01,0.0929747200012207
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,1,power_law_1.01,3.3180267333984377
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,1,power_law_1.01,0.13030079841613768
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,1,power_law_1.01,0.23123584747314455
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,1,power_law_1.01,4.80591552734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,1,power_law_1.01,0.35042049407958986
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,1,power_law_1.01,6.201131591796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,1,power_law_1.01,0.3975667190551758
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,1,power_law_1.01,0.0757855987548828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,1,power_law_1.01,0.4068646240234375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,1,power_law_1.01,0.40950847625732417
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,1,power_law_1.01,0.3804249572753906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,1,power_law_1.01,0.46986942291259765
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,1,power_law_1.01,0.46717056274414065
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,1,power_law_1.01,0.47645694732666016
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,1,power_law_1.01,9.002359008789062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,1,power_law_1.01,0.4972889709472656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,1,power_law_1.01,0.5004288101196289
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,1,power_law_1.01,0.5252959823608399
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,1,power_law_1.01,0.5546137619018554
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,1,power_law_1.01,0.6079321670532226
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,1,power_law_1.01,0.41906688690185545
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,1,power_law_1.01,0.6888703918457031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,1,power_law_1.01,0.8260364532470703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,1,power_law_1.01,1.040738525390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,1,power_law_1.01,11.759248046875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,1,power_law_1.01,1.2582994842529298
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,1,power_law_1.01,1.75219970703125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,1,power_law_1.01,2.1747564697265624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,1,power_law_1.01,0.18233024597167968
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,1,power_law_1.01,3.0839828491210937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,1,1,power_law_1.01,14.51055908203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,1,power_law_1.01,4.010790405273438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,1,1,power_law_1.01,5.034920959472656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,1,power_law_1.01,0.04743936061859131
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,1,power_law_1.01,0.055878400802612305
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,1,power_law_1.01,0.25108415603637696
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,1,power_law_1.01,0.07661248207092285
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,1,power_law_1.01,0.2541017532348633
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,1,power_law_1.01,0.10754048347473144
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,1,power_law_1.01,0.26323455810546875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,1,power_law_1.01,0.13768256187438965
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,1,power_law_1.01,0.26963712692260744
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,1,power_law_1.01,0.2712646484375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,1,power_law_1.01,0.21586559295654295
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,1,power_law_1.01,0.2857683181762695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,1,power_law_1.01,0.23845247268676756
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,1,power_law_1.01,0.24543615341186525
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,1,power_law_1.01,0.30463743209838867
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,1,power_law_1.01,0.2969715118408203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,1,power_law_1.01,0.3276921463012695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,1,power_law_1.01,0.3706444931030274
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.01,0.5254009628295899
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.01,0.41825408935546876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.01,0.6405753326416016
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.01,0.8643475341796876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,1,power_law_1.2,0.09624383926391603
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,1,power_law_1.2,0.18017663955688476
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,1,power_law_1.2,0.2614681625366211
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,1,power_law_1.2,0.08685888290405273
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,1,power_law_1.2,0.35717632293701174
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,1,power_law_1.2,0.71608642578125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,1,power_law_1.2,0.7437561798095703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,1,power_law_1.2,0.7709766387939453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,1,power_law_1.2,0.7955699157714844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,1,power_law_1.2,0.6565164947509766
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,1,power_law_1.2,0.8590873718261719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,1,power_law_1.2,0.881207046508789
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,1,power_law_1.2,0.8931084442138673
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,1,power_law_1.2,0.914343032836914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,1,power_law_1.2,0.9591129302978516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,1,power_law_1.2,0.9716896057128906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,1,power_law_1.2,1.0377484893798827
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,1,power_law_1.2,1.205391387939453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,1,power_law_1.2,1.3618476867675782
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,1,power_law_1.2,1.7440928649902343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,1,power_law_1.2,2.025841979980469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.01,1.1084159851074218
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.01,1.5953471374511718
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,1,power_law_1.2,2.7283154296875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.01,2.0893836975097657
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,1,power_law_1.2,3.4009869384765623
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.01,4.147374572753906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,1,power_law_1.2,4.7362451171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.01,3.1023263549804687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.01,5.211663818359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,1,power_law_1.2,0.11901951789855958
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,1,power_law_1.2,0.164715518951416
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,1,power_law_1.2,0.2084294319152832
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,1,power_law_1.2,0.35490047454833984
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,1,power_law_1.2,0.38461822509765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,1,power_law_1.2,6.225116577148437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,1,power_law_1.2,0.3978656005859375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,1,power_law_1.2,0.40893505096435545
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,1,power_law_1.2,0.4183379364013672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,1,power_law_1.2,0.41949825286865233
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,1,power_law_1.2,0.47106174468994144
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,1,power_law_1.2,0.4814944076538086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,1,power_law_1.2,0.07792640209197998
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,1,power_law_1.2,0.49485694885253906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,1,power_law_1.2,0.5037567901611328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,1,power_law_1.2,0.5129657745361328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,1,power_law_1.2,9.080067138671875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,1,power_law_1.2,0.5705235290527344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,1,power_law_1.2,0.6077542495727539
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,1,power_law_1.2,0.7320787048339843
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,1,power_law_1.2,0.8176595306396484
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,1,power_law_1.2,0.0841215991973877
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,1,power_law_1.2,1.0427123260498046
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,1,power_law_1.2,1.2523731231689452
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,1,power_law_1.2,11.7591357421875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,1,power_law_1.2,1.7026290893554688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,1,power_law_1.2,0.04854015827178955
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,1,power_law_1.2,2.1924589538574217
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,1,power_law_1.2,0.052910718917846676
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,1,power_law_1.2,0.07173952102661132
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,1,power_law_1.2,0.09661952018737793
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,1,power_law_1.2,0.12628864288330077
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,1,power_law_1.2,3.0855007934570313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,1,power_law_1.2,0.22244352340698242
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,1,power_law_1.2,0.24222591400146481
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,1,power_law_1.2,0.24939392089843748
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,1,power_law_1.2,0.531915512084961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,1,1,power_law_1.2,14.543514404296875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,1,power_law_1.2,0.2569241523742676
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,1,power_law_1.2,0.2591360092163086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,1,power_law_1.2,3.977422180175781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,1,power_law_1.2,0.2676697540283203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,1,power_law_1.2,0.2783379173278809
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,1,power_law_1.2,0.28255744934082033
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,1,power_law_1.2,0.32248382568359374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,1,power_law_1.2,0.353070068359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,1,1,power_law_1.2,5.068005676269531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,1,power_law_1.2,0.31180864334106445
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,1,power_law_1.2,0.41647617340087895
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.2,0.7888703918457031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,1,power_law_1.2,0.2975641632080078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.2,1.1143500518798828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.2,0.4851398468017578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.2,0.6378502273559571
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.2,1.4399008178710937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,1,power_law_1.01,0.16557632446289064
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,1,power_law_1.01,0.28083391189575196
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,1,power_law_1.01,0.3939468765258789
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,1,power_law_1.01,0.7274066925048828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,1,power_law_1.01,1.040860824584961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,1,power_law_1.01,1.821278076171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,1,power_law_1.01,1.9969158935546876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,1,power_law_1.01,2.1086778259277343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,1,power_law_1.01,2.2095391845703127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,1,power_law_1.01,2.316392364501953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,1,power_law_1.01,2.509488067626953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,1,power_law_1.01,2.7793356323242184
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,1,power_law_1.01,2.575537414550781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,1,power_law_1.01,2.628465270996094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,1,power_law_1.01,2.722437744140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,1,power_law_1.01,2.798671875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,1,power_law_1.01,3.0029278564453126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,1,power_law_1.01,3.485866088867188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,1,power_law_1.01,3.876837158203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,1,power_law_1.01,4.631688232421875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.2,2.095417022705078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,1,power_law_1.01,5.637159423828125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.2,2.748057556152344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.2,4.050305786132812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,1,power_law_1.01,7.398225708007812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.2,5.3887481689453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.2,6.748732910156249
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,1,power_law_1.01,9.017156372070312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,1,power_law_1.01,0.1900601577758789
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,1,power_law_1.01,0.25831552505493166
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,1,power_law_1.01,0.41147712707519535
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,1,power_law_1.01,0.5593715286254882
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,1,power_law_1.01,0.9046284484863282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,1,power_law_1.01,1.0071347045898438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,1,power_law_1.01,1.0409069061279297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,1,power_law_1.01,1.0789472198486327
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,1,power_law_1.01,12.561737060546875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,1,power_law_1.01,1.115675506591797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,1,power_law_1.01,0.15233920097351075
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,1,power_law_1.01,1.1355443572998047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,1,power_law_1.01,1.160258560180664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,1,power_law_1.01,1.2751084899902343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,1,power_law_1.01,1.2952435302734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,1,power_law_1.01,1.3382981872558593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,1,power_law_1.01,1.3418234252929688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,1,power_law_1.01,16.008797607421876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,1,power_law_1.01,1.4368666076660157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,1,power_law_1.01,1.5483334350585938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,1,power_law_1.01,1.6218194580078127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,1,power_law_1.01,1.8053875732421876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,1,power_law_1.01,2.12927490234375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,1,power_law_1.01,2.575836791992187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,1,power_law_1.01,3.0403866577148437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,1,power_law_1.01,4.049006042480469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,1,power_law_1.01,22.988520507812503
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,1,power_law_1.01,4.947651062011719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,1,power_law_1.01,7.042235107421876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,1,power_law_1.01,9.002687988281249
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,1,power_law_1.01,30.1050537109375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,1,1,power_law_1.01,10.895750732421876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,1,1,power_law_1.01,37.462431640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,1,power_law_1.01,0.07766272068023682
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,1,power_law_1.01,0.10076671600341798
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,1,power_law_1.01,0.6433254241943359
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,1,power_law_1.01,0.13095999717712403
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,1,power_law_1.01,0.6552806091308594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,1,power_law_1.01,0.6757913970947266
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,1,power_law_1.01,0.698050537109375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,1,power_law_1.01,0.22488319396972659
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,1,power_law_1.01,0.7049855804443359
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,1,power_law_1.01,0.7194982147216796
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,1,power_law_1.01,0.746397476196289
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,1,power_law_1.01,0.7654144287109375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,1,power_law_1.01,0.3144710350036621
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,1,power_law_1.01,0.5841702270507813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,1,power_law_1.01,0.5347328186035156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,1,power_law_1.01,0.616632308959961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,1,power_law_1.01,0.8263270568847656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.01,1.0103852844238281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,1,power_law_1.01,0.9255744171142577
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.01,1.1855782318115236
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.01,1.4529248046875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.01,1.8159429931640623
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,1,power_law_1.2,0.16662912368774413
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.01,2.246111297607422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,1,power_law_1.2,0.1942732810974121
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,1,power_law_1.2,0.38255104064941403
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,1,power_law_1.2,0.6398880004882812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,1,power_law_1.2,0.9104383850097657
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,1,power_law_1.2,1.7877017211914061
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,1,power_law_1.2,1.9915525817871091
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,1,power_law_1.2,2.161768341064453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,1,power_law_1.2,2.311267852783203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,1,power_law_1.2,2.4614930725097657
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.01,3.1541842651367187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,1,power_law_1.2,2.6557400512695315
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,1,power_law_1.2,2.833709411621094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,1,power_law_1.2,2.6794558715820314
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,1,power_law_1.2,2.743028564453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,1,power_law_1.2,2.8830810546875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,1,power_law_1.2,2.9578100585937497
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,1,power_law_1.2,3.15026123046875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,1,power_law_1.2,3.5517880249023435
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,1,power_law_1.2,3.9105874633789064
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,1,power_law_1.2,4.751537780761718
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,1,power_law_1.2,5.660020141601562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.01,3.993291015625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,1,power_law_1.2,7.498614501953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,1,power_law_1.2,9.1368896484375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.01,5.825133056640626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.01,7.668592529296875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,1,power_law_1.2,0.15959168434143067
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,1,power_law_1.2,0.2469183921813965
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,1,power_law_1.2,12.5735595703125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,1,1,power_law_1.01,9.593250732421875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,1,power_law_1.2,0.37081409454345704
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,1,power_law_1.2,0.4933260726928711
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,1,power_law_1.2,0.9076812744140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,1,power_law_1.2,0.9833990478515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,1,power_law_1.2,1.068370590209961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,1,power_law_1.2,1.104603500366211
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,1,power_law_1.2,1.1601606750488282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,1,power_law_1.2,1.1239686584472657
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,1,power_law_1.2,0.15233407974243166
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,1,power_law_1.2,16.374669189453122
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,1,power_law_1.2,1.1833337402343749
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,1,power_law_1.2,1.3399853515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,1,power_law_1.2,1.3763116455078124
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,1,power_law_1.2,1.3957688903808594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,1,power_law_1.2,1.4264524841308595
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,1,power_law_1.2,1.5225433349609374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,1,power_law_1.2,1.5781336975097655
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,1,power_law_1.2,1.7175225830078127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,1,power_law_1.2,1.9294918823242189
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,1,power_law_1.2,2.1707846069335934
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,1,power_law_1.2,2.6206597900390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,1,power_law_1.2,3.0955078125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,1,power_law_1.2,4.058334655761719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,1,power_law_1.2,23.77211669921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,1,power_law_1.2,5.012870483398437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,1,power_law_1.2,0.07834176063537598
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,1,power_law_1.2,0.08114751815795898
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,1,power_law_1.2,0.1292268753051758
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,1,power_law_1.2,0.1991961669921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,1,power_law_1.2,0.27596479415893554
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,1,power_law_1.2,7.103818359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,1,power_law_1.2,0.5305344009399414
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,1,power_law_1.2,0.5862963104248047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,1,power_law_1.2,0.6347705459594726
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,1,power_law_1.2,0.6686937713623047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,1,power_law_1.2,0.6769017791748047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,1,power_law_1.2,0.7030540466308594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,1,power_law_1.2,9.110542602539063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,1,power_law_1.2,0.7232044982910156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,1,power_law_1.2,0.7559149169921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,1,power_law_1.2,0.772419204711914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,1,power_law_1.2,31.480405273437498
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,1,1,power_law_1.2,11.0658984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,1,1,power_law_1.2,39.0372265625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,1,power_law_1.2,0.7931468963623047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,1,power_law_1.2,0.8716313934326172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,1,power_law_1.2,1.0017881774902344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.2,1.1154649353027344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.2,1.3827430725097656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,1,power_law_1.2,0.8202169799804688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.2,1.748524169921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,1,power_law_1.01,0.1968729591369629
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,1,power_law_1.01,0.3422943878173828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,1,power_law_1.01,0.6440799713134766
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,1,power_law_1.01,1.4273805236816406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.2,2.2583155822753906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,1,power_law_1.01,2.1082598876953123
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,1,power_law_1.01,2.477205047607422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,1,power_law_1.01,3.369740295410156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,1,power_law_1.01,3.6116229248046876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,1,power_law_1.01,3.76297607421875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,1,power_law_1.01,3.92910400390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.2,2.8818276977539066
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,1,power_law_1.01,4.074141540527344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.2,4.229274291992188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.2,5.593871459960938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,1,power_law_1.01,4.336824340820312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,1,power_law_1.01,4.660008850097656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,1,power_law_1.01,5.1960034179687495
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.2,8.357451171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,1,power_law_1.01,5.476349487304687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,1,power_law_1.01,4.988700866699219
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.2,11.184588623046874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,1,power_law_1.01,5.420536499023437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,1,power_law_1.01,5.850765380859375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,1,power_law_1.01,6.727642211914062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,1,power_law_1.01,7.6012463378906245
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,1,power_law_1.01,9.195670166015626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,1,power_law_1.01,0.23397504806518557
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,1,power_law_1.01,10.824887695312501
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,1,power_law_1.01,0.29021568298339845
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,1,power_law_1.01,0.39790782928466795
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,1,power_law_1.01,14.109194335937499
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,1,power_law_1.01,0.45801216125488275
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,1,power_law_1.01,0.8134496307373047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,1,power_law_1.01,1.1232704162597655
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,1,power_law_1.01,17.345667724609378
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,1,power_law_1.01,1.320938262939453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,1,power_law_1.01,1.709423370361328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,1,power_law_1.01,1.831321563720703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,1,power_law_1.01,1.8979519653320314
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,1,power_law_1.01,1.9694668579101564
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,1,power_law_1.01,2.0340249633789065
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,1,power_law_1.01,2.126739196777344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,1,power_law_1.01,2.1717178344726564
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,1,power_law_1.01,2.231396484375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,1,power_law_1.01,23.86792236328125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,1,power_law_1.01,2.2923545837402344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,1,power_law_1.01,2.5861862182617186
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,1,power_law_1.01,2.654352722167969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,1,power_law_1.01,2.73226318359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,1,power_law_1.01,3.0494720458984377
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,1,power_law_1.01,3.2739801025390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,1,power_law_1.01,3.737994384765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,1,power_law_1.01,4.227290344238281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,1,power_law_1.01,30.17588134765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,1,power_law_1.01,5.108427429199219
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,1,power_law_1.01,6.206365966796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,1,power_law_1.01,8.149165649414062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,1,1,power_law_1.01,36.914033203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,1,power_law_1.01,9.991764526367188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,1,1,power_law_1.01,11.878111572265626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,1,power_law_1.01,0.10732159614562989
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,1,power_law_1.01,0.1076255989074707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,1,power_law_1.01,0.1730419158935547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,1,power_law_1.01,0.18672767639160157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,1,power_law_1.01,0.2136774444580078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,1,power_law_1.01,0.10883520126342774
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,1,power_law_1.01,0.11177408218383789
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,1,power_law_1.01,0.11812992095947265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,1,power_law_1.01,0.14541760444641114
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,1,power_law_1.01,0.13121727943420408
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,1,power_law_1.01,0.15996416091918947
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,1,power_law_1.01,0.2669427108764649
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,1,power_law_1.01,0.3525632095336914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,1,power_law_1.01,0.42819904327392583
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,1,power_law_1.01,0.29449087142944336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,1,power_law_1.01,0.4859020614624024
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,1,power_law_1.01,0.6140991973876953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,1,power_law_1.01,0.8831110382080078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.01,1.1551910400390626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,1,power_law_1.2,0.41203968048095707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,1,power_law_1.2,0.4119615936279297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,1,power_law_1.2,0.6497222137451172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,1,power_law_1.2,1.2138105773925782
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.01,1.6922463989257814
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,1,power_law_1.2,1.8758035278320313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.01,2.2228314208984377
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,1,power_law_1.2,2.1914022827148436
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,1,power_law_1.2,3.2070413208007813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,1,power_law_1.2,3.391080932617187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.01,3.2926956176757813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.01,4.371285095214843
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,1,power_law_1.2,3.5752935791015625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,1,power_law_1.2,3.821013793945313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.01,6.534503784179688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,1,power_law_1.2,4.0875851440429685
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,1,power_law_1.2,4.262547302246094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,1,power_law_1.2,4.662357177734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,1,power_law_1.2,5.208806762695312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,1,power_law_1.2,5.5877050781249995
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,1,power_law_1.2,5.528523559570313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,1,power_law_1.2,5.986322631835938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,1,power_law_1.2,5.060957946777344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,1,power_law_1.2,6.879227905273437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,1,power_law_1.2,7.970857543945312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,1,power_law_1.2,9.694305419921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,1,power_law_1.2,11.386014404296876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,1,power_law_1.2,14.503937988281251
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,1,power_law_1.2,0.29067455291748046
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,1,power_law_1.2,17.820699462890623
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,1,power_law_1.2,0.3418374252319336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,1,power_law_1.2,0.4569977569580078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,1,power_law_1.2,0.7215744018554687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,1,power_law_1.2,1.0129817962646483
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,1,power_law_1.2,0.2336659240722656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,1,power_law_1.2,1.187056655883789
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,1,power_law_1.2,24.10081787109375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,1,power_law_1.2,1.6389414978027343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,1,power_law_1.2,1.7406098937988284
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,1,power_law_1.2,1.8299154663085937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,1,power_law_1.2,1.9264710998535157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,1,power_law_1.2,1.9828282165527344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,1,power_law_1.2,2.039177551269531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,1,power_law_1.2,2.108837738037109
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,1,power_law_1.2,30.863234863281253
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,1,power_law_1.2,2.200084533691406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,1,power_law_1.2,2.276111297607422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,1,power_law_1.2,2.575689697265625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,1,power_law_1.2,2.780181884765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,1,power_law_1.2,3.17515380859375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,1,power_law_1.2,3.4373861694335934
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,1,power_law_1.2,4.464874877929687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,1,1,power_law_1.2,37.409453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,1,power_law_1.2,2.8959686279296877
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,1,power_law_1.2,5.4325634765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,1,power_law_1.2,3.9338891601562502
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,1,power_law_1.2,6.444166259765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,1,power_law_1.2,0.10697407722473144
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,1,power_law_1.2,0.10869567871093751
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,1,power_law_1.2,0.1315283203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,1,power_law_1.2,0.11880640029907226
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,1,power_law_1.2,8.347315673828124
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,1,power_law_1.2,0.14598143577575684
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,1,power_law_1.2,0.15905088424682617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,1,power_law_1.2,0.17314624786376953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,1,power_law_1.2,10.104761352539063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,1,power_law_1.2,0.18715328216552735
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,1,power_law_1.2,0.21336639404296873
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,1,power_law_1.2,0.26627775192260744
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,1,power_law_1.2,0.35260414123535155
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,1,1,power_law_1.2,12.0297509765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,1,power_law_1.2,0.10554112434387206
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,1,power_law_1.2,0.11007103919982911
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,1,power_law_1.2,0.2955615997314453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,1,power_law_1.2,0.42901695251464844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,1,power_law_1.2,0.4853196716308593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,1,power_law_1.2,0.6153855895996093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,1,power_law_1.2,0.8865529632568359
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.2,1.1554764556884767
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.2,3.293680114746094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.2,1.681084747314453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,1,power_law_1.01,0.2651315116882324
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,1,power_law_1.01,0.317137279510498
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.2,2.225183410644531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,1,power_law_1.01,0.3197088050842285
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,1,power_law_1.01,0.3194156837463379
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,1,power_law_1.01,0.32258239746093753
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,1,power_law_1.01,0.330511360168457
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,1,power_law_1.01,0.3520985412597656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,1,power_law_1.01,0.36967041015625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.2,4.376618957519531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,1,power_law_1.01,0.3739168167114258
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,1,power_law_1.01,0.3889689636230469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,1,power_law_1.01,0.4728211212158203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,1,power_law_1.01,0.4269644927978516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,1,power_law_1.01,0.5384249496459961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.2,6.517180175781251
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,1,power_law_1.01,0.6272665786743163
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,1,power_law_1.01,0.8274022674560546
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,1,power_law_1.01,0.9894841766357422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,1,power_law_1.01,1.3893843078613282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,1,power_law_1.01,1.7958348083496094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,1,power_law_1.01,2.6422073364257814
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,1,power_law_1.01,3.53079345703125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,1,power_law_1.01,5.224208984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,1,power_law_1.01,6.9147998046875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,1,power_law_1.01,0.0705625581741333
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,1,power_law_1.01,0.09305343627929688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,1,power_law_1.01,0.10091456413269043
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,1,power_law_1.01,11.151912841796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,1,power_law_1.01,0.16753664016723632
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,1,power_law_1.01,0.16694528579711915
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,1,power_law_1.01,0.1668454360961914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,1,power_law_1.01,0.1481651210784912
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,1,power_law_1.01,0.16470399856567383
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,1,power_law_1.01,0.16655487060546875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,1,power_law_1.01,0.1728544044494629
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,1,power_law_1.01,0.17215232849121093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,1,power_law_1.01,0.1756800079345703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,1,power_law_1.01,0.17931776046752929
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,1,power_law_1.01,0.19971136093139646
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,1,power_law_1.01,0.20918336868286133
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,1,power_law_1.01,0.22221824645996097
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,1,power_law_1.01,0.2345337677001953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,1,power_law_1.01,0.2841740798950195
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,1,power_law_1.01,0.34450111389160154
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,1,power_law_1.01,0.4541632080078125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,1,power_law_1.01,15.671236572265624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,1,power_law_1.01,0.5541228866577148
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,1,power_law_1.01,0.7927072143554688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,1,power_law_1.01,1.0485906982421875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,1,power_law_1.01,1.4955039978027345
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,1,power_law_1.01,1.997480926513672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,1,power_law_1.01,2.9721588134765624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,1,power_law_1.01,3.8707513427734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,2,1,power_law_1.01,20.0117724609375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,2,1,power_law_1.01,4.8693652343750005
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,1,power_law_1.01,0.038419840335845945
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,1,power_law_1.01,0.09199104309082032
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,1,power_law_1.01,0.05053376197814942
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,1,power_law_1.01,0.0903001594543457
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,1,power_law_1.01,0.0912831974029541
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,1,power_law_1.01,0.09288960456848144
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,1,power_law_1.01,0.09386816024780273
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,1,power_law_1.01,0.06099199771881103
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,1,power_law_1.01,0.08946304321289063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,1,power_law_1.01,0.08958656311035156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,1,power_law_1.01,0.09039936065673829
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,1,power_law_1.01,0.09592063903808594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,1,power_law_1.01,0.10250240325927736
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,1,power_law_1.01,0.09308863639831542
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,1,power_law_1.01,0.10535296440124511
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,1,power_law_1.01,0.10954815864562988
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,1,power_law_1.01,0.12130175590515138
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,1,power_law_1.01,0.1504697608947754
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.01,0.17972864151000975
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.01,0.23747007369995118
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.01,0.29422975540161134
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,1,power_law_1.2,0.12485055923461914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.01,0.4138848114013672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,1,power_law_1.2,0.14678336143493653
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,1,power_law_1.2,0.17960895538330077
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,1,power_law_1.2,0.2706163215637207
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,1,power_law_1.2,0.31216255187988284
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,1,power_law_1.2,0.3205759811401367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,1,power_law_1.2,0.3207833480834961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,1,power_law_1.2,0.32108863830566403
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,1,power_law_1.2,0.3323686218261719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,1,power_law_1.2,0.3541670227050781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,1,power_law_1.2,0.37214401245117185
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,1,power_law_1.2,0.3794681549072266
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,1,power_law_1.2,0.38322624206542966
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,1,power_law_1.2,0.42051841735839846
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,1,power_law_1.2,0.4660665512084961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,1,power_law_1.2,0.5573043060302735
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,1,power_law_1.2,0.6261062240600586
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,1,power_law_1.2,0.8102079772949219
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,1,power_law_1.2,1.0030828857421876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,1,power_law_1.2,1.4037362670898437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.01,0.543377914428711
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,1,power_law_1.2,1.8761241149902343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,1,power_law_1.2,2.69927490234375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,1,power_law_1.2,3.4523678588867184
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,1,power_law_1.2,5.179437255859375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,1,power_law_1.2,6.9453540039062505
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.01,0.7859827423095703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.01,1.0407532501220704
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,1,power_law_1.2,0.06699776172637939
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,1,power_law_1.2,0.09859328269958496
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,1,power_law_1.2,0.15066304206848144
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,1,power_law_1.2,0.1559552001953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,1,power_law_1.2,11.42802490234375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,1,power_law_1.2,0.16622655868530273
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,1,power_law_1.2,0.17207296371459962
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,1,power_law_1.2,0.16373056411743164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,1,power_law_1.2,0.17321279525756836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,1,power_law_1.2,0.08731904029846191
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,1,power_law_1.2,0.17330623626708985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,1,power_law_1.2,0.175545597076416
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,1,power_law_1.2,0.17756032943725586
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,1,power_law_1.2,0.17845823287963866
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,1,power_law_1.2,0.1987718391418457
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,1,power_law_1.2,0.21568063735961912
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,1,power_law_1.2,0.2291360092163086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,1,power_law_1.2,0.23512960433959962
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,1,power_law_1.2,0.282744312286377
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,1,power_law_1.2,0.34568511962890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,1,power_law_1.2,0.4542208099365235
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,1,power_law_1.2,0.5537996673583985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,1,power_law_1.2,0.7893452453613281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,1,power_law_1.2,0.9912051391601562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,1,power_law_1.2,1.5082342529296875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,1,power_law_1.2,1.992615661621094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,1,power_law_1.2,16.005899658203127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,1,power_law_1.2,3.0056985473632816
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,1,power_law_1.2,0.038870398998260496
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,1,power_law_1.2,0.056201601028442384
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.01,1.610174102783203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,1,power_law_1.2,0.08957119941711425
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,1,power_law_1.2,0.09095040321350098
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,1,power_law_1.2,0.09169535636901856
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.01,2.1743756103515626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,1,power_law_1.2,0.09480064392089844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,1,power_law_1.2,0.09052736282348632
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,1,power_law_1.2,0.09738176345825196
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,1,power_law_1.2,0.09397248268127442
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,1,power_law_1.2,0.09378560066223145
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.01,2.686861572265625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,1,power_law_1.2,3.911904602050781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,1,power_law_1.2,0.10266048431396484
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,1,power_law_1.2,0.12234111785888672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,2,1,power_law_1.2,4.9593395996093745
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,2,1,power_law_1.2,20.021146240234376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,1,power_law_1.2,0.09507455825805664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,1,power_law_1.2,0.04724607944488525
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,1,power_law_1.2,0.09470208168029785
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,1,power_law_1.2,0.12459903717041017
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,1,power_law_1.2,0.10863615989685058
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,1,power_law_1.2,0.15467840194702148
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.2,0.1832921600341797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.2,0.23978111267089847
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.2,0.29918848037719725
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.2,0.4197267150878906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.2,0.5522751998901367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,1,power_law_1.01,0.23321151733398438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,1,power_law_1.01,0.4424281692504882
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.2,1.0880147552490234
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,1,power_law_1.01,0.5381715011596679
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,1,power_law_1.01,0.5522988891601562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.2,0.825934066772461
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,1,power_law_1.01,0.5557683181762696
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.2,1.6745152282714844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,1,power_law_1.01,0.5474367904663086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,1,power_law_1.01,0.1661881637573242
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,1,power_law_1.01,0.5740703964233399
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,1,power_law_1.01,0.566099853515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,1,power_law_1.01,0.29943103790283204
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,1,power_law_1.01,0.6238694381713867
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,1,power_law_1.01,0.6359283065795898
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,1,power_law_1.01,0.684181137084961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,1,power_law_1.01,0.740723876953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,1,power_law_1.01,0.7945645141601563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,1,power_law_1.01,0.8396339416503906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,1,power_law_1.01,1.0307481384277344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,1,power_law_1.01,1.3858758544921874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,1,power_law_1.01,1.607749786376953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.2,2.1720594787597656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,1,power_law_1.01,2.4576722717285158
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,1,power_law_1.01,3.1331130981445314
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,1,power_law_1.01,4.525360107421875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,1,power_law_1.01,0.09620800018310546
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.2,2.732655334472656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,1,power_law_1.01,0.12681983947753905
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,1,power_law_1.01,0.15473983764648438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,1,power_law_1.01,0.2386739158630371
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,1,power_law_1.01,0.258474235534668
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,1,power_law_1.01,5.941649780273438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,1,power_law_1.01,0.2573638343811035
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,1,power_law_1.01,0.25976064682006833
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,1,power_law_1.01,0.27508863449096677
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,1,power_law_1.01,0.26249984741210936
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,1,power_law_1.01,0.26692991256713866
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,1,power_law_1.01,0.2722681617736816
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,1,power_law_1.01,0.2730035209655762
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,1,power_law_1.01,0.27981760025024416
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,1,power_law_1.01,0.29331647872924804
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,1,power_law_1.01,0.2997907257080078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,1,power_law_1.01,0.31526975631713866
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,1,power_law_1.01,0.35190078735351565
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,1,power_law_1.01,0.4285996627807617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,1,power_law_1.01,0.5135942459106445
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,1,power_law_1.01,0.6763692474365234
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,1,power_law_1.01,8.734318237304688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,1,power_law_1.01,0.889560317993164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,1,power_law_1.01,1.1880499267578126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,1,power_law_1.01,1.580999755859375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,1,power_law_1.01,2.4102330017089844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,1,power_law_1.01,3.059788818359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,1,power_law_1.01,11.85872802734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,1,power_law_1.01,4.547616577148437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,1,power_law_1.01,5.994833984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,2,1,power_law_1.01,7.569625244140624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,1,power_law_1.01,19.096387939453127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,1,power_law_1.01,27.584702148437497
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,2,1,power_law_1.01,35.9906591796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,1,power_law_1.01,0.06993152141571045
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,1,power_law_1.01,0.13294079780578613
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,1,power_law_1.01,0.04798272132873535
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,1,power_law_1.01,0.13470144271850587
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,1,power_law_1.01,0.13591487884521486
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,1,power_law_1.01,0.0859494400024414
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,1,power_law_1.01,0.13187199592590332
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,1,power_law_1.01,0.13280832290649414
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,1,power_law_1.01,0.13427200317382812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,1,power_law_1.01,0.13372032165527342
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,1,power_law_1.01,0.13232128143310545
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,1,power_law_1.01,0.1532479953765869
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,1,power_law_1.01,0.13624640464782717
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,1,power_law_1.01,0.14017024040222167
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,1,power_law_1.01,0.15804863929748536
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,1,power_law_1.01,0.1670137596130371
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,1,power_law_1.01,0.19144960403442385
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,1,power_law_1.01,0.2412735939025879
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.01,0.2842515182495117
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.01,0.473034896850586
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.01,0.6899616241455078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.01,0.374648323059082
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,1,power_law_1.2,0.18954111099243165
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,1,power_law_1.2,0.22118976593017575
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,1,power_law_1.2,0.2801478385925293
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,1,power_law_1.2,0.43977729797363285
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,1,power_law_1.2,0.5018092727661132
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,1,power_law_1.2,0.5460134506225586
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,1,power_law_1.2,0.5553094482421874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,1,power_law_1.2,0.5441439819335938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,1,power_law_1.2,0.5632447814941406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,1,power_law_1.2,0.561952018737793
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,1,power_law_1.2,0.6142681503295899
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,1,power_law_1.2,0.6566719818115234
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,1,power_law_1.2,0.6700665283203124
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,1,power_law_1.2,0.7416812896728515
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.01,0.8877375793457031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,1,power_law_1.2,0.7941951751708984
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,1,power_law_1.2,0.8741516876220704
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,1,power_law_1.2,1.0097497558593749
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,1,power_law_1.2,1.354285430908203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,1,power_law_1.2,1.7613734436035156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.01,1.3385580444335938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,1,power_law_1.2,2.390277099609375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.01,1.8148768615722655
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,1,power_law_1.2,3.1552102661132815
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,1,power_law_1.2,4.5112646484375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,1,power_law_1.2,5.9486413574218755
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.01,2.680445556640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,1,power_law_1.2,0.09803647994995117
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,1,power_law_1.2,0.12475456237792967
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,1,power_law_1.2,0.1368166446685791
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,1,power_law_1.2,0.2401804733276367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,1,power_law_1.2,0.2523046493530273
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,1,power_law_1.2,8.755675659179687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,1,power_law_1.2,0.260383358001709
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,1,power_law_1.2,0.2764703941345215
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,1,power_law_1.2,0.2601600074768066
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,1,power_law_1.2,0.26848703384399414
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,1,power_law_1.2,0.26805440902709965
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,1,power_law_1.2,0.27177024841308595
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,1,power_law_1.2,0.2757472038269043
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,1,power_law_1.2,0.2742803192138672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,1,power_law_1.2,0.29334144592285155
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,1,power_law_1.2,0.30718528747558593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,1,power_law_1.2,0.32076160430908207
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,1,power_law_1.2,0.3580390548706055
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,1,power_law_1.2,0.42837375640869146
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,1,power_law_1.2,12.05921630859375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,1,power_law_1.2,0.5143308639526367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,1,power_law_1.2,0.6722227478027344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,1,power_law_1.2,0.8521766662597656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,1,power_law_1.2,1.2123353576660156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,1,power_law_1.2,1.6062521362304687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,1,power_law_1.2,2.4229702758789062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,1,power_law_1.2,3.0603955078125002
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.01,3.4373477172851565
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,1,power_law_1.2,0.052771840095520016
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,1,power_law_1.2,0.06280191898345946
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,1,power_law_1.2,4.529888610839843
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,1,power_law_1.2,0.08072575569152832
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,1,power_law_1.2,0.13720767974853515
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,1,power_law_1.2,0.13057408332824708
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,1,power_law_1.2,0.1356262397766113
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,1,power_law_1.2,0.13317119598388671
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,1,power_law_1.2,0.13462656021118163
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,1,power_law_1.2,19.53856201171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,1,power_law_1.2,0.1341600036621094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,1,power_law_1.2,0.13466815948486327
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,1,power_law_1.2,0.1391487979888916
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,1,power_law_1.2,0.154903039932251
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,1,power_law_1.2,6.018969116210938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,2,1,power_law_1.2,7.495205688476562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.01,4.361024780273437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,1,power_law_1.2,28.134697265625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,2,1,power_law_1.2,38.04463134765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,1,power_law_1.2,0.13752384185791017
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,1,power_law_1.2,0.14031807899475096
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,1,power_law_1.2,0.16229568481445314
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,1,power_law_1.2,0.17493696212768556
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,1,power_law_1.2,0.19493120193481445
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,1,power_law_1.2,0.24180416107177732
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.2,0.2852377510070801
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.2,0.37468097686767576
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.2,0.4679097747802734
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.2,0.7040537261962891
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.2,0.8919500732421876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.2,1.314217529296875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.2,1.7962361145019532
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,1,power_law_1.01,0.12865920066833497
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,1,power_law_1.01,0.6461081695556641
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,1,power_law_1.01,0.17855104446411132
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,1,power_law_1.01,0.20446144104003908
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,1,power_law_1.01,1.2734041595458985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.2,2.650940246582031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,1,power_law_1.01,1.332422332763672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,1,power_law_1.01,0.40915584564208984
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,1,power_law_1.01,0.8784223937988281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,1,power_law_1.01,1.368802490234375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,1,power_law_1.01,1.397022705078125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.2,3.5218945312499996
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,1,power_law_1.01,1.6469529724121095
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,1,power_law_1.01,1.4936659240722656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,1,power_law_1.01,1.7011724853515624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,1,power_law_1.01,1.8229408264160156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,1,power_law_1.01,1.8550469970703127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,1,power_law_1.01,1.8995948791503907
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,1,power_law_1.01,1.9449746704101565
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,1,power_law_1.01,2.1042092895507816
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,1,power_law_1.01,2.2911750793457033
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,1,power_law_1.01,2.7356378173828126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,1,power_law_1.01,3.1100180053710935
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,1,power_law_1.01,3.9221874999999997
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,1,power_law_1.01,4.6935577392578125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,1,power_law_1.01,6.507133178710937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,1,power_law_1.01,0.1281382369995117
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,1,power_law_1.01,0.16435199737548828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,1,power_law_1.01,0.2747616004943848
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,1,power_law_1.01,0.39539840698242185
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,1,power_law_1.01,8.144782104492188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.2,4.419197998046875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,1,power_law_1.01,0.5161318588256836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,1,power_law_1.01,11.195090332031251
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,1,power_law_1.01,0.6942585754394531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,1,power_law_1.01,0.1831705665588379
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,1,power_law_1.01,0.7199654388427734
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,1,power_law_1.01,0.7673792266845704
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,1,power_law_1.01,0.7994918060302735
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,1,power_law_1.01,0.7441702270507813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,1,power_law_1.01,14.572047119140626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,1,power_law_1.01,0.8319519805908204
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,1,power_law_1.01,0.8489356994628906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,1,power_law_1.01,0.8720812988281249
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,1,power_law_1.01,0.9745651245117187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,1,power_law_1.01,0.9888896179199218
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,1,power_law_1.01,1.0127244567871094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,1,power_law_1.01,1.0765331268310547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,1,power_law_1.01,1.12308349609375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,1,power_law_1.01,1.2926611328125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,1,power_law_1.01,1.4215846252441406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,2,1,power_law_1.01,17.83622802734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,1,power_law_1.01,1.67451904296875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,1,power_law_1.01,1.9184646606445312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,1,power_law_1.01,2.50042236328125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,1,power_law_1.01,3.0465625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,1,power_law_1.01,4.096756591796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,1,power_law_1.01,5.173133544921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,2,1,power_law_1.01,6.360780029296874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,1,power_law_1.01,0.07013760089874269
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,1,power_law_1.01,0.09743680000305176
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,1,power_law_1.01,0.08856320381164551
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,1,power_law_1.01,0.14636672019958497
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,1,power_law_1.01,0.4550995254516602
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,1,power_law_1.01,0.22116928100585936
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,1,power_law_1.01,0.46657791137695315
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,1,power_law_1.01,0.489153938293457
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,1,power_law_1.01,0.29275007247924806
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,1,power_law_1.01,0.5092492675781249
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,1,power_law_1.01,0.5236172866821289
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,1,power_law_1.01,0.5440531158447266
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,1,power_law_1.01,0.5630419158935547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,1,power_law_1.01,0.577966079711914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,1,power_law_1.01,0.6100249481201172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,1,power_law_1.01,0.43700542449951174
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,1,power_law_1.01,0.4178476715087891
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,1,power_law_1.01,0.6743084716796874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.01,0.7314380645751953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.01,0.9636268615722656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.01,0.8641350555419922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.01,1.530059509277344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.01,1.2397676849365236
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,1,power_law_1.2,0.17293312072753905
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,1,power_law_1.2,0.18725183486938476
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,1,power_law_1.2,0.3841030502319336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.01,2.116774444580078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,1,power_law_1.2,0.5713203048706055
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,1,power_law_1.2,0.126112642288208
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.01,2.7094137573242185
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,1,power_law_1.2,0.7781394958496094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,1,power_law_1.2,1.2346355438232421
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,1,power_law_1.2,1.307523193359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,1,power_law_1.2,1.3597702026367187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,1,power_law_1.2,1.4039295959472657
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,1,power_law_1.2,1.5050968933105469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,1,power_law_1.2,1.6539173889160157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,1,power_law_1.2,1.7619520568847655
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,1,power_law_1.2,1.9773683166503908
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,1,power_law_1.2,1.872202911376953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,1,power_law_1.2,1.9300991821289064
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,1,power_law_1.2,2.0509088134765627
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,1,power_law_1.2,2.2510861206054686
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,1,power_law_1.2,2.4806072998046877
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,1,power_law_1.2,2.8776632690429684
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,1,power_law_1.2,3.263194274902344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,1,power_law_1.2,4.043043823242188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,1,power_law_1.2,4.900120239257812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,1,power_law_1.2,6.609010009765624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,1,power_law_1.2,8.252168579101562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,1,power_law_1.2,0.1610361671447754
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,1,power_law_1.2,0.12877056121826172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,1,power_law_1.2,0.17435327529907227
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,1,power_law_1.2,0.35551742553710936
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,1,power_law_1.2,11.49591796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,1,power_law_1.2,0.2592563247680664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,1,power_law_1.2,0.46430206298828125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,1,power_law_1.2,0.6734214019775391
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,1,power_law_1.2,0.7123219299316406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.01,3.913510437011719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,1,power_law_1.2,0.7399987030029297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,1,power_law_1.2,0.7614924621582031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,1,power_law_1.2,0.7962911987304688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,1,power_law_1.2,0.8231590270996094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,1,power_law_1.2,0.8518508911132813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,1,power_law_1.2,0.8937964630126952
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,1,power_law_1.2,14.666370849609375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.01,5.2173791503906255
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,1,power_law_1.2,0.9833318328857421
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,2,1,power_law_1.01,6.408408203125001
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,1,power_law_1.2,1.0370687866210937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,1,power_law_1.2,1.0782899475097656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,1,power_law_1.2,1.1309158325195312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,1,power_law_1.2,1.2042700958251953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,1,power_law_1.2,1.3690226745605467
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,1,power_law_1.2,1.4736575317382812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,1,power_law_1.2,1.6856282043457032
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,1,power_law_1.2,2.0306719970703124
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,2,1,power_law_1.2,17.937711181640626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,1,power_law_1.2,2.540952911376953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,1,power_law_1.2,0.06926208019256593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,1,power_law_1.2,0.0882316780090332
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,1,power_law_1.2,3.139336853027344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,1,power_law_1.2,0.08743359565734862
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,1,power_law_1.2,0.1362335968017578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,1,power_law_1.2,0.19638591766357422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,1,power_law_1.2,4.192170104980469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,1,power_law_1.2,0.26491264343261717
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,1,power_law_1.2,0.4037184143066407
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,1,power_law_1.2,0.43131584167480475
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,1,power_law_1.2,0.44970943450927736
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,1,power_law_1.2,0.46396865844726565
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,1,power_law_1.2,5.269002075195313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,1,power_law_1.2,0.5158899307250977
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,1,power_law_1.2,0.4906739044189453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,1,power_law_1.2,0.5307302474975586
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,1,power_law_1.2,0.5670495986938476
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,2,1,power_law_1.2,6.398851318359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,1,power_law_1.2,0.5879347229003906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,1,power_law_1.2,0.6120230484008788
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,1,power_law_1.2,0.6593510437011718
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,1,power_law_1.2,0.7523833465576172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.2,0.827286376953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.2,1.0178643035888673
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.2,1.1962118530273438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.2,1.6454208374023438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,1,power_law_1.01,0.030432000160217288
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,1,power_law_1.01,0.05006207942962646
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,1,power_law_1.01,0.04128896236419678
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.2,2.0698016357421873
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,1,power_law_1.01,0.07719295978546142
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,1,power_law_1.01,0.11933695793151855
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,1,power_law_1.01,0.12737343788146974
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,1,power_law_1.01,0.1320556831359863
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.2,4.044343566894531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,1,power_law_1.01,0.1309657573699951
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,1,power_law_1.01,0.13988160133361816
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,1,power_law_1.01,0.14889599800109862
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.2,2.988065185546875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,1,power_law_1.01,0.15964223861694335
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,1,power_law_1.01,0.16274431228637695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,1,power_law_1.01,0.16981504440307618
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,1,power_law_1.01,0.15598527908325194
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,1,power_law_1.01,0.20218240737915041
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,1,power_law_1.01,0.21997568130493167
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,1,power_law_1.01,0.26847551345825194
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,1,power_law_1.01,0.3169926452636719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,1,power_law_1.01,0.4190208053588867
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,1,power_law_1.01,0.5243360137939452
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,1,power_law_1.01,0.7357453155517578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,1,power_law_1.01,0.9456396484375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,1,power_law_1.01,0.042720642089843754
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,1,power_law_1.01,1.3780992126464844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,1,power_law_1.01,0.04145472049713135
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,1,power_law_1.01,0.05044288158416747
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,1,power_law_1.01,1.7930201721191406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,1,power_law_1.01,0.09271743774414062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,1,power_law_1.01,0.10186880111694337
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,1,power_law_1.01,0.05854271888732911
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,1,power_law_1.01,0.103438081741333
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,1,power_law_1.01,0.07438848018646241
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,2,1,power_law_1.01,2.2281471252441407
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,1,power_law_1.01,0.10666239738464356
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,1,power_law_1.01,0.10697279930114748
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,1,power_law_1.01,0.10836799621582031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,1,power_law_1.01,0.12360960006713868
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,1,power_law_1.01,0.12413311958312989
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,1,power_law_1.01,0.12808575630187988
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,1,power_law_1.01,0.12870464324951172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,1,power_law_1.01,0.13072064399719238
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,1,power_law_1.01,0.14043456077575683
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,1,power_law_1.01,0.14922112464904785
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,1,power_law_1.01,0.16679424285888672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,1,power_law_1.01,0.1888863945007324
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,1,power_law_1.01,0.22229312896728515
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,1,power_law_1.01,0.2781702423095703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,1,power_law_1.01,0.3352716827392578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,1,power_law_1.01,0.45813438415527347
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,1,power_law_1.01,0.5783238220214844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,1,power_law_1.01,0.8215462493896485
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,1,power_law_1.01,1.0601849365234375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,2,1,power_law_1.01,1.3001043701171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,1,power_law_1.01,0.03459520101547241
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,1,power_law_1.01,0.038724479675292967
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,1,power_law_1.01,0.04236032009124756
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,1,power_law_1.01,0.06337152004241944
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,1,power_law_1.01,0.06750080108642578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,1,power_law_1.01,0.04650112152099609
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,1,power_law_1.01,0.05125120162963868
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,1,power_law_1.01,0.06845376014709473
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,1,power_law_1.01,0.06971263885498047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,1,power_law_1.01,0.07064640045166015
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,1,power_law_1.01,0.07172287940979004
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,1,power_law_1.01,0.0751091194152832
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,1,power_law_1.01,0.0791807985305786
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,1,power_law_1.01,0.07452799797058104
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,1,power_law_1.01,0.09449024200439453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,1,power_law_1.01,0.08361408233642578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,1,power_law_1.01,0.0864633560180664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,1,power_law_1.01,0.12531519889831544
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,1,power_law_1.01,0.10972479820251464
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,1,power_law_1.01,0.15934911727905274
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,1,power_law_1.01,0.19532480239868164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,1,power_law_1.01,0.2657414436340332
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,1,power_law_1.01,0.5033055877685546
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,1,power_law_1.01,0.3427660751342773
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,1,power_law_1.2,0.032116479873657226
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,1,power_law_1.2,0.04078976154327393
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,1,power_law_1.2,0.049374718666076664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,1,power_law_1.2,0.061662721633911136
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,1,power_law_1.2,0.0743827199935913
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,1,power_law_1.2,0.1213203239440918
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,1,power_law_1.2,0.12551232337951662
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,1,power_law_1.2,0.13014528274536133
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,1,power_law_1.2,0.1305401611328125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,1,power_law_1.2,0.13187264442443847
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,1,power_law_1.2,0.1414694404602051
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,1,power_law_1.2,0.15151103973388672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,1,power_law_1.2,0.1529587173461914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,1,power_law_1.2,0.15696767807006834
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,1,power_law_1.2,0.16111103057861328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,1,power_law_1.2,0.2041049575805664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,1,power_law_1.2,0.222608642578125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,1,power_law_1.2,0.2751897621154785
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,1,power_law_1.2,0.31813631057739256
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,1,power_law_1.2,0.42304000854492185
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,1,power_law_1.2,0.5291948699951172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,1,power_law_1.2,0.16019264221191407
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,1,power_law_1.2,0.1680659294128418
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,1,power_law_1.01,0.6666738891601562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,1,power_law_1.2,0.7419276428222656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,1,power_law_1.2,0.9497670745849609
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,1,power_law_1.2,1.3737606811523437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,1,power_law_1.01,1.0013785552978516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,1,power_law_1.01,1.3378924560546874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,1,power_law_1.2,1.8017855834960936
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,1,power_law_1.2,0.0496563196182251
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,1,power_law_1.2,0.05743360042572021
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,2,1,power_law_1.01,1.6863565063476564
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,1,power_law_1.2,0.03778304100036621
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,1,power_law_1.2,0.06740352153778076
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,1,power_law_1.2,0.09571007728576661
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,1,power_law_1.2,0.10456768035888672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,1,power_law_1.2,0.10526016235351562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,1,power_law_1.2,0.10710847854614258
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,1,power_law_1.2,0.10550784111022948
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,1,power_law_1.2,0.1073414421081543
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,2,1,power_law_1.2,2.2466348266601566
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,1,power_law_1.2,0.12500864028930664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,1,power_law_1.2,0.1275052833557129
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,1,power_law_1.2,0.12361023902893067
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,1,power_law_1.2,0.13123711585998535
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,1,power_law_1.2,0.1355020809173584
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,1,power_law_1.2,0.14124095916748047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,1,power_law_1.2,0.15424320220947266
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,1,power_law_1.2,0.16822463989257813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,1,power_law_1.2,0.19756032943725585
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,1,power_law_1.2,0.22399744033813476
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,1,power_law_1.2,0.27992000579833987
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,1,power_law_1.2,0.3432268905639648
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,1,power_law_1.2,0.45972736358642574
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,1,power_law_1.2,0.5837804794311523
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,1,power_law_1.2,0.8254061126708985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,1,power_law_1.2,0.03535552024841308
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,1,power_law_1.2,0.04156223773956299
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,1,power_law_1.2,1.065560302734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,2,1,power_law_1.2,1.302558135986328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,1,power_law_1.2,0.041866240501403806
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,1,power_law_1.2,0.04518400192260742
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,1,power_law_1.2,0.049687681198120115
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,1,power_law_1.2,0.06465343952178955
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,1,power_law_1.2,0.06800000190734863
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,1,power_law_1.2,0.06907008171081544
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,1,power_law_1.2,0.07285376071929932
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,1,power_law_1.2,0.07121407985687256
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,1,power_law_1.2,0.07680448055267333
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,1,power_law_1.2,0.0772985601425171
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,1,power_law_1.2,0.08258560180664062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,1,power_law_1.2,0.08820096015930176
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,1,power_law_1.2,0.09241087913513184
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,1,power_law_1.2,0.03903743982315063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,1,power_law_1.2,0.14778880119323728
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,1,power_law_1.2,0.25135040283203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,1,power_law_1.2,0.19936447143554686
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,1,power_law_1.2,0.07055615901947021
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,1,power_law_1.2,0.10211392402648926
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,1,power_law_1.2,0.12300352096557617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,1,power_law_1.01,0.060323839187622075
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,1,power_law_1.2,0.47218238830566406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,1,power_law_1.2,0.36171199798583986
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,1,power_law_1.2,0.7018745422363282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,1,power_law_1.01,0.17546432495117187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,1,power_law_1.01,0.34680255889892575
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,1,power_law_1.01,0.06865856170654297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,1,power_law_1.01,0.3736211013793945
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,1,power_law_1.01,0.10788607597351074
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,1,power_law_1.01,0.3882790374755859
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,1,power_law_1.01,0.40341758728027344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,1,power_law_1.2,0.9265644836425782
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,1,power_law_1.2,1.368076171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,2,1,power_law_1.2,2.27988037109375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,1,power_law_1.01,0.41194305419921873
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,1,power_law_1.01,0.4403071975708007
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,1,power_law_1.01,0.4475270462036133
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,1,power_law_1.01,0.448540153503418
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,1,power_law_1.01,0.4856179046630859
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,1,power_law_1.01,0.5060748672485351
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,1,power_law_1.01,0.5149184036254882
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,1,power_law_1.01,0.6049055862426758
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,1,power_law_1.01,0.6598918151855468
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,1,power_law_1.2,1.8222508239746094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,1,power_law_1.01,0.8129990386962891
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,1,power_law_1.01,0.9801497650146483
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,1,power_law_1.01,0.22644287109375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,1,power_law_1.01,0.5295622253417969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,1,power_law_1.01,1.3663737487792968
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,1,power_law_1.01,1.6853094482421873
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,1,power_law_1.01,0.05807487964630127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,1,power_law_1.01,2.5013702392578123
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,1,power_law_1.01,0.06827072143554688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,1,power_law_1.01,0.09726655960083008
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,1,power_law_1.01,3.1505453491210935
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,1,power_law_1.01,0.14950271606445314
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,1,power_law_1.01,0.21075967788696287
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,1,power_law_1.01,0.23360639572143554
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,1,power_law_1.01,0.24157119750976563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,1,power_law_1.01,0.2415551948547363
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,1,power_law_1.01,0.24440832138061525
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,1,power_law_1.01,0.25227008819580077
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,1,power_law_1.01,0.27897600173950193
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,1,power_law_1.01,0.2912499237060547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,1,power_law_1.01,0.3034169578552246
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,1,power_law_1.01,0.3010560035705566
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,1,power_law_1.01,0.31992000579833985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,1,power_law_1.01,0.3392531204223633
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,1,power_law_1.01,4.654649658203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,1,power_law_1.01,0.3692364883422852
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,1,power_law_1.01,0.4242278289794922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,1,power_law_1.01,0.500058250427246
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,1,power_law_1.01,0.63133056640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,1,power_law_1.01,6.161923217773437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,1,power_law_1.01,0.7495590209960937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,1,power_law_1.01,0.28254911422729495
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,1,power_law_1.01,1.028551712036133
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,1,power_law_1.01,0.11811584472656249
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,1,power_law_1.01,1.2938803100585938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,2,1,power_law_1.01,7.63712890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,1,power_law_1.01,1.8393823242187501
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,1,power_law_1.01,2.3954847717285155
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,2,1,power_law_1.01,2.9577005004882815
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,1,power_law_1.01,0.03616767883300781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,1,power_law_1.01,0.04441279888153076
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,1,power_law_1.01,0.15005056381225584
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,1,power_law_1.01,0.15234623908996583
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,1,power_law_1.01,0.08867072105407715
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,1,power_law_1.01,0.15797887802124025
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,1,power_law_1.01,0.1613190460205078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,1,power_law_1.01,0.05641920089721679
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,1,power_law_1.01,0.16314880371093748
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,1,power_law_1.01,0.0730457592010498
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,1,power_law_1.01,0.1412051200866699
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,1,power_law_1.01,0.17320896148681642
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,1,power_law_1.01,0.14561408042907714
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,1,power_law_1.01,0.1294809627532959
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,1,power_law_1.01,0.1834809684753418
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,1,power_law_1.01,0.18840576171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,1,power_law_1.01,0.2062156867980957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,1,power_law_1.01,0.24011520385742186
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.01,0.27314943313598633
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.01,0.353265266418457
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.01,0.4380704116821289
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,1,power_law_1.2,0.06431231975555421
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.01,0.6042335891723634
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,1,power_law_1.2,0.06898687839508057
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,1,power_law_1.2,0.10275199890136719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,1,power_law_1.2,0.14876031875610352
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,1,power_law_1.2,0.21042303085327146
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,1,power_law_1.2,0.3506662368774414
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,1,power_law_1.2,0.3827987289428711
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,1,power_law_1.2,0.39947200775146485
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,1,power_law_1.2,0.4144569778442383
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,1,power_law_1.2,0.4191436767578125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,1,power_law_1.2,0.4482617568969727
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,1,power_law_1.2,0.45866367340087894
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,1,power_law_1.2,0.46657150268554687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,1,power_law_1.2,0.502419204711914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,1,power_law_1.2,0.5167046356201171
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,1,power_law_1.2,0.5188179016113281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,1,power_law_1.2,0.5355839920043945
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,1,power_law_1.2,0.6173798370361328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,1,power_law_1.2,0.6773101043701172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,1,power_law_1.2,0.8513785552978514
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,1,power_law_1.2,1.044510726928711
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,1,power_law_1.2,1.4125100708007812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,1,power_law_1.2,1.7852557373046873
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,1,power_law_1.2,2.540156097412109
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,1,power_law_1.2,3.2099761962890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.01,0.7787980651855468
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,1,power_law_1.2,4.675926513671875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.01,1.136779556274414
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,1,power_law_1.2,0.059103999137878414
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,1,power_law_1.2,0.06579391956329346
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,1,power_law_1.2,0.08566656112670898
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.01,1.5079718017578125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,1,power_law_1.2,0.11082176208496093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,1,power_law_1.2,0.13828479766845703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,1,power_law_1.2,0.22009855270385742
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,1,power_law_1.2,0.2334009552001953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.01,2.2497906494140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,1,power_law_1.2,0.23951168060302735
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,1,power_law_1.2,0.24733760833740231
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,1,power_law_1.2,0.24650560379028322
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,1,power_law_1.2,0.2545100784301758
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,1,power_law_1.2,0.28758272171020505
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.01,3.01003662109375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,1,power_law_1.2,0.29200000762939454
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,1,power_law_1.2,6.22587890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,1,power_law_1.2,0.3109004783630371
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,1,power_law_1.2,0.3113798332214356
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,1,power_law_1.2,0.323631362915039
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,1,power_law_1.2,0.3490265655517578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.01,3.7960076904296876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,1,power_law_1.2,0.37731136322021486
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,1,power_law_1.2,0.4419424057006836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,1,power_law_1.2,0.5087500762939453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,1,power_law_1.2,0.6281472015380859
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,1,power_law_1.2,0.7643212890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,1,power_law_1.2,1.0329869079589844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,1,power_law_1.2,1.3013253784179688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,1,power_law_1.2,0.037360639572143556
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,1,power_law_1.2,0.04348544120788574
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,1,power_law_1.2,1.8429766845703124
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,1,power_law_1.2,0.05403711795806885
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,1,power_law_1.2,0.06759039878845215
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,2,1,power_law_1.2,7.733696899414062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,1,power_law_1.2,2.414187469482422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,1,power_law_1.2,0.08348480224609375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,1,power_law_1.2,0.13329983711242677
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,1,power_law_1.2,0.14373696327209473
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,1,power_law_1.2,0.14868800163269041
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,1,power_law_1.2,0.1530470371246338
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,1,power_law_1.2,0.15506943702697754
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,1,power_law_1.2,0.16023872375488282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,1,power_law_1.2,0.16724416732788086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,2,1,power_law_1.2,2.938170776367188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,1,power_law_1.2,0.1711097526550293
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,1,power_law_1.2,0.18262271881103515
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,1,power_law_1.2,0.19319551467895507
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,1,power_law_1.2,0.2986425590515137
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,1,power_law_1.2,0.22633407592773436
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,1,power_law_1.2,0.20019456863403323
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,1,power_law_1.2,0.2714867210388184
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.2,0.43641918182373046
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.2,0.5503327941894531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.2,0.7951308441162109
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.2,0.32384639739990234
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,1,power_law_1.01,0.11157055854797364
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,1,power_law_1.01,0.1513875198364258
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.2,1.0363654327392577
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,1,power_law_1.01,0.3897452926635742
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,1,power_law_1.01,0.5463641738891601
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,1,power_law_1.01,0.22172992706298827
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,1,power_law_1.01,0.9247296142578125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,1,power_law_1.01,0.9986943817138672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,1,power_law_1.01,1.0663314819335938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,1,power_law_1.01,1.0974002838134767
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,1,power_law_1.01,1.14607421875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,1,power_law_1.01,1.2384320068359376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,1,power_law_1.01,1.3555308532714843
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,1,power_law_1.01,1.3220460510253906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,1,power_law_1.01,1.3365061950683592
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,1,power_law_1.01,1.3905459594726561
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,1,power_law_1.01,1.436895294189453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,1,power_law_1.01,1.5266816711425781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,1,power_law_1.01,1.7311520385742187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,1,power_law_1.01,1.9457203674316408
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,1,power_law_1.01,2.3489862060546876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,1,power_law_1.01,2.8142880249023436
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,1,power_law_1.01,3.665718994140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,1,power_law_1.01,4.494383850097656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.2,1.5218815612792969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,1,power_law_1.01,6.3591552734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.2,2.9888735961914064
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.2,2.009349822998047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,1,power_law_1.01,0.1314899158477783
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,1,power_law_1.01,0.10611840248107911
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.2,3.9820690917968746
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,1,power_law_1.01,0.16083456039428712
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,1,power_law_1.01,8.14926025390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.2,4.995395202636719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,1,power_law_1.01,0.32653953552246096
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,1,power_law_1.01,0.2513676834106445
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,1,power_law_1.01,0.5051359939575195
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,1,power_law_1.01,0.5475795364379883
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,1,power_law_1.01,0.5719929504394531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,1,power_law_1.01,0.5978956985473632
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,1,power_law_1.01,0.6084812927246094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,1,power_law_1.01,0.6219289779663086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,1,power_law_1.01,0.6386304092407227
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,1,power_law_1.01,0.6955072021484374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,1,power_law_1.01,0.7087648010253906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,1,power_law_1.01,0.7290156555175782
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,1,power_law_1.01,0.7305741119384765
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,1,power_law_1.01,0.7732390594482422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,1,power_law_1.01,0.8376959991455077
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,1,power_law_1.01,0.8942098999023438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,1,power_law_1.01,11.500340576171876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,1,power_law_1.01,0.9774982452392578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,1,power_law_1.01,1.1517369842529297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,1,power_law_1.01,1.431356201171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,1,power_law_1.01,1.7316192626953124
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,1,power_law_1.01,2.2912147521972654
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,1,power_law_1.01,2.850262451171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,1,power_law_1.01,3.9497778320312498
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,1,power_law_1.01,14.943197021484375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,1,power_law_1.01,5.104813537597656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,2,1,power_law_1.01,6.310552978515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,2,1,power_law_1.01,18.479261474609377
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,1,power_law_1.01,0.07243135929107666
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,1,power_law_1.01,0.06066624164581299
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,1,power_law_1.01,0.08669183731079103
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,1,power_law_1.01,0.3514150238037109
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,1,power_law_1.01,0.1348057556152344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,1,power_law_1.01,0.35967296600341797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,1,power_law_1.01,0.3715897750854492
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,1,power_law_1.01,0.38390464782714845
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,1,power_law_1.01,0.38878273010253905
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,1,power_law_1.01,0.3999270248413086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,1,power_law_1.01,0.4154336166381836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,1,power_law_1.01,0.29461183547973635
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,1,power_law_1.01,0.42635456085205076
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,1,power_law_1.01,0.18140352249145508
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,1,power_law_1.01,0.32179519653320315
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,1,power_law_1.01,0.3386431884765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,1,power_law_1.01,0.5224774551391602
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,1,power_law_1.01,0.45948543548583987
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.01,0.5762598419189453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.01,0.703567352294922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.01,1.1275033569335937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.01,0.8493440246582031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,1,power_law_1.2,0.12895423889160157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.01,1.4186714172363282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,1,power_law_1.2,0.2108678436279297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,1,power_law_1.2,0.3488198471069336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,1,power_law_1.2,0.48336704254150387
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,1,power_law_1.2,0.9166547393798827
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.01,2.02849853515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,1,power_law_1.2,1.0081830596923829
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,1,power_law_1.2,0.11500736236572266
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,1,power_law_1.2,1.073587875366211
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,1,power_law_1.2,1.137947540283203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,1,power_law_1.2,1.3022988891601561
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,1,power_law_1.2,1.346604766845703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,1,power_law_1.2,1.3752787780761717
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,1,power_law_1.2,1.189512939453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,1,power_law_1.2,1.396194610595703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,1,power_law_1.2,1.4837055969238282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,1,power_law_1.2,1.5200416564941406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,1,power_law_1.2,1.5773817443847657
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,1,power_law_1.2,1.769276123046875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,1,power_law_1.2,1.9590605163574217
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,1,power_law_1.2,2.4125573730468752
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,1,power_law_1.2,2.8086367797851564
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,1,power_law_1.2,3.7064697265625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,1,power_law_1.2,4.539607543945313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,1,power_law_1.2,6.334007568359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,1,power_law_1.2,8.068259277343751
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,1,power_law_1.2,0.10775232315063477
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.01,2.6530535888671873
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,1,power_law_1.2,0.15850239753723144
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,1,power_law_1.2,0.11149312019348145
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,1,power_law_1.2,0.22365119934082028
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,1,power_law_1.2,0.29783615112304684
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,1,power_law_1.2,0.5004217529296875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,1,power_law_1.2,0.5493708801269531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.01,3.900704650878906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,1,power_law_1.2,11.5761376953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,1,power_law_1.2,0.5864275360107423
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,1,power_law_1.2,0.610953598022461
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,1,power_law_1.2,0.6376665496826173
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,1,power_law_1.2,0.621146240234375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,1,power_law_1.2,0.655462417602539
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,2,1,power_law_1.01,6.548918457031251
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.01,5.20299560546875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,1,power_law_1.2,0.7414662170410156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,1,power_law_1.2,0.7570706939697265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,1,power_law_1.2,0.7475686645507812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,1,power_law_1.2,0.7677011108398437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,1,power_law_1.2,0.7967327880859375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,1,power_law_1.2,0.8352339172363281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,1,power_law_1.2,0.9043603515624999
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,1,power_law_1.2,1.038449249267578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,1,power_law_1.2,14.996722412109374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,1,power_law_1.2,1.1871084594726562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,1,power_law_1.2,1.4972370910644532
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,1,power_law_1.2,1.7983302307128906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,1,power_law_1.2,0.0608025598526001
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,1,power_law_1.2,2.307923126220703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,1,power_law_1.2,0.06249983787536621
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,1,power_law_1.2,2.8692410278320315
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,1,power_law_1.2,0.08457407951354981
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,1,power_law_1.2,0.12162688255310058
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,1,power_law_1.2,0.16034559249877928
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,1,power_law_1.2,0.2931398391723633
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,1,power_law_1.2,0.3224691009521484
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,1,power_law_1.2,3.9963214111328127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,1,power_law_1.2,0.3469971084594726
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,1,power_law_1.2,0.36043968200683596
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,1,power_law_1.2,0.3709280014038086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,1,power_law_1.2,0.3839321517944336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,2,1,power_law_1.2,18.496226806640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,1,power_law_1.2,5.177855224609376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,1,power_law_1.2,0.3964236831665039
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,1,power_law_1.2,0.40997886657714844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,1,power_law_1.2,0.42338111877441403
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,2,1,power_law_1.2,6.31645263671875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,1,power_law_1.2,0.44450942993164066
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,1,power_law_1.2,0.46384639739990235
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,1,power_law_1.2,0.49913150787353516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.2,0.6586483001708985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,1,power_law_1.2,0.5773657608032227
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.2,0.8468147277832031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.2,1.0533805084228516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,1,power_law_1.01,0.14828672409057617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.2,1.4874624633789062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,1,power_law_1.01,0.1977881622314453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,1,power_law_1.01,0.3314924621582031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,1,power_law_1.01,0.7688832092285156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,1,power_law_1.01,1.1009817504882813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,1,power_law_1.01,1.298616943359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,1,power_law_1.01,1.721543731689453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,1,power_law_1.01,1.8194099426269532
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,1,power_law_1.01,1.8933235168457032
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,1,power_law_1.01,2.0015193176269532
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,1,power_law_1.01,2.0746394348144532
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,1,power_law_1.01,2.2043431091308596
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,1,power_law_1.01,2.375196228027344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,1,power_law_1.01,2.553283233642578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,1,power_law_1.01,2.7393811035156252
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,1,power_law_1.01,2.586182250976562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.2,1.9391743469238283
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,1,power_law_1.01,2.79131591796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.2,3.926414184570312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.2,2.9204852294921873
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,1,power_law_1.01,2.9824038696289064
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,1,power_law_1.01,3.387576904296875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.2,5.974185791015625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,1,power_law_1.01,3.8202810668945313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.2,7.963438110351563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,1,power_law_1.01,4.710654602050782
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,1,power_law_1.01,0.1769887924194336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,1,power_law_1.01,5.533964233398438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,1,power_law_1.01,0.21215999603271482
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,1,power_law_1.01,0.26478591918945316
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,1,power_law_1.01,7.233013305664062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,1,power_law_1.01,0.2988185691833496
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,1,power_law_1.01,0.4789587020874023
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,1,power_law_1.01,8.997984008789063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,1,power_law_1.01,0.6508672332763672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,1,power_law_1.01,0.7338944244384765
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,1,power_law_1.01,0.9293548583984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,1,power_law_1.01,0.9900262451171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,1,power_law_1.01,1.0577951812744142
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,1,power_law_1.01,1.01744384765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,1,power_law_1.01,12.119870605468751
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,1,power_law_1.01,1.1044633483886719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,1,power_law_1.01,1.1587583923339844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,1,power_law_1.01,1.1254617309570312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,1,power_law_1.01,1.200576629638672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,1,power_law_1.01,1.2259891510009766
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,1,power_law_1.01,1.4022015380859376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,1,power_law_1.01,15.478029785156249
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,1,power_law_1.01,1.4872767639160156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,1,power_law_1.01,1.535021514892578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,1,power_law_1.01,1.6544274902343752
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,1,power_law_1.01,1.817371520996094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,1,power_law_1.01,2.108326416015625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,1,power_law_1.01,2.369395904541016
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,1,power_law_1.01,2.9585592651367185
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,2,1,power_law_1.01,18.6876220703125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,1,power_law_1.01,3.5480569458007816
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,1,power_law_1.01,4.665433044433594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,1,power_law_1.01,5.88804931640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,2,1,power_law_1.01,7.007658691406251
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,1,power_law_1.01,0.09010239601135253
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,1,power_law_1.01,0.08962559700012207
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,1,power_law_1.01,0.09056256294250489
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,1,power_law_1.01,0.09407551765441895
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,1,power_law_1.01,0.13633664131164552
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,1,power_law_1.01,0.09992447853088379
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,1,power_law_1.01,0.1593727970123291
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,1,power_law_1.01,0.18384960174560547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,1,power_law_1.01,0.11148672103881836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,1,power_law_1.01,0.12336447715759277
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,1,power_law_1.01,0.21876287460327148
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,1,power_law_1.01,0.2431660842895508
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,1,power_law_1.01,0.14647040367126465
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,1,power_law_1.01,0.4014995193481445
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,1,power_law_1.01,0.2914816093444824
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,1,power_law_1.01,0.35438526153564454
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,1,power_law_1.01,0.5165004730224609
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,1,power_law_1.01,0.7392134094238282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.01,0.9546495819091797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,1,power_law_1.2,0.14764479637145994
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.01,1.3966726684570312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.01,1.8338528442382813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,1,power_law_1.2,0.19775039672851563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,1,power_law_1.2,0.24594560623168946
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.01,3.5956295776367186
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,1,power_law_1.2,0.6477625274658203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.01,2.7130047607421877
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,1,power_law_1.2,0.9768025970458984
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.01,5.339481811523437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,1,power_law_1.2,1.1401805114746093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,1,power_law_1.2,1.6192454528808593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,1,power_law_1.2,1.701130828857422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,1,power_law_1.2,1.8151271057128906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,1,power_law_1.2,2.0746873474121097
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,1,power_law_1.2,1.939849548339844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,1,power_law_1.2,2.1287129211425784
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,1,power_law_1.2,2.3918380737304688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,1,power_law_1.2,2.5910629272460937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,1,power_law_1.2,2.7917132568359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,1,power_law_1.2,2.6300775146484376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,1,power_law_1.2,2.876922912597656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,1,power_law_1.2,0.3997452926635742
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,1,power_law_1.2,3.095733642578125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,1,power_law_1.2,3.540597839355469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,1,power_law_1.2,3.9782046508789066
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,1,power_law_1.2,4.828208618164062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,1,power_law_1.2,5.761992187500001
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,1,power_law_1.2,0.1749785614013672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,1,power_law_1.2,7.336109008789063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,1,power_law_1.2,9.01296875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,1,power_law_1.2,0.20967615127563474
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,1,power_law_1.2,0.2983168029785156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,1,power_law_1.2,0.42930366516113283
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,1,power_law_1.2,12.344371337890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,1,power_law_1.2,0.5785228729248046
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,1,power_law_1.2,0.2380134391784668
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,1,power_law_1.2,0.6663648223876952
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,1,power_law_1.2,0.8886726379394532
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,1,power_law_1.2,0.9363104248046875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,1,power_law_1.2,0.9780953979492187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,1,power_law_1.2,1.0384774780273438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,1,power_law_1.2,1.075976333618164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,1,power_law_1.2,15.65361328125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,1,power_law_1.2,1.0981011199951172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,1,power_law_1.2,1.143824005126953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,1,power_law_1.2,1.1947532653808595
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,1,power_law_1.2,1.220221405029297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,1,power_law_1.2,1.4108819580078125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,1,power_law_1.2,1.5103872680664063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,1,power_law_1.2,1.580699462890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,1,power_law_1.2,1.760675811767578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,1,power_law_1.2,1.8704325866699218
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,2,1,power_law_1.2,19.06093994140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,1,power_law_1.2,2.234410858154297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,1,power_law_1.2,2.5426060485839845
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,1,power_law_1.2,3.071904602050781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,1,power_law_1.2,0.08907135963439941
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,1,power_law_1.2,3.6898764038085936
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,1,power_law_1.2,0.08944640159606934
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,1,power_law_1.2,0.09095808029174804
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,1,power_law_1.2,4.830841064453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,1,power_law_1.2,0.09268287658691407
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,1,power_law_1.2,0.10000063896179198
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,1,power_law_1.2,6.040435791015625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,1,power_law_1.2,0.11107199668884278
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,1,power_law_1.2,0.12367424011230468
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,2,1,power_law_1.2,7.1429553222656255
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,1,power_law_1.2,0.13583871841430664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,1,power_law_1.2,0.14820927619934082
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,1,power_law_1.2,0.15931712150573732
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,1,power_law_1.2,0.18331199645996094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,1,power_law_1.2,0.24376127243041995
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,1,power_law_1.2,0.29133184432983394
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,1,power_law_1.2,0.4021100616455078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,1,power_law_1.2,0.2184127998352051
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,1,power_law_1.2,0.3513996887207031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,1,power_law_1.2,0.5158316802978515
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,1,power_law_1.2,0.7404934692382812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.2,0.9587372589111329
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.2,1.3947360229492187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.2,1.8402394104003907
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.2,2.7175238037109373
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,1,power_law_1.01,0.15476223945617676
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,1,power_law_1.01,0.1722707176208496
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,1,power_law_1.01,0.17720640182495118
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,1,power_law_1.01,0.17610111236572265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,1,power_law_1.01,0.17663103103637695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,1,power_law_1.01,0.18725055694580078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,1,power_law_1.01,0.2240928077697754
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,1,power_law_1.01,0.22659584045410158
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,1,power_law_1.01,0.2355936050415039
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,1,power_law_1.01,0.24667072296142578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,1,power_law_1.01,0.16750335693359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,1,power_law_1.01,0.25631744384765626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,1,power_law_1.01,0.2923648071289063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,1,power_law_1.01,0.34799617767333985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,1,power_law_1.01,0.42335166931152346
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,1,power_law_1.01,0.5122399902343749
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,1,power_law_1.01,0.6864268493652343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,1,power_law_1.01,0.8777574157714844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,1,power_law_1.01,1.2346144104003907
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,1,power_law_1.01,1.6854162597656248
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.2,3.5879147338867186
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,1,power_law_1.01,2.5240249633789062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,1,power_law_1.01,3.2620166015625003
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.2,5.324639892578125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,1,power_law_1.01,0.06513792037963867
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,1,power_law_1.01,0.060629758834838864
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,1,power_law_1.01,0.06529983997344971
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,1,power_law_1.01,0.09072832107543946
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,1,power_law_1.01,0.09726847648620605
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,1,power_law_1.01,0.09836159706115723
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,1,power_law_1.01,0.0993126392364502
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,1,power_law_1.01,4.988850708007813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,1,power_law_1.01,0.09928319931030274
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,1,power_law_1.01,0.10224575996398926
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,1,power_law_1.01,0.10316608428955079
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,1,power_law_1.01,0.10410816192626952
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,1,power_law_1.01,6.6779736328125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,1,power_law_1.01,0.10681920051574707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,1,power_law_1.01,0.12388671874999999
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,1,power_law_1.01,0.1267686367034912
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,1,power_law_1.01,0.13849727630615233
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,1,power_law_1.01,0.14013824462890626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,1,power_law_1.01,0.16657855987548828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,1,power_law_1.01,0.19602943420410157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,1,power_law_1.01,0.2527891159057617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,1,power_law_1.01,0.3037664031982422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,1,power_law_1.01,0.4127372741699219
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,1,power_law_1.01,0.5362771224975587
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,4,1,power_law_1.01,8.503301391601562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,1,power_law_1.01,0.09990655899047851
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,1,power_law_1.01,0.7638412475585937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,1,power_law_1.01,1.0198777770996093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,1,power_law_1.01,1.5412147521972657
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,1,power_law_1.01,2.0323532104492186
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,4,1,power_law_1.01,2.5095558166503906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,1,power_law_1.01,0.02705984115600586
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,1,power_law_1.01,0.05690239906311036
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,1,power_law_1.01,0.05728576183319092
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,1,power_law_1.01,0.057377920150756836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,1,power_law_1.01,0.034903678894042965
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,1,power_law_1.01,0.05623167991638184
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,1,power_law_1.01,0.05663040161132813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,1,power_law_1.01,0.057518081665039064
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,1,power_law_1.01,0.05847871780395508
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,1,power_law_1.01,0.05616703987121582
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,1,power_law_1.01,0.03934015989303589
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,1,power_law_1.01,0.05779136180877685
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,1,power_law_1.01,0.05947648048400879
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,1,power_law_1.01,0.061783680915832515
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,1,power_law_1.01,0.06444287776947022
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,1,power_law_1.01,0.06738431930541992
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,1,power_law_1.01,0.07273791790008545
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.01,0.10505727767944335
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,1,power_law_1.01,0.08948991775512696
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.01,0.13431551933288574
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.01,0.16623231887817383
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.01,0.23081344604492188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.01,0.29502847671508786
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,1,power_law_1.2,0.07615680217742919
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,1,power_law_1.2,0.15327487945556642
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,1,power_law_1.2,0.08622783660888672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,1,power_law_1.2,0.16853376388549807
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,1,power_law_1.2,0.1694041633605957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,1,power_law_1.2,0.17469951629638672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,1,power_law_1.2,0.17357055664062498
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,1,power_law_1.2,0.17554304122924805
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,1,power_law_1.2,0.0962611198425293
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,1,power_law_1.2,0.19466047286987304
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,1,power_law_1.2,0.22345087051391604
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,1,power_law_1.2,0.2236012840270996
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,1,power_law_1.2,0.23406463623046875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,1,power_law_1.2,0.24787328720092772
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,1,power_law_1.2,0.26127872467041013
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,1,power_law_1.2,0.30847295761108395
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,1,power_law_1.2,0.3480281448364258
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,1,power_law_1.2,0.4245087814331055
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,1,power_law_1.2,0.5132735824584961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.01,0.42605758666992183
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,1,power_law_1.2,0.6960562896728516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,1,power_law_1.2,0.8800991821289064
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,1,power_law_1.2,1.2770931243896484
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,1,power_law_1.2,1.7017625427246095
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,1,power_law_1.2,2.528663024902344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.01,0.5675487899780274
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.01,0.8379756927490234
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,1,power_law_1.2,0.05456192016601562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,1,power_law_1.2,0.06367104053497315
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,1,power_law_1.2,0.06663040161132813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,1,power_law_1.2,3.379587097167969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.01,1.103829116821289
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,1,power_law_1.2,0.09255231857299805
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,1,power_law_1.2,0.09571328163146972
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,1,power_law_1.2,0.10402496337890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,1,power_law_1.2,0.09951807975769042
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,1,power_law_1.2,0.10271743774414062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,1,power_law_1.2,0.100830078125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,1,power_law_1.2,0.10423551559448244
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,1,power_law_1.2,0.10354240417480469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,1,power_law_1.2,0.108090238571167
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,1,power_law_1.2,0.10573951721191406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,1,power_law_1.2,0.11842623710632325
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,1,power_law_1.2,0.1300979232788086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,1,power_law_1.2,0.14270848274230957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,1,power_law_1.2,0.14010239601135255
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,1,power_law_1.2,0.1657721519470215
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,1,power_law_1.2,0.19643840789794922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,1,power_law_1.2,0.2499475288391113
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,1,power_law_1.2,0.3062137603759766
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,1,power_law_1.2,0.41502719879150385
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,1,power_law_1.2,4.993717651367187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,1,power_law_1.2,0.5312614440917969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,1,power_law_1.2,0.766468505859375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,1,power_law_1.2,0.028334081172943115
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,1,power_law_1.2,1.009667205810547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,1,power_law_1.2,0.033522560596466064
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,1,power_law_1.2,0.03701247930526733
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,1,power_law_1.2,0.05639488220214843
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,1,power_law_1.2,0.05493247985839844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,1,power_law_1.2,0.057677440643310554
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,1,power_law_1.2,1.5151437377929686
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,1,power_law_1.2,0.057877120971679685
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,1,power_law_1.2,0.056816000938415524
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,1,power_law_1.2,0.057994241714477535
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,1,power_law_1.2,0.05857791900634766
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,1,power_law_1.2,0.05960063934326172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,1,power_law_1.2,2.060414123535156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,1,power_law_1.2,6.609755249023438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,1,power_law_1.2,0.05877439975738526
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,1,power_law_1.2,0.06165887832641601
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,1,power_law_1.2,0.07361855983734131
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,4,1,power_law_1.2,2.5041477966308596
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.2,0.11145919799804686
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,4,1,power_law_1.2,8.670247192382813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.01,1.443317108154297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,1,power_law_1.2,0.05740543842315674
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,1,power_law_1.2,0.06564735889434814
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,1,power_law_1.2,0.06968832015991211
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,1,power_law_1.2,0.09109824180603028
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.2,0.13886400222778322
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.2,0.17132608413696288
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,1,power_law_1.01,0.10328512191772461
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,1,power_law_1.01,0.13123968124389648
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.2,0.2402124786376953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,1,power_law_1.01,0.1536070442199707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,1,power_law_1.01,0.2505862426757813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,1,power_law_1.01,0.25507392883300783
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,1,power_law_1.01,0.2575436782836914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,1,power_law_1.01,0.26214975357055664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,1,power_law_1.01,0.25566144943237307
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,1,power_law_1.01,0.26520256042480467
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,1,power_law_1.01,0.28555904388427733
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,1,power_law_1.01,0.3570771026611328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,1,power_law_1.01,0.3672864151000977
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,1,power_law_1.01,0.38010177612304685
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,1,power_law_1.01,0.4085913467407226
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,1,power_law_1.01,0.42025089263916016
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,1,power_law_1.01,0.46080257415771486
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,1,power_law_1.01,0.5305779266357422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,1,power_law_1.01,0.6827161407470703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,1,power_law_1.01,0.8412537384033204
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,1,power_law_1.01,1.2088185882568359
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.2,0.3025811195373535
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,1,power_law_1.01,1.5100717163085937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,1,power_law_1.01,2.2543775939941404
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,1,power_law_1.01,4.3786553955078125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,1,power_law_1.01,2.897486572265625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,1,power_law_1.01,5.752747802734374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,1,power_law_1.01,8.590075073242186
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,1,power_law_1.01,0.06574975967407226
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,1,power_law_1.01,0.07782464027404785
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,1,power_law_1.01,0.09153471946716309
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.2,0.4403481674194336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,1,power_law_1.01,0.13439040184020995
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,1,power_law_1.01,0.14403648376464845
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,1,power_law_1.01,0.14694592475891113
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,1,power_law_1.01,0.14769408226013184
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,1,power_law_1.01,0.14770432472229006
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,1,power_law_1.01,0.14942655563354493
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,1,power_law_1.01,0.15129023551940918
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.2,0.8573241424560546
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,1,power_law_1.01,0.15325823783874512
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,1,power_law_1.01,0.15512255668640135
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,1,power_law_1.01,0.16409151077270506
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,1,power_law_1.01,0.17263551712036135
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,1,power_law_1.01,0.1777907180786133
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.2,1.170170211791992
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,1,power_law_1.01,0.18357120513916017
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,1,power_law_1.01,0.19713472366333007
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,1,power_law_1.01,0.23371263504028322
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.2,0.5729663848876954
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,1,power_law_1.01,11.501533203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,1,power_law_1.01,0.2863936042785645
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,1,power_law_1.01,0.35986942291259766
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,1,power_law_1.01,0.442729606628418
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,1,power_law_1.01,0.6218624114990234
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,1,power_law_1.01,0.8414393615722655
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,1,power_law_1.01,1.147869415283203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,1,power_law_1.01,1.5633453369140624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.2,1.4437023925781252
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,1,power_law_1.01,2.33227783203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,1,power_law_1.01,3.0949914550781252
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,4,1,power_law_1.01,3.915224914550781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,4,1,power_law_1.01,15.4115087890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,1,power_law_1.01,0.03217152118682861
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,1,power_law_1.01,0.08017279624938964
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,1,power_law_1.01,0.0806553554534912
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,1,power_law_1.01,0.04633664131164551
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,1,power_law_1.01,0.08173376083374023
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,1,power_law_1.01,0.054202241897583006
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,1,power_law_1.01,0.0796895980834961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,1,power_law_1.01,0.0786841583251953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,1,power_law_1.01,0.07991615772247315
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,1,power_law_1.01,0.07977344036102295
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,1,power_law_1.01,0.07988096237182617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,1,power_law_1.01,0.08317824363708495
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,1,power_law_1.01,0.08919743537902833
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,1,power_law_1.01,0.09496319770812989
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,1,power_law_1.01,0.08166399955749512
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,1,power_law_1.01,0.09215871810913086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,1,power_law_1.01,0.10498944282531739
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.01,0.1556544017791748
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,1,power_law_1.01,0.13043519973754883
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.01,0.2018976020812988
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.01,0.24983360290527346
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.01,0.34772735595703125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,1,power_law_1.2,0.12293760299682617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,1,power_law_1.2,0.14443584442138674
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.01,0.44733119964599605
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,1,power_law_1.2,0.25146495819091796
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,1,power_law_1.2,0.24650304794311523
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,1,power_law_1.2,0.10234880447387695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,1,power_law_1.2,0.25651968002319336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,1,power_law_1.2,0.26160192489624023
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.01,0.6702777862548828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.01,0.8800102233886719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,1,power_law_1.2,0.254400634765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,1,power_law_1.2,0.2706060791015625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,1,power_law_1.2,0.3062374305725098
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,1,power_law_1.2,0.3726675033569336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,1,power_law_1.2,0.39431808471679686
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,1,power_law_1.2,0.37228031158447267
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,1,power_law_1.2,0.4061740875244141
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,1,power_law_1.2,0.4257171249389648
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,1,power_law_1.2,0.4733695983886719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,1,power_law_1.2,0.535665283203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,1,power_law_1.2,0.6858483123779296
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.01,1.3541708374023438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,1,power_law_1.2,0.8447865295410157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,1,power_law_1.2,1.1878278350830078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,1,power_law_1.2,1.5488140869140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,1,power_law_1.2,2.2435935974121093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.01,1.8479032897949217
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,1,power_law_1.2,0.06955647945404053
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,1,power_law_1.2,2.917198791503906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,1,power_law_1.2,0.07512320041656494
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,1,power_law_1.2,0.08730496406555176
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,1,power_law_1.2,0.1364896011352539
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,1,power_law_1.2,0.15167807579040526
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,1,power_law_1.2,0.1474944019317627
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,1,power_law_1.2,0.1498380756378174
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,1,power_law_1.2,0.1483910369873047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,1,power_law_1.2,0.1513542366027832
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,1,power_law_1.2,0.15217280387878418
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,1,power_law_1.2,4.333923950195312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,1,power_law_1.2,0.15357248306274413
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,1,power_law_1.2,0.15666687965393067
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,1,power_law_1.2,0.1675775909423828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,1,power_law_1.2,0.17228992462158205
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,1,power_law_1.2,0.1829203224182129
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,1,power_law_1.2,0.20032255172729493
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,1,power_law_1.2,0.19951936721801758
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,1,power_law_1.2,0.23442304611206052
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,1,power_law_1.2,0.2798559951782226
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,1,power_law_1.2,0.35600128173828127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,1,power_law_1.2,5.74584228515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,1,power_law_1.2,0.4442502212524414
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,1,power_law_1.2,0.6148204803466797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,1,power_law_1.2,0.81893310546875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,1,power_law_1.2,1.220880661010742
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,1,power_law_1.2,1.5715213012695313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,1,power_law_1.2,0.03403007984161377
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,1,power_law_1.2,0.04510655879974365
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,1,power_law_1.2,0.0509984016418457
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,1,power_law_1.2,2.3563334655761716
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,1,power_law_1.2,0.0789305591583252
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,1,power_law_1.2,8.575173339843749
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,1,power_law_1.2,0.08126591682434083
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,1,power_law_1.2,0.07976704120635987
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,1,power_law_1.2,0.08424896240234375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,1,power_law_1.2,0.08052224159240723
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,1,power_law_1.2,0.08031167984008789
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,1,power_law_1.2,0.08502976417541504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,1,power_law_1.2,0.08073920249938965
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,1,power_law_1.2,3.1027508544921876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,4,1,power_law_1.2,3.896936950683594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,1,power_law_1.2,0.10520319938659668
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,1,power_law_1.2,11.73076171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,4,1,power_law_1.2,16.429384765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.01,2.2300070190429686
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.2,0.2104198455810547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,1,power_law_1.2,0.0822719955444336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,1,power_law_1.2,0.08259584426879882
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,1,power_law_1.2,0.08879551887512208
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,1,power_law_1.2,0.09374848365783692
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,1,power_law_1.2,0.10808256149291992
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,1,power_law_1.2,0.13208000183105467
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.2,0.15531904220581055
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.2,0.2526668739318848
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,1,power_law_1.01,0.07805312156677247
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.2,0.45383937835693355
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.2,0.9021740722656251
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.2,0.35416831970214846
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.2,0.6586067199707031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.2,1.3470060729980469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,1,power_law_1.01,0.10169856071472168
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,1,power_law_1.01,0.11500800132751465
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.2,1.8478291320800782
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,1,power_law_1.01,0.5865753555297852
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,1,power_law_1.01,0.669070053100586
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,1,power_law_1.01,0.19656063079833985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,1,power_law_1.01,0.601377296447754
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,1,power_law_1.01,0.7077734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,1,power_law_1.01,0.6313241577148437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,1,power_law_1.01,0.30975679397583006
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,1,power_law_1.01,0.6434726715087891
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,1,power_law_1.01,0.8366572570800781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,1,power_law_1.01,0.4246694564819336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,1,power_law_1.01,0.8854483032226563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.2,2.317704315185547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,1,power_law_1.01,0.7484038543701172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,1,power_law_1.01,0.7773779296875001
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,1,power_law_1.01,0.9828179168701172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,1,power_law_1.01,1.2103238677978516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,1,power_law_1.01,1.0440985870361328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,1,power_law_1.01,1.188033905029297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,1,power_law_1.01,1.319475860595703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,1,power_law_1.01,1.530471649169922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,1,power_law_1.01,1.7699320983886717
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,1,power_law_1.01,0.13284543991088865
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,1,power_law_1.01,2.229799041748047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,1,power_law_1.01,3.6910943603515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,1,power_law_1.01,0.20074304580688476
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,4,1,power_law_1.01,5.635502319335937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,1,power_law_1.01,0.11411968231201172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,1,power_law_1.01,2.7210943603515623
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,1,power_law_1.01,0.4337561416625976
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,1,power_law_1.01,0.4484294509887695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,1,power_law_1.01,4.602500610351562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,1,power_law_1.01,0.26936639785766603
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,1,power_law_1.01,0.4687155151367188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,1,power_law_1.01,0.14253824234008788
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,1,power_law_1.01,0.48035327911376957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,1,power_law_1.01,0.498455696105957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,1,power_law_1.01,0.5245619201660157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,1,power_law_1.01,0.5385337448120117
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,1,power_law_1.01,0.5126982498168945
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,1,power_law_1.01,0.6157516860961915
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,1,power_law_1.01,0.5969574356079101
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,1,power_law_1.01,0.6341702270507812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,1,power_law_1.01,0.6703545379638671
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,1,power_law_1.01,0.7023257446289063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,1,power_law_1.01,0.7921868896484374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,1,power_law_1.01,0.8689529418945312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,1,power_law_1.01,1.0328358459472657
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,1,power_law_1.01,0.3341305541992187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,1,power_law_1.01,1.2112992095947266
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,1,power_law_1.01,1.569036102294922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,1,power_law_1.01,1.9439181518554687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,1,power_law_1.01,2.639054870605469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,1,power_law_1.01,3.362836608886719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,4,1,power_law_1.01,4.096817932128906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,1,power_law_1.01,0.07268799781799316
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,1,power_law_1.01,0.059481601715087894
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,1,power_law_1.01,0.07980224132537841
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,1,power_law_1.01,0.10787839889526367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,1,power_law_1.01,0.290064640045166
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,1,power_law_1.01,0.2985670471191406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,1,power_law_1.01,0.14985471725463867
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,1,power_law_1.01,0.31415552139282227
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,1,power_law_1.01,0.32531200408935546
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,1,power_law_1.01,0.33434814453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,1,power_law_1.01,0.34995006561279296
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,1,power_law_1.01,0.2655379295349121
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,1,power_law_1.01,0.3642444610595703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,1,power_law_1.01,0.37783935546874997
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,1,power_law_1.01,0.40187007904052735
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,1,power_law_1.01,0.1893836784362793
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,1,power_law_1.01,0.2771161651611328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.01,0.6015859222412109
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.01,0.5053049468994141
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.01,0.708760986328125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.01,0.9074438476562501
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.01,1.1343148803710936
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,1,power_law_1.2,0.07799615859985351
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,1,power_law_1.01,0.4526847839355469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.01,1.6110330200195313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,1,power_law_1.2,0.1081113624572754
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.01,2.086316833496094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,1,power_law_1.2,0.36043392181396483
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,1,power_law_1.2,0.09936575889587403
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,1,power_law_1.2,0.5616831970214844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,1,power_law_1.2,0.1812556838989258
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,1,power_law_1.2,0.6375711822509766
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,1,power_law_1.2,0.5927116775512695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,1,power_law_1.2,0.29351871490478515
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,1,power_law_1.2,0.7253401947021485
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,1,power_law_1.2,0.7480268859863282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,1,power_law_1.2,0.8091014099121093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,1,power_law_1.2,0.6195852661132812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,1,power_law_1.2,0.8683315277099609
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,1,power_law_1.2,0.9200409698486329
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,1,power_law_1.2,0.6693881225585938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,1,power_law_1.2,1.2846995544433593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,1,power_law_1.2,1.086135711669922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,1,power_law_1.2,1.2586291503906248
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,1,power_law_1.2,1.3547608947753906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,1,power_law_1.2,1.6126617431640624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,1,power_law_1.2,1.8775917053222657
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,1,power_law_1.2,1.0558464050292968
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,1,power_law_1.2,2.3066867065429686
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,1,power_law_1.2,2.792476806640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,1,power_law_1.2,0.11271167755126954
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,1,power_law_1.2,0.13179072380065918
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,1,power_law_1.2,0.1359500789642334
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,1,power_law_1.2,0.19439424514770506
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,1,power_law_1.2,3.7600286865234374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,1,power_law_1.2,0.23946752548217773
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,1,power_law_1.2,4.7285159301757815
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,1,power_law_1.2,0.3021023941040039
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,1,power_law_1.2,0.42572799682617185
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,4,1,power_law_1.2,5.6439208984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,1,power_law_1.2,0.44133567810058594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,1,power_law_1.2,0.4608448028564453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,1,power_law_1.2,0.4767155075073243
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,1,power_law_1.2,0.49913791656494144
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,1,power_law_1.2,0.5157062530517578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,1,power_law_1.2,0.5255744171142578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,1,power_law_1.2,0.5523884963989258
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,1,power_law_1.2,0.6151385498046875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,1,power_law_1.2,0.6380640029907226
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,1,power_law_1.2,0.6573907470703124
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,1,power_law_1.2,0.7004863739013671
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,1,power_law_1.2,0.7489555358886719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,1,power_law_1.2,0.8875647735595702
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,1,power_law_1.2,0.912163848876953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,1,power_law_1.2,1.0771654510498048
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,1,power_law_1.2,1.2639615631103516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,1,power_law_1.2,1.6065632629394533
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.01,3.0782891845703126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,1,power_law_1.2,1.9778279113769532
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.01,4.070316772460937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,1,power_law_1.2,0.058405761718750004
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,1,power_law_1.2,2.69323974609375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,1,power_law_1.2,0.07279744148254394
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,1,power_law_1.2,0.07285056114196778
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,4,1,power_law_1.01,5.023841247558594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,1,power_law_1.2,3.400517883300781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,1,power_law_1.2,0.13625920295715332
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,1,power_law_1.2,0.17406911849975587
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,1,power_law_1.2,0.10336000442504882
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,1,power_law_1.2,0.25797119140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,1,power_law_1.2,0.2747212791442871
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,4,1,power_law_1.2,4.126422424316407
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,1,power_law_1.2,0.28563135147094726
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,1,power_law_1.2,0.29667135238647463
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,1,power_law_1.2,0.3298553466796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,1,power_law_1.2,0.31594047546386717
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,1,power_law_1.2,0.3440512084960937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,1,power_law_1.2,0.3656889724731445
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,1,power_law_1.2,0.3845920181274414
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,1,power_law_1.2,0.40418304443359376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,1,power_law_1.2,0.4421964645385742
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.2,0.585984001159668
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.2,0.7393727874755859
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.2,0.9025138854980469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.2,1.245185317993164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,1,power_law_1.2,0.510711669921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.2,1.6127545166015627
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,1,power_law_1.01,0.028399360179901124
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,1,power_law_1.01,0.0442412805557251
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,1,power_law_1.01,0.03795968055725098
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,1,power_law_1.01,0.056985602378845215
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,1,power_law_1.01,0.07927103996276855
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,1,power_law_1.01,0.08083135604858398
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,1,power_law_1.01,0.09112319946289063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,1,power_law_1.01,0.09968640327453612
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.2,2.4276544189453126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,1,power_law_1.01,0.09974528312683105
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,1,power_law_1.01,0.10463423728942871
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,1,power_law_1.01,0.048473601341247556
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,1,power_law_1.01,0.11006272315979004
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,1,power_law_1.01,0.1181503963470459
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,1,power_law_1.01,0.10760640144348144
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,1,power_law_1.01,0.14013567924499512
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,1,power_law_1.01,0.16224704742431642
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,1,power_law_1.01,0.19261695861816405
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,1,power_law_1.01,0.2281657600402832
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,1,power_law_1.01,0.29738304138183597
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,1,power_law_1.01,0.3712307357788086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,1,power_law_1.01,0.5157382583618164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,1,power_law_1.01,0.6619129943847656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,1,power_law_1.01,0.9533331298828125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,1,power_law_1.01,1.2440953826904297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,4,1,power_law_1.01,1.5339474487304687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,1,power_law_1.01,0.03861119985580445
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,1,power_law_1.01,0.040240001678466794
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,1,power_law_1.01,0.048219518661499025
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,1,power_law_1.01,0.07527935981750489
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,1,power_law_1.01,0.0841004753112793
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,1,power_law_1.01,0.08348095893859864
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,1,power_law_1.01,0.05268288135528565
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,1,power_law_1.01,0.08720704078674316
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,1,power_law_1.01,0.08613311767578125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,1,power_law_1.01,0.08841856002807617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,1,power_law_1.01,0.056524801254272464
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,1,power_law_1.01,0.10108608245849608
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,1,power_law_1.01,0.10034496307373046
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,1,power_law_1.01,0.10407168388366698
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,1,power_law_1.01,0.10471039772033693
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,1,power_law_1.01,0.10901951789855957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,1,power_law_1.01,0.11568256378173829
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,1,power_law_1.01,0.12447296142578126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,1,power_law_1.01,0.14116928100585938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,1,power_law_1.01,0.16396352767944336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,1,power_law_1.01,0.1927008056640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,1,power_law_1.01,0.24067455291748047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,1,power_law_1.01,0.2940671920776367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,1,power_law_1.01,0.4027212905883789
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,1,power_law_1.01,0.5092428970336914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,1,power_law_1.01,0.7244242858886719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,1,power_law_1.01,0.9363308715820313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,4,1,power_law_1.01,1.1489727783203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.2,3.248262939453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,1,power_law_1.01,0.037360000610351565
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,1,power_law_1.01,0.059617919921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,1,power_law_1.01,0.04036992073059082
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,1,power_law_1.01,0.06185728073120117
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,1,power_law_1.01,0.04414912223815918
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,1,power_law_1.01,0.04628608226776123
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,1,power_law_1.01,0.06328256130218506
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,1,power_law_1.01,0.06440959930419922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,1,power_law_1.01,0.06437248229980469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,1,power_law_1.01,0.06582975864410401
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,1,power_law_1.01,0.06783423900604249
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,1,power_law_1.01,0.0691923189163208
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,1,power_law_1.01,0.07149119853973389
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,1,power_law_1.01,0.05288576126098633
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,1,power_law_1.01,0.07471168041229248
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,1,power_law_1.01,0.09768896102905274
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,1,power_law_1.01,0.07771008014678955
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,1,power_law_1.01,0.08368639945983887
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,1,power_law_1.01,0.11103360176086426
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,1,power_law_1.01,0.17218048095703126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,1,power_law_1.01,0.14040127754211426
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,1,power_law_1.01,0.23346752166748047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,1,power_law_1.01,0.29788799285888673
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,1,power_law_1.01,0.4329830551147461
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,1,power_law_1.2,0.03166016101837158
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,1,power_law_1.2,0.03630592107772827
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,1,power_law_1.2,0.04339712142944336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,1,power_law_1.2,0.04770944118499756
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,1,power_law_1.2,0.05464128017425537
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,1,power_law_1.2,0.07655424118041992
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,1,power_law_1.2,0.08002176284790039
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,1,power_law_1.2,0.08228927612304687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,1,power_law_1.2,0.085730562210083
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,1,power_law_1.2,0.0858784008026123
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,1,power_law_1.2,0.0934598445892334
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,1,power_law_1.2,0.10034624099731446
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,1,power_law_1.2,0.10237567901611329
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,1,power_law_1.2,0.10645376205444337
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,1,power_law_1.2,0.1091001605987549
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,1,power_law_1.2,0.11095168113708495
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,1,power_law_1.2,0.11928064346313476
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,1,power_law_1.2,0.14110464096069336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,1,power_law_1.2,0.16279104232788086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,1,power_law_1.2,0.19368448257446289
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,1,power_law_1.2,0.22652992248535156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,1,power_law_1.2,0.29849920272827146
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,1,power_law_1.2,0.37199935913085935
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,1,power_law_1.2,0.5170073699951172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,1,power_law_1.2,0.6633952331542969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,1,power_law_1.01,0.5700806427001953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,1,power_law_1.2,0.9578816223144532
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,1,power_law_1.2,1.2479129791259767
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,1,power_law_1.2,0.05203648090362549
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,1,power_law_1.2,0.05587967872619629
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,1,power_law_1.2,0.07776383876800537
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,1,power_law_1.2,0.08329728126525879
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,1,power_law_1.2,0.04581183910369873
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,1,power_law_1.2,0.08594304084777832
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,1,power_law_1.01,0.8551904296874999
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,1,power_law_1.2,0.08751935958862304
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,1,power_law_1.2,0.0879321575164795
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,1,power_law_1.2,0.08742912292480469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,1,power_law_1.2,0.10166015625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,1,power_law_1.2,0.10526592254638671
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,1,power_law_1.01,1.1474253082275392
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,1,power_law_1.2,0.10848704338073731
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,1,power_law_1.2,0.1107436752319336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,1,power_law_1.2,0.11672191619873047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,1,power_law_1.2,0.13138303756713868
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,1,power_law_1.2,0.14377535820007323
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,1,power_law_1.2,0.16839807510375976
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,1,power_law_1.2,0.19738304138183593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,4,1,power_law_1.2,1.5347398376464843
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,1,power_law_1.2,0.24199487686157228
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,1,power_law_1.2,0.2979884719848633
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,1,power_law_1.2,0.40480895996093746
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,4,1,power_law_1.01,1.4371232604980468
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,1,power_law_1.2,0.10197888374328615
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,1,power_law_1.2,0.5092889785766601
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,1,power_law_1.2,0.03889663934707642
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,1,power_law_1.2,0.726635513305664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,1,power_law_1.2,0.038211839199066164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,1,power_law_1.2,0.04089024066925049
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,1,power_law_1.2,0.9411078643798827
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,1,power_law_1.2,0.04887551784515381
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,1,power_law_1.2,0.05938176155090332
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,4,1,power_law_1.2,1.1524806213378906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,1,power_law_1.2,0.03900160074234009
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,1,power_law_1.2,0.0633184003829956
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,1,power_law_1.2,0.0648524808883667
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,1,power_law_1.2,0.06257984161376953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,1,power_law_1.2,0.06538112163543701
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,1,power_law_1.2,0.0668838405609131
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,1,power_law_1.2,0.06868607997894287
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,1,power_law_1.2,0.07093503952026367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,1,power_law_1.2,0.07872511863708495
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,1,power_law_1.2,0.08203071594238281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,1,power_law_1.2,0.04399744033813476
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,1,power_law_1.2,0.042641921043395994
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,1,power_law_1.2,0.11047295570373535
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,1,power_law_1.2,0.07457791805267334
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,1,power_law_1.2,0.1307808017730713
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,1,power_law_1.2,0.0912063980102539
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,1,power_law_1.2,0.17493631362915038
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,1,power_law_1.2,0.22101823806762697
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,1,power_law_1.01,0.04749311923980713
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,1,power_law_1.01,0.06246719837188721
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,1,power_law_1.01,0.07842368125915528
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,1,power_law_1.01,0.10285759925842286
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,1,power_law_1.01,0.12800064086914062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,1,power_law_1.01,0.2093292808532715
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,1,power_law_1.01,0.22328128814697265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,1,power_law_1.01,0.22877632141113283
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,1,power_law_1.01,0.2310643196105957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,1,power_law_1.01,0.23417152404785155
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,1,power_law_1.01,0.2486803245544434
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,1,power_law_1.01,0.267063045501709
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,1,power_law_1.01,0.26846336364746093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,1,power_law_1.01,0.27621759414672853
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,1,power_law_1.01,0.28022016525268556
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,1,power_law_1.01,0.285404167175293
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,1,power_law_1.01,0.30472383499145506
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,1,power_law_1.2,0.3177484893798828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,1,power_law_1.01,0.36386112213134764
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,1,power_law_1.01,0.393570556640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,1,power_law_1.01,0.48768062591552735
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,1,power_law_1.2,0.41357952117919916
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,1,power_law_1.01,0.5795743942260743
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,1,power_law_1.01,0.764554214477539
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,1,power_law_1.01,0.957183380126953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,1,power_law_1.2,0.7998790740966797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,1,power_law_1.2,1.184307861328125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,1,power_law_1.01,1.3751654052734374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,1,power_law_1.2,0.6099564743041992
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,1,power_law_1.01,1.7631443786621095
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,1,power_law_1.01,0.06260799884796142
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,1,power_law_1.01,0.07406271934509277
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,1,power_law_1.01,0.054952960014343265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,1,power_law_1.01,0.08779456138610839
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,1,power_law_1.2,1.5759065246582031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,1,power_law_1.01,0.11190336227416993
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,1,power_law_1.01,2.5865396118164066
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,1,power_law_1.01,0.16013952255249025
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,1,power_law_1.01,0.1654047966003418
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,4,1,power_law_1.2,1.9684524536132812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,1,power_law_1.01,0.1675449562072754
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,1,power_law_1.01,0.1709267234802246
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,1,power_law_1.01,0.17097087860107424
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,1,power_law_1.01,0.14813440322875976
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,1,power_law_1.01,0.1959878349304199
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,1,power_law_1.01,3.3902835083007816
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,1,power_law_1.01,0.2025388717651367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,1,power_law_1.01,0.19611839294433592
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,1,power_law_1.01,0.2059980773925781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,1,power_law_1.01,0.20849151611328126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,1,power_law_1.01,0.21943296432495116
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,1,power_law_1.01,0.2370310401916504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,1,power_law_1.01,0.26615936279296876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,1,power_law_1.01,0.30107711791992187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,1,power_law_1.01,0.35680255889892576
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,1,power_law_1.01,0.44919872283935547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,1,power_law_1.01,0.5429248046875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,1,power_law_1.01,0.7374674987792968
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,4,1,power_law_1.01,4.172085266113281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,1,power_law_1.01,0.9299571228027343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,1,power_law_1.01,1.318396759033203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,1,power_law_1.01,1.7102438354492187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,4,1,power_law_1.01,2.0932672119140627
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,1,power_law_1.01,0.03667968034744263
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,1,power_law_1.01,0.042688641548156735
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,1,power_law_1.01,0.04871103763580322
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,1,power_law_1.01,0.10945023536682127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,1,power_law_1.01,0.07153600215911865
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,1,power_law_1.01,0.1107040023803711
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,1,power_law_1.01,0.06039231777191162
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,1,power_law_1.01,0.11522239685058593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,1,power_law_1.01,0.11835712432861327
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,1,power_law_1.01,0.11911744117736815
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,1,power_law_1.01,0.12799615859985353
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,1,power_law_1.01,0.10634048461914063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,1,power_law_1.01,0.10362560272216796
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,1,power_law_1.01,0.09680895805358887
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,1,power_law_1.01,0.14162367820739746
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,1,power_law_1.01,0.1354355239868164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,1,power_law_1.01,0.18722368240356446
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.01,0.21598527908325194
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.01,0.2824102401733398
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,1,power_law_1.01,0.15699968338012696
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.01,0.3546803283691406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.01,0.4999673461914062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,1,power_law_1.2,0.06046144008636475
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,1,power_law_1.2,0.07456448078155517
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,1,power_law_1.2,0.0939743995666504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,1,power_law_1.2,0.11815360069274902
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,1,power_law_1.2,0.2204051208496094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,1,power_law_1.2,0.23274368286132813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,1,power_law_1.2,0.05221183776855469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,1,power_law_1.2,0.23470848083496093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,1,power_law_1.2,0.23548095703125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,1,power_law_1.2,0.2368998336791992
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,1,power_law_1.2,0.25290559768676757
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,1,power_law_1.2,0.26922624588012695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,1,power_law_1.2,0.2733600044250488
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,1,power_law_1.2,0.2791315269470215
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,1,power_law_1.2,0.28584703445434567
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,1,power_law_1.2,0.2878630447387695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,1,power_law_1.2,0.32395774841308594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,1,power_law_1.2,0.36705856323242186
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,1,power_law_1.2,0.398416633605957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,1,power_law_1.2,0.4994611358642578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,1,power_law_1.2,0.5764588928222656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,1,power_law_1.2,0.7695442962646484
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,1,power_law_1.2,0.9726278686523437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,1,power_law_1.2,1.3841625976562502
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,1,power_law_1.2,1.775910339355469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,1,power_law_1.2,2.5883566284179684
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.01,0.6480198669433593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,1,power_law_1.2,0.056179199218749996
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.01,0.955506591796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.01,1.2691846466064454
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,1,power_law_1.2,0.06268479824066162
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,1,power_law_1.2,0.07061503887176514
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.01,1.9141261291503906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,1,power_law_1.2,3.396970825195312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,1,power_law_1.2,0.08476479530334473
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,1,power_law_1.2,0.10251839637756348
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,1,power_law_1.2,0.14902015686035158
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,1,power_law_1.2,0.1615180778503418
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,1,power_law_1.2,0.16654016494750976
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,1,power_law_1.2,0.1688287925720215
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.01,2.576399230957031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,1,power_law_1.2,0.16935552597045897
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,1,power_law_1.2,0.17242496490478515
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,1,power_law_1.2,0.19578496932983397
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,1,power_law_1.2,0.1990969657897949
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,1,power_law_1.2,0.20510208129882815
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,1,power_law_1.2,0.21153343200683591
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,1,power_law_1.2,0.20905344009399415
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,4,1,power_law_1.2,4.215210876464844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,1,power_law_1.2,0.22174272537231446
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,1,power_law_1.2,0.24807104110717773
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,1,power_law_1.2,0.26840063095092775
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,1,power_law_1.2,0.31556928634643555
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,1,power_law_1.2,0.36000064849853514
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,1,power_law_1.2,0.45035518646240236
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,1,power_law_1.2,0.5573529434204102
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.01,3.260089111328125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,1,power_law_1.2,0.7416806030273437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,1,power_law_1.2,0.03769727945327759
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,1,power_law_1.2,0.9365023803710937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,1,power_law_1.2,0.04312895774841309
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,1,power_law_1.2,0.05717440128326416
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,1,power_law_1.2,0.04974400043487549
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,1,power_law_1.2,1.3225382995605468
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,1,power_law_1.2,0.06787712097167968
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,1,power_law_1.2,0.09895039558410644
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,1,power_law_1.2,0.10562879562377929
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,1,power_law_1.2,1.7161048889160155
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,1,power_law_1.2,0.10755135536193847
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,1,power_law_1.2,0.11132991790771485
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,1,power_law_1.2,0.11272128105163574
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,1,power_law_1.2,0.1166425609588623
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,1,power_law_1.2,0.12195199966430663
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,4,1,power_law_1.2,2.104206085205078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,1,power_law_1.2,0.12553919792175292
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,1,power_law_1.2,0.14387392044067382
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,1,power_law_1.2,0.1712179183959961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,1,power_law_1.2,0.1345107173919678
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,1,power_law_1.2,0.1530016040802002
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.2,0.25744255065917965
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.2,0.3542675018310547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.2,0.4534156799316406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,1,power_law_1.2,0.2134630393981934
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,1,power_law_1.01,0.0967084789276123
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.2,0.6636799621582031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,1,power_law_1.01,0.134204158782959
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,1,power_law_1.01,0.10972928047180175
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,1,power_law_1.01,0.22833984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,1,power_law_1.01,0.3247577667236328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,1,power_law_1.01,0.49933761596679693
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,1,power_law_1.01,0.5211180877685547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,1,power_law_1.01,0.5546700668334961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,1,power_law_1.01,0.585041275024414
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,1,power_law_1.01,0.6050604629516602
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,1,power_law_1.01,0.6313638305664062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,1,power_law_1.01,0.6752889251708984
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,1,power_law_1.01,0.7383968353271484
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,1,power_law_1.01,0.7622303771972656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,1,power_law_1.01,0.7009056091308594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,1,power_law_1.01,0.7777196502685546
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,1,power_law_1.01,0.8099897766113282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,1,power_law_1.01,0.911022720336914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,1,power_law_1.01,1.0027993774414061
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,1,power_law_1.01,1.2210662078857422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,1,power_law_1.01,1.4385356140136718
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.2,0.8774272155761718
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,1,power_law_1.01,1.9267359924316405
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,1,power_law_1.01,2.366021728515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.2,1.7181484985351563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.2,1.2968255615234374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.2,2.576324462890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,1,power_law_1.01,3.303236999511719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.2,3.4435122680664065
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,1,power_law_1.01,4.265647888183594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,1,power_law_1.01,0.1192319965362549
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,1,power_law_1.01,0.10449215888977051
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,1,power_law_1.01,0.1693926429748535
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,1,power_law_1.01,0.21588735580444335
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.2,4.338913879394531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,1,power_law_1.01,0.32285438537597655
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,1,power_law_1.01,0.33670207977294925
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,1,power_law_1.01,0.35078590393066406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,1,power_law_1.01,0.08837120056152344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,1,power_law_1.01,0.3598284912109375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,1,power_law_1.01,0.3688467025756836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,1,power_law_1.01,6.149782104492187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,1,power_law_1.01,0.37579071044921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,1,power_law_1.01,0.38764606475830077
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,1,power_law_1.01,0.4301414489746094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,1,power_law_1.01,0.4389420700073242
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,1,power_law_1.01,0.4458412933349609
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,1,power_law_1.01,0.45757118225097654
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,1,power_law_1.01,0.477151985168457
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,1,power_law_1.01,0.510063362121582
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,1,power_law_1.01,0.5509439849853516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,1,power_law_1.01,0.624334716796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,1,power_law_1.01,0.7226233673095703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,1,power_law_1.01,0.9024800109863282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,1,power_law_1.01,8.085385131835938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,1,power_law_1.01,1.0822112274169922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,1,power_law_1.01,1.4619200134277344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,1,power_law_1.01,1.8141600036621093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,1,power_law_1.01,2.528401336669922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,4,1,power_law_1.01,9.871506958007812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,1,power_law_1.01,3.2896588134765623
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,4,1,power_law_1.01,4.013614807128906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,1,power_law_1.01,0.05845695972442627
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,1,power_law_1.01,0.09462143898010253
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,1,power_law_1.01,0.2164691162109375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,1,power_law_1.01,0.1204640007019043
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,1,power_law_1.01,0.18411584854125979
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,1,power_law_1.01,0.22196863174438475
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,1,power_law_1.01,0.22998464584350584
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,1,power_law_1.01,0.04478271961212158
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,1,power_law_1.01,0.06738687992095947
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,1,power_law_1.01,0.24076032638549805
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,1,power_law_1.01,0.23632192611694336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,1,power_law_1.01,0.24946495056152346
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,1,power_law_1.01,0.26115583419799804
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,1,power_law_1.01,0.27200960159301757
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,1,power_law_1.01,0.20824512481689453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,1,power_law_1.01,0.19762239456176758
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,1,power_law_1.01,0.29694400787353514
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,1,power_law_1.01,0.34598976135253906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.01,0.6122963333129883
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.01,0.8367865753173828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.01,1.0730054473876953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.01,0.39422527313232425
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,1,power_law_1.2,0.09632831573486328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.01,0.4957555389404297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.01,1.5657029724121094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,1,power_law_1.2,0.09961536407470703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,1,power_law_1.2,0.1353664016723633
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,1,power_law_1.2,0.19492351531982421
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,1,power_law_1.2,0.29174400329589845
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,1,power_law_1.2,0.49712894439697264
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,1,power_law_1.2,0.5277619171142578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,1,power_law_1.2,0.5712351989746094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,1,power_law_1.2,0.6015103912353515
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,1,power_law_1.2,0.6248524856567383
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,1,power_law_1.2,0.6571315002441407
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,1,power_law_1.2,0.7113209533691406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,1,power_law_1.2,0.7464243316650391
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,1,power_law_1.2,0.7701203155517578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,1,power_law_1.2,0.7917606353759765
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,1,power_law_1.2,0.795960922241211
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,1,power_law_1.2,0.8369132995605468
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,1,power_law_1.2,0.9421971130371094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,1,power_law_1.2,1.0064800262451172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,1,power_law_1.2,1.2378636932373046
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,1,power_law_1.2,1.4862905883789064
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,1,power_law_1.2,1.9632453918457031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,1,power_law_1.2,2.3737887573242187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,1,power_law_1.2,3.319412536621094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,1,power_law_1.2,4.273363952636719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,1,power_law_1.2,6.191161499023438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,1,power_law_1.2,0.09749247550964354
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,1,power_law_1.2,0.09183679580688477
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,1,power_law_1.2,0.11788607597351075
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,1,power_law_1.2,0.16005056381225585
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,1,power_law_1.2,8.060200805664063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,1,power_law_1.2,0.19222784042358398
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,1,power_law_1.2,0.32168575286865236
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,1,power_law_1.2,0.3385260772705078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,1,power_law_1.2,0.35561473846435543
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,1,power_law_1.2,0.36891902923583986
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,1,power_law_1.2,0.3762617492675781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,1,power_law_1.2,0.3841030502319336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,1,power_law_1.2,0.400445442199707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,1,power_law_1.2,0.45406654357910153
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,1,power_law_1.2,0.4598444747924805
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,4,1,power_law_1.2,9.971624145507812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,1,power_law_1.2,0.4695321655273437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,1,power_law_1.2,0.4760153579711914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,1,power_law_1.2,0.49260223388671875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,1,power_law_1.2,0.5250182342529297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,1,power_law_1.2,0.5614035034179687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.01,2.0683251953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,1,power_law_1.2,0.6431263732910156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,1,power_law_1.2,0.7370272064208985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,1,power_law_1.2,0.9257107543945313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,1,power_law_1.2,1.1048729705810547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,1,power_law_1.2,1.4657562255859375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.01,3.0804409790039062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,1,power_law_1.2,1.836974792480469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,1,power_law_1.2,0.044374399185180664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,1,power_law_1.2,0.053196158409118656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,1,power_law_1.2,2.546229705810547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,1,power_law_1.2,0.0667136001586914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,1,power_law_1.2,0.08691519737243653
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.01,4.137161560058594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,1,power_law_1.2,0.10957504272460938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,1,power_law_1.2,3.3103988647460936
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,1,power_law_1.2,0.19860416412353515
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,1,power_law_1.2,0.18264896392822266
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,1,power_law_1.2,0.21353727340698242
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,4,1,power_law_1.01,5.203138427734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,1,power_law_1.2,0.22760831832885745
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,1,power_law_1.2,0.22274303436279297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,4,1,power_law_1.2,4.020501708984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,1,power_law_1.2,0.23568895339965817
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,1,power_law_1.2,0.2454035186767578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,1,power_law_1.2,0.25523263931274415
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,1,power_law_1.2,0.26667968750000004
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,1,power_law_1.2,0.29869056701660157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,1,power_law_1.2,0.2843052864074707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,1,power_law_1.2,0.3295692825317383
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,1,power_law_1.2,0.3933209609985352
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.2,0.46507007598876954
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.2,0.6215411376953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.2,0.7982009887695313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.2,1.1518675231933595
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,1,power_law_1.01,0.12978176116943357
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,1,power_law_1.01,0.14714624404907226
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,1,power_law_1.01,0.2237343978881836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,1,power_law_1.01,0.42693630218505857
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,1,power_law_1.01,0.6024703979492188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,1,power_law_1.01,0.6885388946533203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,1,power_law_1.01,0.892980499267578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,1,power_law_1.01,0.9542515563964844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,1,power_law_1.01,0.987805404663086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,1,power_law_1.01,1.0322252655029298
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,1,power_law_1.01,1.0683251190185548
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,1,power_law_1.01,1.1140268707275391
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,1,power_law_1.01,1.2016397094726563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,1,power_law_1.01,1.3226527404785156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,1,power_law_1.01,1.4254917907714844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,1,power_law_1.01,1.3677830505371094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,1,power_law_1.01,1.5124447631835938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,1,power_law_1.01,1.5973727416992187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,1,power_law_1.01,1.8413912963867187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,1,power_law_1.01,2.0620736694335937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.2,2.3550643920898438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.2,1.530789794921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,1,power_law_1.01,2.566168212890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.2,3.18035400390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,1,power_law_1.01,2.957679443359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.2,4.922785339355469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,1,power_law_1.01,3.794547729492187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,1,power_law_1.01,0.15164799690246583
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,1,power_law_1.01,4.63890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,1,power_law_1.01,0.17313791275024415
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.2,6.597686767578125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,1,power_law_1.01,0.20446720123291015
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,1,power_law_1.01,0.2307027244567871
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,1,power_law_1.01,0.3260940933227539
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,1,power_law_1.01,0.4163590240478515
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,1,power_law_1.01,0.47342529296875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,1,power_law_1.01,6.586287841796874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,1,power_law_1.01,0.588078727722168
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,1,power_law_1.01,0.6185305786132813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,1,power_law_1.01,0.6338886260986328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,1,power_law_1.01,0.6605356597900391
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,1,power_law_1.01,0.6774976348876953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,1,power_law_1.01,0.6960095977783203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,1,power_law_1.01,8.468236083984376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,1,power_law_1.01,0.7211289978027343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,1,power_law_1.01,0.737567367553711
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,1,power_law_1.01,0.7562598419189454
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,1,power_law_1.01,0.8614080047607422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,1,power_law_1.01,0.8954822540283203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,4,1,power_law_1.01,10.168102416992188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,1,power_law_1.01,0.9332717132568359
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,1,power_law_1.01,1.0236691284179689
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,1,power_law_1.01,1.1159117126464844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,1,power_law_1.01,1.3179718017578126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,1,power_law_1.01,1.4807565307617188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,1,power_law_1.01,1.875383605957031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,1,power_law_1.01,2.264610595703125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,1,power_law_1.01,3.017134094238281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,1,power_law_1.01,3.7927865600585937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,4,1,power_law_1.01,4.55254150390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,1,power_law_1.01,0.07791552066802979
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,1,power_law_1.01,0.07866047859191895
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,1,power_law_1.01,0.07935872077941894
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,1,power_law_1.01,0.08033791542053223
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,1,power_law_1.01,0.11747455596923828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,1,power_law_1.01,0.13137727737426758
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,1,power_law_1.01,0.13735872268676758
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,1,power_law_1.01,0.09371456146240234
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,1,power_law_1.01,0.15024895668029786
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,1,power_law_1.01,0.10506815910339355
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,1,power_law_1.01,0.08491519927978516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,1,power_law_1.01,0.19665407180786132
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,1,power_law_1.01,0.21969984054565428
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,1,power_law_1.01,0.26383935928344726
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,1,power_law_1.01,0.3661721420288086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,1,power_law_1.01,0.6528012847900391
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.01,0.8501222229003906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,1,power_law_1.01,0.3209983825683594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,1,power_law_1.01,0.45919040679931644
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.01,1.2465420532226563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.01,1.6361465454101562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,1,power_law_1.2,0.16637439727783204
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.01,2.4060127258300783
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,1,power_law_1.2,0.23044095993041994
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,1,power_law_1.2,0.40449344635009765
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,1,power_law_1.2,0.5525190353393554
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,1,power_law_1.2,0.6126745605468751
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,1,power_law_1.2,0.838124771118164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.01,3.2001358032226563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.01,4.761078491210937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,1,power_law_1.2,0.9063763427734376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,1,power_law_1.2,0.944106903076172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,1,power_law_1.2,1.0025183868408203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,1,power_law_1.2,1.0563040161132813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,1,power_law_1.2,1.1926675415039063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,1,power_law_1.2,1.0933382415771484
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,1,power_law_1.2,1.3027436828613281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,1,power_law_1.2,1.4054637145996094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,1,power_law_1.2,1.4135897827148436
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,1,power_law_1.2,1.552737274169922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,1,power_law_1.2,1.6799853515624998
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,1,power_law_1.2,1.8885151672363283
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,1,power_law_1.2,2.129423370361328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,1,power_law_1.2,2.6108331298828125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,1,power_law_1.2,3.0468344116210937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,1,power_law_1.2,3.907185363769531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,1,power_law_1.2,0.15017024040222168
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,1,power_law_1.2,4.788839111328125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,1,power_law_1.2,0.1724620819091797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,1,power_law_1.2,0.1849273681640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,1,power_law_1.2,0.2334367942810059
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,1,power_law_1.2,0.30669120788574217
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,1,power_law_1.2,6.668453979492187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,1,power_law_1.2,0.40105663299560546
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,1,power_law_1.2,0.4400153732299804
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,1,power_law_1.2,0.5588768005371094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,1,power_law_1.2,8.607363281249999
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,1,power_law_1.2,0.5916870498657227
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,1,power_law_1.2,0.6167116928100586
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,4,1,power_law_1.2,10.30038818359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,1,power_law_1.2,0.6448467254638672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,1,power_law_1.2,0.6829228973388671
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,1,power_law_1.2,0.7110495758056641
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,1,power_law_1.2,0.665499496459961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,1,power_law_1.2,0.7543814086914062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,1,power_law_1.2,0.8559410858154297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,1,power_law_1.2,0.7283897399902344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,1,power_law_1.2,0.9156646728515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,1,power_law_1.2,0.9737190246582031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,1,power_law_1.2,1.0740684509277343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,1,power_law_1.2,1.1733779144287109
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,1,power_law_1.2,1.3923199462890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,1,power_law_1.2,1.5745893859863282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,1,power_law_1.2,1.9560850524902342
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,1,power_law_1.2,2.342395477294922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,1,power_law_1.2,3.0904083251953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,1,power_law_1.2,0.07750336170196534
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,1,power_law_1.2,0.07831744194030762
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,1,power_law_1.2,3.904578552246094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,1,power_law_1.2,0.07970431804656983
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,1,power_law_1.2,0.08034624099731445
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,4,1,power_law_1.2,4.638396301269531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,1,power_law_1.2,0.08547200202941894
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,1,power_law_1.2,0.09409024238586425
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,1,power_law_1.2,0.10207039833068847
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,1,power_law_1.2,0.11950592041015624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,1,power_law_1.2,0.1271449565887451
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,1,power_law_1.2,0.13520383834838867
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,1,power_law_1.2,0.15845248222351074
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,1,power_law_1.2,0.22021120071411132
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,1,power_law_1.2,0.2652095985412598
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,1,power_law_1.2,0.19826496124267579
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,1,power_law_1.2,0.32026496887207034
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,1,power_law_1.2,0.3662688064575195
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,1,power_law_1.2,0.46066238403320314
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,1,power_law_1.2,0.6505427551269531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.2,4.759081726074219
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,1,power_law_1.01,0.07679679870605469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,1,power_law_1.01,0.06841663837432861
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,1,power_law_1.01,0.0719814395904541
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,1,power_law_1.01,0.09055616378784179
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,1,power_law_1.01,0.11959551811218261
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,1,power_law_1.01,0.12102848052978517
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,1,power_law_1.01,0.12204480171203613
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,1,power_law_1.01,0.1219372844696045
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,1,power_law_1.01,0.12300800323486327
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,1,power_law_1.01,0.12458687782287599
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,1,power_law_1.01,0.12881407737731934
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,1,power_law_1.01,0.13087807655334474
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,1,power_law_1.01,0.1363308811187744
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,1,power_law_1.01,0.14148863792419433
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,1,power_law_1.01,0.1471449565887451
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,1,power_law_1.01,0.18656639099121092
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,1,power_law_1.01,0.21370431900024417
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,1,power_law_1.01,0.24729856491088867
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,1,power_law_1.01,0.29300928115844727
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,1,power_law_1.01,0.3826598358154297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,1,power_law_1.01,0.4750265502929688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,1,power_law_1.01,0.6639014434814453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,1,power_law_1.01,0.8544300842285157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.2,1.2490451049804687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,1,power_law_1.01,1.2738540649414063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,1,power_law_1.01,2.4911155700683594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.2,0.8447564697265625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,1,power_law_1.01,1.6373548889160154
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.2,1.6312486267089845
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.2,2.4158029174804687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.2,3.180738525390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,1,power_law_1.01,3.370348205566406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,1,power_law_1.01,0.07577600002288817
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,1,power_law_1.01,0.0639680004119873
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,1,power_law_1.01,0.04291135787963867
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,1,power_law_1.01,0.06495872020721435
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,1,power_law_1.01,0.0661030387878418
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,1,power_law_1.01,0.0460211181640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,1,power_law_1.01,0.06724671840667724
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,1,power_law_1.01,0.069552001953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,1,power_law_1.01,0.06525568008422852
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,1,power_law_1.01,0.07085311889648438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,1,power_law_1.01,0.07904575824737549
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,1,power_law_1.01,0.08467967987060547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,1,power_law_1.01,0.09624832153320313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,1,power_law_1.01,0.09760831832885743
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,1,power_law_1.01,0.1060422420501709
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,1,power_law_1.01,0.12605759620666504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,1,power_law_1.01,0.15836735725402834
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,1,power_law_1.01,0.1928659248352051
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,1,power_law_1.01,0.052334079742431636
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,1,power_law_1.01,0.25975679397583007
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,1,power_law_1.01,0.330843505859375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,1,power_law_1.01,0.06412479877471924
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,1,power_law_1.01,0.46414848327636715
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,1,power_law_1.01,0.6077638244628906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,8,1,power_law_1.01,4.092319946289062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,1,power_law_1.01,0.0656108808517456
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,1,power_law_1.01,0.8853286743164063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,1,power_law_1.01,1.1700057220458984
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,8,1,power_law_1.01,1.45474365234375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,1,power_law_1.01,0.02252095937728882
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,1,power_law_1.01,0.0377728009223938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,1,power_law_1.01,0.025143039226531983
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,1,power_law_1.01,0.03696831941604614
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,1,power_law_1.01,0.020968320369720458
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,1,power_law_1.01,0.03714623928070068
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,1,power_law_1.01,0.038197119235992436
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,1,power_law_1.01,0.03713664054870606
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,1,power_law_1.01,0.03865344047546387
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,1,power_law_1.01,0.0377836799621582
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,1,power_law_1.01,0.03746047973632812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,1,power_law_1.01,0.04123519897460938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,1,power_law_1.01,0.038499839305877685
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,1,power_law_1.01,0.03947776079177857
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,1,power_law_1.01,0.049663357734680176
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,1,power_law_1.01,0.04343103885650635
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,1,power_law_1.01,0.04598911762237549
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,1,power_law_1.01,0.05831999778747558
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.01,0.08548671722412109
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.01,0.103253116607666
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.01,0.06709887981414794
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.01,0.14002752304077148
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.01,0.17753023147583008
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,1,power_law_1.2,0.06688511848449707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,1,power_law_1.2,0.0691590404510498
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,1,power_law_1.2,0.09072511672973634
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,1,power_law_1.2,0.11753536224365235
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,1,power_law_1.2,0.12058176040649413
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,1,power_law_1.2,0.12093952178955078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,1,power_law_1.2,0.05882175922393799
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,1,power_law_1.2,0.12209983825683594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,1,power_law_1.2,0.12316608428955078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,1,power_law_1.2,0.1238924789428711
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,1,power_law_1.2,0.12893823623657225
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,1,power_law_1.2,0.12953856468200683
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,1,power_law_1.2,0.13484479904174804
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,1,power_law_1.2,0.14219008445739748
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,1,power_law_1.2,0.14795519828796386
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,1,power_law_1.2,0.19825279235839843
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,1,power_law_1.2,0.21423807144165039
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,1,power_law_1.2,0.2482310485839844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,1,power_law_1.2,0.29539072036743164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,1,power_law_1.2,0.38633342742919924
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,1,power_law_1.2,0.47768383026123046
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,1,power_law_1.2,0.6639321899414062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,1,power_law_1.2,0.854471664428711
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,1,power_law_1.2,1.2313247680664063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,1,power_law_1.2,1.6913133239746094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,1,power_law_1.2,2.4960295104980466
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,1,power_law_1.2,3.344461364746094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.01,0.2524857521057129
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,1,power_law_1.2,0.05028160095214844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,1,power_law_1.2,0.06091904163360595
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,1,power_law_1.2,0.06284671783447265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,1,power_law_1.2,0.06523007869720458
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,1,power_law_1.2,0.06559679985046388
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,1,power_law_1.2,0.04521152019500733
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.01,0.3289440155029297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,1,power_law_1.2,0.06628928184509278
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,1,power_law_1.2,0.06816448211669922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,8,1,power_law_1.2,4.192186279296875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,1,power_law_1.2,0.06963391780853272
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,1,power_law_1.2,0.07117184162139892
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,1,power_law_1.2,0.07177984237670898
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,1,power_law_1.2,0.07846911907196044
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,1,power_law_1.2,0.0879462432861328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,1,power_law_1.2,0.100512638092041
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,1,power_law_1.2,0.09812159538269043
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,1,power_law_1.2,0.10614784240722656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,1,power_law_1.2,0.12734399795532228
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,1,power_law_1.2,0.15554176330566408
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,1,power_law_1.2,0.19375423431396482
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,1,power_law_1.2,0.2596479988098145
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,1,power_law_1.2,0.04382207870483398
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,1,power_law_1.2,0.3320544052124023
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,1,power_law_1.2,0.4678131103515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,1,power_law_1.2,0.6065343856811524
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,1,power_law_1.2,0.021073920726776125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,1,power_law_1.2,0.02336127996444702
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,1,power_law_1.2,0.8813433837890624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,1,power_law_1.2,0.02470207929611206
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,1,power_law_1.2,0.0644927978515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,1,power_law_1.2,1.1576211547851563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,1,power_law_1.2,0.03569472074508667
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,1,power_law_1.2,0.03825023889541626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,1,power_law_1.2,0.037419519424438476
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,1,power_law_1.2,0.038496639728546146
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,8,1,power_law_1.2,1.4444557189941407
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,1,power_law_1.2,0.03823744058609009
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,1,power_law_1.2,0.038609919548034666
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,1,power_law_1.2,0.03904000043869019
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.01,0.48329345703125004
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.01,0.8029964447021485
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.01,0.6372268676757813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,1,power_law_1.2,0.050716800689697264
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,1,power_law_1.2,0.036755840778350826
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,1,power_law_1.2,0.03651776075363159
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,1,power_law_1.2,0.039584639072418215
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,1,power_law_1.2,0.04372928142547607
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,1,power_law_1.2,0.0413753604888916
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.2,0.1455667209625244
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.2,0.06784895896911622
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,1,power_law_1.2,0.046776318550109865
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,1,power_law_1.2,0.05933055877685547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.2,0.08735296249389649
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,1,power_law_1.01,0.07743296146392822
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,1,power_law_1.01,0.08908224105834961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,1,power_law_1.01,0.09660991668701172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,1,power_law_1.01,0.12971776008605956
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,1,power_law_1.01,0.163384952545166
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.2,0.2629644775390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,1,power_law_1.01,0.16495296478271484
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,1,power_law_1.01,0.1675040054321289
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.2,0.3411814498901367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.2,0.18296064376831053
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,1,power_law_1.01,0.16585727691650393
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,1,power_law_1.01,0.16791679382324218
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,1,power_law_1.01,0.16973823547363281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,1,power_law_1.01,0.18427007675170898
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,1,power_law_1.01,0.19449344635009766
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,1,power_law_1.01,0.20846847534179686
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,1,power_law_1.01,0.2182694435119629
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,1,power_law_1.01,0.2429529571533203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.2,0.10536640167236329
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,1,power_law_1.01,0.2868838310241699
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,1,power_law_1.01,0.31227840423583986
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,1,power_law_1.01,0.38479358673095704
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,1,power_law_1.01,0.4661286544799805
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,1,power_law_1.01,0.6168985748291016
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,1,power_law_1.01,0.7740223693847657
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,1,power_law_1.01,1.09512451171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,1,power_law_1.01,1.4422930908203124
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,1,power_law_1.01,2.157085418701172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,1,power_law_1.01,0.053877758979797366
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,1,power_law_1.01,0.05992767810821533
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,1,power_law_1.01,0.06296063899993896
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,1,power_law_1.01,2.848497314453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,1,power_law_1.01,0.08285504341125488
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,1,power_law_1.01,0.09552255630493164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,1,power_law_1.01,0.09136639595031738
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,1,power_law_1.01,0.09684160232543945
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,1,power_law_1.01,0.09173376083374023
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,1,power_law_1.01,0.09546367645263672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,1,power_law_1.01,0.09361472129821777
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,1,power_law_1.01,0.09752639770507812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,1,power_law_1.01,0.09587519645690919
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,1,power_law_1.01,0.10112000465393067
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,1,power_law_1.01,0.11260607719421387
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,1,power_law_1.01,4.285322875976563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,1,power_law_1.01,0.11964991569519043
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,1,power_law_1.01,0.12085375785827637
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,1,power_law_1.01,0.13396736145019533
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,1,power_law_1.01,0.14611264228820803
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,1,power_law_1.01,0.18168767929077148
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,1,power_law_1.01,0.23070335388183594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,1,power_law_1.01,0.27193216323852537
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.2,0.49887615203857416
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,1,power_law_1.01,0.3723545455932617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,1,power_law_1.01,0.4772499084472656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.2,0.6597856140136719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.2,0.8196614074707032
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,1,power_law_1.01,0.6764198303222656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,1,power_law_1.01,5.7125006103515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,1,power_law_1.01,0.9085420989990235
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,1,power_law_1.01,1.3578099060058595
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,1,power_law_1.01,1.799085388183594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,8,1,power_law_1.01,7.1667041015625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,8,1,power_law_1.01,2.216705932617187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,1,power_law_1.01,0.025763199329376218
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,1,power_law_1.01,0.029955840110778807
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,1,power_law_1.01,0.047931518554687504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,1,power_law_1.01,0.03537280082702637
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,1,power_law_1.01,0.04858240127563477
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,1,power_law_1.01,0.04931839942932129
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,1,power_law_1.01,0.049745922088623044
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,1,power_law_1.01,0.04834432125091553
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,1,power_law_1.01,0.04846911907196045
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,1,power_law_1.01,0.04848896026611328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,1,power_law_1.01,0.04829823970794678
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,1,power_law_1.01,0.051584000587463374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,1,power_law_1.01,0.06064383983612061
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,1,power_law_1.01,0.06695231914520264
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,1,power_law_1.01,0.059618558883666996
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,1,power_law_1.01,0.05735743999481201
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,1,power_law_1.01,0.08077183723449707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,1,power_law_1.01,0.05215104103088379
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.01,0.0947327995300293
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.01,0.149934720993042
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.01,0.2622713661193848
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.01,0.12251392364501953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,1,power_law_1.2,0.07935935974121093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.01,0.37667007446289064
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.01,0.4971481704711914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,1,power_law_1.2,0.08783295631408691
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,1,power_law_1.2,0.09515135765075684
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,1,power_law_1.2,0.13025792121887209
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,1,power_law_1.2,0.16552576065063476
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,1,power_law_1.2,0.16871936798095705
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,1,power_law_1.2,0.16618431091308594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,1,power_law_1.2,0.16654912948608397
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.01,0.7256716918945313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,1,power_law_1.2,0.16787584304809572
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,1,power_law_1.2,0.17169599533081054
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,1,power_law_1.2,0.18682367324829102
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,1,power_law_1.2,0.19570240020751953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,1,power_law_1.2,0.21174463272094726
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,1,power_law_1.2,0.22158592224121093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,1,power_law_1.2,0.24432256698608396
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,1,power_law_1.2,0.29764223098754883
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.01,0.20632127761840818
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,1,power_law_1.2,0.31225856781005856
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,1,power_law_1.2,0.38610111236572264
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,1,power_law_1.2,0.46929088592529294
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,1,power_law_1.2,0.6160435104370118
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,1,power_law_1.2,0.7813088226318359
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,1,power_law_1.2,1.100338592529297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,1,power_law_1.2,1.4531890869140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.01,0.9988671874999999
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,1,power_law_1.2,2.148356475830078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,1,power_law_1.2,0.05495168209075928
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,1,power_law_1.2,0.05974143981933594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,1,power_law_1.2,0.06291584014892579
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,1,power_law_1.2,0.08429375648498535
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,1,power_law_1.2,0.08985856056213379
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,1,power_law_1.2,0.0934879970550537
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,1,power_law_1.2,2.819512939453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,1,power_law_1.2,0.09209152221679687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,1,power_law_1.2,0.09330752372741699
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,1,power_law_1.2,0.09526335716247558
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,1,power_law_1.2,0.09504768371582031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,1,power_law_1.2,0.0982688045501709
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,1,power_law_1.2,0.09720319747924805
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,1,power_law_1.2,0.09905471801757812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,1,power_law_1.2,0.11243647575378417
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,1,power_law_1.2,0.12550335884094238
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,1,power_law_1.2,0.12678463935852052
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,1,power_law_1.2,0.13333184242248536
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,1,power_law_1.2,0.14552512168884277
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,1,power_law_1.2,0.18247295379638673
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,1,power_law_1.2,0.2262892723083496
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,1,power_law_1.2,0.27347455978393553
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,1,power_law_1.2,0.37211071014404296
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,1,power_law_1.2,4.210418701171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,1,power_law_1.2,0.4766764831542969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,1,power_law_1.2,0.6869261169433594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,1,power_law_1.2,0.026746239662170414
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,1,power_law_1.2,0.9106374359130859
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,1,power_law_1.2,0.028497281074523928
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,1,power_law_1.2,0.03285887956619263
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,1,power_law_1.2,0.047689599990844725
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,1,power_law_1.2,0.04734655857086182
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,1,power_law_1.2,1.3428466796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,1,power_law_1.2,0.048824319839477534
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,1,power_law_1.2,0.048798079490661624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,1,power_law_1.2,5.697976684570312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,1,power_law_1.2,0.04892416000366211
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,1,power_law_1.2,0.04886528015136719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,1,power_law_1.2,0.04916416168212891
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,1,power_law_1.2,0.05113855838775635
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,1,power_law_1.2,0.051825280189514156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,1,power_law_1.2,1.8189120483398438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,1,power_law_1.2,0.061781120300292966
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,8,1,power_law_1.2,2.2078175354003906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,8,1,power_law_1.2,7.44721435546875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.01,1.2445081329345702
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.2,0.16375232696533204
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,1,power_law_1.2,0.05209343910217286
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,1,power_law_1.2,0.06260096073150635
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,1,power_law_1.2,0.08180607795715332
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,1,power_law_1.2,0.05768383979797363
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.2,0.09564224243164063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,1,power_law_1.2,0.06743296146392821
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.2,0.12394944190979004
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,1,power_law_1.01,0.06502272129058838
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,1,power_law_1.01,0.08337151527404785
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,1,power_law_1.01,0.08793791770935058
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,1,power_law_1.01,0.13203264236450196
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,1,power_law_1.01,0.183623046875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,1,power_law_1.01,0.2726176071166992
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,1,power_law_1.01,0.33551937103271484
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,1,power_law_1.01,0.34639102935791016
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,1,power_law_1.01,0.3566745758056641
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.2,0.21292224884033203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,1,power_law_1.01,0.3725337600708008
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,1,power_law_1.01,0.3889606475830078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,1,power_law_1.01,0.40342273712158205
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,1,power_law_1.01,0.41160766601562504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,1,power_law_1.01,0.43280769348144527
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,1,power_law_1.01,0.4910745620727539
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,1,power_law_1.01,0.5151462554931641
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.2,0.2687321662902832
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.2,0.5067219161987305
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.2,0.3899027252197266
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.2,0.7625408172607422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.2,0.9997100830078125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.2,1.282860870361328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,1,power_law_1.01,0.529887351989746
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,1,power_law_1.01,0.5940089416503906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,1,power_law_1.01,0.6123884963989258
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,1,power_law_1.01,0.7283340454101562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,1,power_law_1.01,0.896637420654297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,1,power_law_1.01,1.0095148468017578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,1,power_law_1.01,1.1152236938476563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,1,power_law_1.01,0.11902400016784669
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,1,power_law_1.01,0.1548416042327881
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,1,power_law_1.01,1.4378988647460937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,8,1,power_law_1.01,3.6264306640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,1,power_law_1.01,0.20881216049194337
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,1,power_law_1.01,0.1283615970611572
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,1,power_law_1.01,1.76901123046875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,1,power_law_1.01,0.31676544189453126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,1,power_law_1.01,0.32170623779296875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,1,power_law_1.01,0.3331212615966797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,1,power_law_1.01,0.338271369934082
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,1,power_law_1.01,2.4183282470703125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,1,power_law_1.01,0.350695686340332
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,1,power_law_1.01,0.36019073486328124
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,1,power_law_1.01,0.3694572830200195
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,1,power_law_1.01,0.38108734130859373
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,1,power_law_1.01,3.067110900878906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,1,power_law_1.01,0.42863807678222654
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,1,power_law_1.01,0.44055809020996095
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,1,power_law_1.01,0.4582988739013672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,1,power_law_1.01,0.4887276840209961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,1,power_law_1.01,0.10745535850524904
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,1,power_law_1.01,0.5174009704589844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,1,power_law_1.01,0.5752358245849609
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,1,power_law_1.01,0.6520755004882812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,1,power_law_1.01,0.7824748992919922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,1,power_law_1.01,0.2443174362182617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,1,power_law_1.01,0.9140268707275391
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,1,power_law_1.01,1.1913478088378906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,1,power_law_1.01,1.486670684814453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,1,power_law_1.01,2.0383628845214843
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,1,power_law_1.01,2.610758972167969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,8,1,power_law_1.01,3.1684722900390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,1,power_law_1.01,0.05990208148956298
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,1,power_law_1.01,0.06503359794616699
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,1,power_law_1.01,0.07339647769927979
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,1,power_law_1.01,0.12275391578674318
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,1,power_law_1.01,0.09516160011291504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,1,power_law_1.01,0.2056185531616211
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,1,power_law_1.01,0.2116908836364746
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,1,power_law_1.01,0.2204652786254883
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,1,power_law_1.01,0.1492396831512451
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,1,power_law_1.01,0.22916799545288086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,1,power_law_1.01,0.23492671966552736
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,1,power_law_1.01,0.1913484764099121
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,1,power_law_1.01,0.24605823516845704
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,1,power_law_1.01,0.2680384063720703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,1,power_law_1.01,0.258218879699707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,1,power_law_1.01,0.2890956878662109
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,1,power_law_1.01,0.1978995132446289
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,1,power_law_1.01,0.33288448333740234
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.01,0.3742809677124023
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.01,0.46385471343994145
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.01,0.5611116790771484
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.01,0.7509062194824219
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.01,0.9640345764160155
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.01,1.3849606323242187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,1,power_law_1.2,0.0649567985534668
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.01,1.8177127075195312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,1,power_law_1.2,0.2143161582946777
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,1,power_law_1.2,0.32636222839355467
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,1,power_law_1.2,0.12202752113342286
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,1,power_law_1.2,0.34045631408691407
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,1,power_law_1.2,0.16385984420776367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,1,power_law_1.2,0.35170944213867184
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,1,power_law_1.2,0.08217087745666504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,1,power_law_1.2,0.3730329513549805
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,1,power_law_1.2,0.38882431030273434
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,1,power_law_1.2,0.08412544250488281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,1,power_law_1.2,0.44333759307861326
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,1,power_law_1.2,0.5021126556396485
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,1,power_law_1.2,0.5465651321411132
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,1,power_law_1.2,0.6342425537109375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,1,power_law_1.2,0.4032787322998047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,1,power_law_1.2,0.41620223999023437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,1,power_law_1.2,0.6337862396240235
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,1,power_law_1.2,0.752069091796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,1,power_law_1.2,0.531082878112793
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,1,power_law_1.2,0.9323046112060547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,1,power_law_1.2,1.0395449829101562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,1,power_law_1.2,1.2516204833984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,1,power_law_1.2,1.4692454528808594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,1,power_law_1.2,0.10952896118164061
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,1,power_law_1.2,1.804485168457031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,1,power_law_1.2,0.12090432167053222
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,1,power_law_1.2,2.4542573547363284
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,1,power_law_1.2,0.1256159973144531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,1,power_law_1.2,0.1515999984741211
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,1,power_law_1.2,3.0979559326171873
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,1,power_law_1.2,0.1829324722290039
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,8,1,power_law_1.2,3.670841064453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,1,power_law_1.2,0.23138303756713868
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,1,power_law_1.2,0.3110297584533691
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,1,power_law_1.2,0.3220415878295898
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,1,power_law_1.2,0.3293382263183594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,1,power_law_1.2,0.3361209487915039
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,1,power_law_1.2,0.34959423065185546
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,1,power_law_1.2,0.36103870391845705
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,1,power_law_1.2,0.37121536254882814
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,1,power_law_1.2,0.3906182479858399
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,1,power_law_1.2,0.4382252883911133
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,1,power_law_1.2,0.4591072082519531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,1,power_law_1.2,0.477723503112793
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,1,power_law_1.2,0.5109190368652344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,1,power_law_1.2,0.5402163314819336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,1,power_law_1.2,0.6176544189453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,1,power_law_1.2,0.677352294921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,1,power_law_1.2,0.8192518615722657
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,1,power_law_1.2,0.962239990234375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,1,power_law_1.2,1.2130233764648437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.01,2.78357177734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,1,power_law_1.2,1.5103269958496095
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,1,power_law_1.2,2.0788543701171873
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,1,power_law_1.2,0.05982399940490722
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.01,3.63781494140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,1,power_law_1.2,0.06615615844726562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,1,power_law_1.2,0.07192319869995117
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,8,1,power_law_1.01,4.544974060058594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,1,power_law_1.2,2.6372308349609375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,1,power_law_1.2,0.0918988800048828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,1,power_law_1.2,0.11783103942871094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,1,power_law_1.2,0.13902336120605469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,1,power_law_1.2,0.18795200347900393
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,8,1,power_law_1.2,3.1918829345703124
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,1,power_law_1.2,0.19698112487792968
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,1,power_law_1.2,0.2234476852416992
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,1,power_law_1.2,0.21078655242919925
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,1,power_law_1.2,0.20953407287597656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,1,power_law_1.2,0.23188287734985352
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,1,power_law_1.2,0.24126272201538085
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,1,power_law_1.2,0.2566470336914063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,1,power_law_1.2,0.2720211219787598
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,1,power_law_1.2,0.2877440071105957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.2,0.4510617446899414
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,1,power_law_1.2,0.3179078483581543
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.2,1.059651870727539
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.2,0.7456851196289063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.2,0.5958342361450195
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,1,power_law_1.2,0.38011520385742187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.2,2.1257554626464845
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.2,1.409786834716797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,1,power_law_1.01,0.05844672203063965
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,1,power_law_1.01,0.12087295532226563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,1,power_law_1.01,0.038066558837890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,1,power_law_1.01,0.09289407730102539
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,1,power_law_1.01,0.13318143844604494
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,1,power_law_1.01,0.13827839851379395
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,1,power_law_1.01,0.14324416160583495
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,1,power_law_1.01,0.07055424213409424
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,1,power_law_1.01,0.15255680084228515
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,1,power_law_1.01,0.16915712356567383
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,1,power_law_1.01,0.17771776199340822
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,1,power_law_1.01,0.18078720092773437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,1,power_law_1.01,0.18700288772583007
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,1,power_law_1.01,0.19815040588378907
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,1,power_law_1.01,0.24728960037231446
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.2,2.9534808349609376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,1,power_law_1.01,0.28320703506469724
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,1,power_law_1.01,0.3313273620605469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,1,power_law_1.01,0.5232083129882812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,1,power_law_1.01,0.39585983276367187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,1,power_law_1.01,0.6533382415771485
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,1,power_law_1.01,0.9071257781982421
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,1,power_law_1.01,1.163875198364258
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,1,power_law_1.01,0.06186751842498779
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,1,power_law_1.01,1.6788287353515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,1,power_law_1.01,0.06935488224029542
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,1,power_law_1.01,0.08926336288452148
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,1,power_law_1.01,2.197420196533203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,1,power_law_1.01,0.12020992279052733
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,1,power_law_1.01,0.12689344406127928
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,1,power_law_1.01,0.13305600166320802
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,8,1,power_law_1.01,2.71089599609375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,1,power_law_1.01,0.1339475154876709
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,1,power_law_1.01,0.05466432094573974
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,1,power_law_1.01,0.13714112281799315
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,1,power_law_1.01,0.15696831703186037
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,1,power_law_1.01,0.15981056213378905
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,1,power_law_1.01,0.1647929573059082
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,1,power_law_1.01,0.07947135925292968
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,1,power_law_1.01,0.16593536376953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,1,power_law_1.01,0.16988927841186524
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,1,power_law_1.01,0.18116735458374023
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,1,power_law_1.01,0.22526271820068358
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,1,power_law_1.01,0.26033279418945315
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,1,power_law_1.01,0.310120964050293
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,1,power_law_1.01,0.38839935302734374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,1,power_law_1.01,0.1388268756866455
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,1,power_law_1.01,0.47502590179443355
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,1,power_law_1.01,0.6507981109619141
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,1,power_law_1.01,0.8245043182373047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,1,power_law_1.01,0.19845567703247072
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,1,power_law_1.01,1.5199142456054688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,8,1,power_law_1.01,1.8690240478515627
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,1,power_law_1.01,1.1701785278320311
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,1,power_law_1.01,0.039839999675750734
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,1,power_law_1.01,0.08286975860595704
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,1,power_law_1.01,0.0888646411895752
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,1,power_law_1.01,0.04513216018676758
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,1,power_law_1.01,0.09067071914672852
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,1,power_law_1.01,0.04892735958099365
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,1,power_law_1.01,0.056009597778320305
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,1,power_law_1.01,0.09288703918457031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,1,power_law_1.01,0.0945644760131836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,1,power_law_1.01,0.09845248222351075
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,1,power_law_1.01,0.10086015701293945
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,1,power_law_1.01,0.10380736351013184
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,1,power_law_1.01,0.10929727554321289
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,1,power_law_1.01,0.06476672172546387
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,1,power_law_1.01,0.11623807907104493
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,1,power_law_1.01,0.12169983863830566
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,1,power_law_1.01,0.13537216186523438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.01,0.19267520904541016
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,1,power_law_1.01,0.1628428840637207
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.01,0.2511404800415039
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.01,0.3142009544372558
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.01,0.44390209197998043
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.01,0.8502623748779297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.01,0.5756659317016601
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,1,power_law_1.2,0.04148223876953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,1,power_law_1.2,0.058106880187988284
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,1,power_law_1.2,0.06563199996948242
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,1,power_law_1.2,0.0762278413772583
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,1,power_law_1.2,0.08891712188720703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,1,power_law_1.2,0.13223039627075195
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,1,power_law_1.2,0.13081855773925782
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,1,power_law_1.2,0.13616448402404785
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,1,power_law_1.2,0.14119487762451172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,1,power_law_1.2,0.1435436820983887
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,1,power_law_1.2,0.15653056144714356
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,1,power_law_1.2,0.16940671920776368
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,1,power_law_1.2,0.17228160858154298
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,1,power_law_1.2,0.18318656921386717
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,1,power_law_1.2,0.18772031784057616
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,1,power_law_1.2,0.26139392852783205
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,1,power_law_1.2,0.3334611129760742
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,1,power_law_1.2,0.396743049621582
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,1,power_law_1.2,0.5248640060424805
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,1,power_law_1.2,0.17314815521240234
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,1,power_law_1.2,0.6555526733398438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,1,power_law_1.2,0.1932192039489746
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,1,power_law_1.2,0.9063040161132813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,1,power_law_1.2,0.2804979133605957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,1,power_law_1.2,1.1681804656982422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.01,1.1337926483154297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,1,power_law_1.2,0.055724802017211916
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,1,power_law_1.2,0.05985599994659424
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,1,power_law_1.2,0.06842495918273925
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,1,power_law_1.2,1.680756530761719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,1,power_law_1.2,0.07453311920166016
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,1,power_law_1.2,0.08892416000366211
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,1,power_law_1.2,2.1989811706542968
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,1,power_law_1.2,0.1296876811981201
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,1,power_law_1.2,0.13416447639465331
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,1,power_law_1.2,0.13476863861083985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,1,power_law_1.2,0.13736319541931152
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,8,1,power_law_1.2,2.7170343017578125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.01,1.7069792175292968
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,1,power_law_1.2,0.1396076774597168
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,1,power_law_1.2,0.15908991813659668
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,1,power_law_1.2,0.1619660758972168
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,1,power_law_1.2,0.16146112442016602
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,1,power_law_1.2,0.17489984512329101
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,1,power_law_1.2,0.1701024055480957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.01,2.306830749511719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,1,power_law_1.2,0.18283456802368164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,1,power_law_1.2,0.20873088836669923
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,1,power_law_1.2,0.22877567291259765
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,1,power_law_1.2,0.2678643226623535
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,1,power_law_1.2,0.3172083282470703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,1,power_law_1.2,0.39077632904052734
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,1,power_law_1.2,0.4810796737670898
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,1,power_law_1.2,0.6558067321777343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,1,power_law_1.2,0.8314739227294922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,8,1,power_law_1.01,2.9387457275390623
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,1,power_law_1.2,1.173878402709961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,1,power_law_1.2,0.04550784111022949
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,1,power_law_1.2,1.5245887756347656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,1,power_law_1.2,0.0553433609008789
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,1,power_law_1.2,0.0849407958984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,1,power_law_1.2,0.11634431838989259
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,1,power_law_1.2,0.061850237846374514
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,1,power_law_1.2,0.08932991981506347
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,8,1,power_law_1.2,1.8757766723632812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,1,power_law_1.2,0.0925267219543457
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,1,power_law_1.2,0.09423616409301758
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,1,power_law_1.2,0.1040774440765381
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,1,power_law_1.2,0.10758975982666015
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,1,power_law_1.2,0.09999039649963379
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,1,power_law_1.2,0.12318400382995605
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,1,power_law_1.2,0.1306387233734131
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,1,power_law_1.2,0.18770816802978516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,1,power_law_1.2,0.04035264015197754
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,1,power_law_1.2,0.04862656116485596
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,1,power_law_1.2,0.11562687873840331
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,1,power_law_1.2,0.148853120803833
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,1,power_law_1.2,0.09612607955932617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.2,0.3178892707824707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.2,0.40888446807861334
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.2,0.2287820816040039
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,1,power_law_1.01,0.0819929599761963
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,1,power_law_1.01,0.09387647628784179
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,1,power_law_1.01,0.10576831817626953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,1,power_law_1.01,0.14067456245422363
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,1,power_law_1.01,0.17910591125488282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,1,power_law_1.01,0.30264896392822266
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,1,power_law_1.01,0.32245311737060545
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,1,power_law_1.01,0.3375705718994141
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,1,power_law_1.01,0.34421630859375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,1,power_law_1.01,0.3483174514770508
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,1,power_law_1.01,0.3579916763305664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,1,power_law_1.01,0.3863193511962891
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,1,power_law_1.01,0.41090049743652346
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,1,power_law_1.01,0.4209177780151367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,1,power_law_1.01,0.4357734298706054
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,1,power_law_1.01,0.4616524887084961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.2,0.597201271057129
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,1,power_law_1.01,0.5196863937377929
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,1,power_law_1.01,0.5560902404785156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,1,power_law_1.01,0.5949049758911132
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.2,0.7880691528320313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,1,power_law_1.01,0.6977056121826172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,1,power_law_1.01,0.8327699279785156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,1,power_law_1.01,1.0826464080810547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.2,1.163465576171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,1,power_law_1.01,1.3456448364257814
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.2,2.322334747314453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,1,power_law_1.01,1.874295654296875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.2,1.5566943359374998
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,1,power_law_1.01,0.08202048301696777
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,1,power_law_1.01,2.4815910339355467
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,1,power_law_1.01,0.1035852813720703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,1,power_law_1.01,0.09272319793701171
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,1,power_law_1.01,0.16516416549682617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,8,1,power_law_1.2,3.934461364746094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,1,power_law_1.01,0.2207846450805664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,1,power_law_1.01,0.24139328002929689
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,1,power_law_1.01,3.474036560058594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,1,power_law_1.01,0.24990783691406251
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,1,power_law_1.01,0.2555385589599609
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,1,power_law_1.01,0.2595923233032226
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,1,power_law_1.01,0.2654591941833496
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.2,3.120873107910156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,1,power_law_1.01,0.2723443222045899
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,1,power_law_1.01,0.31649663925170896
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,1,power_law_1.01,4.546757202148437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,1,power_law_1.01,0.3217248153686524
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,1,power_law_1.01,0.32541248321533206
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,1,power_law_1.01,0.33732990264892576
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,1,power_law_1.01,0.3585696029663086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,1,power_law_1.01,0.3870745468139648
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,1,power_law_1.01,0.4129452896118164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,1,power_law_1.01,0.4848652648925781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,1,power_law_1.01,0.5630502319335937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,8,1,power_law_1.01,5.693075561523438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,1,power_law_1.01,0.7101459503173828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,1,power_law_1.01,0.8480319976806641
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,1,power_law_1.01,1.1578656005859376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,1,power_law_1.01,1.4407066345214843
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,1,power_law_1.01,2.0256448364257813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,1,power_law_1.01,2.6294580078125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,1,power_law_1.01,0.12840703964233396
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,8,1,power_law_1.01,3.2342919921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,1,power_law_1.01,0.04466047763824463
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,1,power_law_1.01,0.05868607997894287
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,1,power_law_1.01,0.1560799980163574
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,1,power_law_1.01,0.07808000087738037
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,1,power_law_1.01,0.15930368423461913
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,1,power_law_1.01,0.09620608329772949
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,1,power_law_1.01,0.16582592010498048
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,1,power_law_1.01,0.17211135864257812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,1,power_law_1.01,0.051361918449401855
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,1,power_law_1.01,0.1346540832519531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,1,power_law_1.01,0.14601152420043945
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,1,power_law_1.01,0.17565120697021483
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,1,power_law_1.01,0.1938572883605957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,1,power_law_1.01,0.18459455490112303
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,1,power_law_1.01,0.2025222396850586
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,1,power_law_1.01,0.15146559715270996
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,1,power_law_1.01,0.22592063903808596
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.01,0.31065408706665043
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,1,power_law_1.01,0.2692288017272949
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.01,0.40198593139648436
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.01,0.5082617568969726
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.01,0.6987494659423829
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,1,power_law_1.2,0.08870528221130372
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,1,power_law_1.2,0.08279487609863281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,1,power_law_1.2,0.10504768371582032
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.01,0.9043430328369141
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,1,power_law_1.2,0.13116607666015626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,1,power_law_1.2,0.16328575134277343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,1,power_law_1.2,0.32158016204833983
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,1,power_law_1.2,0.3396012878417969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,1,power_law_1.2,0.3456927871704102
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,1,power_law_1.2,0.35282432556152343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,1,power_law_1.2,0.36860481262207034
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,1,power_law_1.2,0.40207809448242193
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,1,power_law_1.2,0.4172556686401367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,1,power_law_1.2,0.4272422409057617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.01,1.3354931640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,1,power_law_1.2,0.45381439208984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,1,power_law_1.2,0.49019390106201166
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,1,power_law_1.2,0.5245248031616211
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,1,power_law_1.2,0.5605427169799805
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,1,power_law_1.2,0.31622976303100586
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,1,power_law_1.2,0.5965983963012695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,1,power_law_1.2,0.713135986328125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,1,power_law_1.2,0.836289291381836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,1,power_law_1.2,1.103937301635742
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,1,power_law_1.2,1.369165496826172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,1,power_law_1.2,1.8893324279785155
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,1,power_law_1.2,2.4207302856445314
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,1,power_law_1.2,3.4807360839843753
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,1,power_law_1.2,0.09136704444885255
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,1,power_law_1.2,4.588800659179688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,1,power_law_1.2,0.10279616355895996
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,1,power_law_1.2,0.14463168144226074
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,1,power_law_1.2,0.12188799858093262
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,1,power_law_1.2,0.21914688110351563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,1,power_law_1.2,0.23933504104614256
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,1,power_law_1.2,0.2526412773132324
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,8,1,power_law_1.2,5.716919555664062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,1,power_law_1.2,0.08136704444885254
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,1,power_law_1.2,0.26934463500976563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,1,power_law_1.2,0.2586476707458496
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,1,power_law_1.2,0.2633939170837402
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,1,power_law_1.2,0.2748883247375488
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,1,power_law_1.2,0.33598079681396487
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,1,power_law_1.2,0.34127552032470704
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,1,power_law_1.2,0.34690303802490235
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,1,power_law_1.2,0.3540691375732422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,1,power_law_1.2,0.3704191970825195
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,1,power_law_1.2,0.3916524887084961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,1,power_law_1.2,0.42111358642578123
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,1,power_law_1.2,0.49529983520507814
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,1,power_law_1.2,0.5757932662963867
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,1,power_law_1.2,0.7206470489501953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,1,power_law_1.2,0.8579539489746093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.01,1.7775846862792968
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,1,power_law_1.2,1.1660179138183593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.01,2.6870214843749998
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,1,power_law_1.2,1.4491424560546875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,1,power_law_1.2,0.0443558406829834
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,1,power_law_1.2,0.05232895851135254
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,1,power_law_1.2,0.05796544075012207
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.01,3.602735290527344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,1,power_law_1.2,2.0405389404296876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,1,power_law_1.2,0.07388351917266846
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,1,power_law_1.2,0.08849920272827148
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,1,power_law_1.2,0.13460415840148926
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,1,power_law_1.2,0.14513919830322267
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,1,power_law_1.2,2.659393920898437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,1,power_law_1.2,0.15494720458984373
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,1,power_law_1.2,0.16076543807983398
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,8,1,power_law_1.01,4.549666442871094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,8,1,power_law_1.2,3.2374700927734374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,1,power_law_1.2,0.16396223068237306
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,1,power_law_1.2,0.17137216567993163
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,1,power_law_1.2,0.17833536148071288
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,1,power_law_1.2,0.18660032272338867
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,1,power_law_1.2,0.19823551177978516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,1,power_law_1.2,0.31220863342285154
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,1,power_law_1.2,0.21291967391967775
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,1,power_law_1.2,0.22631103515625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,1,power_law_1.2,0.2514143943786621
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.2,0.3743948745727539
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.2,0.6621337890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.2,0.9835545349121094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.2,1.3274534606933595
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.2,0.5185683059692383
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,1,power_law_1.01,0.15241984367370603
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,1,power_law_1.01,0.13012928009033203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,1,power_law_1.01,0.1646944046020508
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,1,power_law_1.01,0.11479680061340332
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,1,power_law_1.01,0.3804972839355469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,1,power_law_1.01,0.41179584503173833
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,1,power_law_1.01,0.5072351837158203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,1,power_law_1.01,0.5495251083374024
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,1,power_law_1.01,0.5580902481079102
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,1,power_law_1.01,0.5869740676879883
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,1,power_law_1.01,0.6011897659301757
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,1,power_law_1.01,0.6167014312744141
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,1,power_law_1.01,0.6611942291259766
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,1,power_law_1.01,0.7111398315429687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,1,power_law_1.01,0.7542291259765624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,1,power_law_1.01,0.8296134185791015
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,1,power_law_1.01,0.8938835144042969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,1,power_law_1.01,0.9798995208740233
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,1,power_law_1.01,1.0887296295166016
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,1,power_law_1.01,1.185225601196289
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,1,power_law_1.01,1.4527320861816406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,1,power_law_1.01,1.733278045654297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,1,power_law_1.01,2.2842892456054686
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,1,power_law_1.01,2.807131652832031
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.2,2.071603240966797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,1,power_law_1.01,3.8335699462890624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,1,power_law_1.01,4.975213317871093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,1,power_law_1.01,0.1538086414337158
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,1,power_law_1.01,0.1454259204864502
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,1,power_law_1.01,0.1774412727355957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,1,power_law_1.01,0.1860799980163574
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.2,2.81229248046875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,8,1,power_law_1.01,6.05653076171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,1,power_law_1.01,0.2580396842956543
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,1,power_law_1.01,0.31007551193237304
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.2,4.392236938476563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,1,power_law_1.01,0.35293888092041015
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,1,power_law_1.01,0.44420608520507815
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,1,power_law_1.01,0.42292926788330076
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,1,power_law_1.01,0.45316543579101565
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.2,5.874741821289062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,1,power_law_1.01,0.4796467208862305
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,1,power_law_1.01,0.46880767822265623
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,1,power_law_1.01,0.49191104888916015
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,1,power_law_1.01,0.5105958557128906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,1,power_law_1.01,0.5248857498168945
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,1,power_law_1.01,0.618611183166504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,1,power_law_1.01,0.6517337799072266
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,1,power_law_1.01,0.6817791748046875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,1,power_law_1.01,0.5396275329589844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,1,power_law_1.01,0.7524358367919921
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,1,power_law_1.01,0.835504608154297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,1,power_law_1.01,1.1236793518066406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,1,power_law_1.01,1.4315878295898439
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,1,power_law_1.01,1.74716796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,1,power_law_1.01,2.3626937866210938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,1,power_law_1.01,0.9872544097900391
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,1,power_law_1.01,2.996007080078125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,8,1,power_law_1.01,3.605252380371094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,1,power_law_1.01,0.0791103982925415
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,1,power_law_1.01,0.07912320137023926
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,1,power_law_1.01,0.07955647945404053
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,1,power_law_1.01,0.10586048126220704
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,1,power_law_1.01,0.09163200378417968
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,1,power_law_1.01,0.11244095802307127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,1,power_law_1.01,0.11894335746765136
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,1,power_law_1.01,0.08120512008666993
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,1,power_law_1.01,0.0995686435699463
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,1,power_law_1.01,0.084453763961792
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,1,power_law_1.01,0.13800064086914063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,1,power_law_1.01,0.18448511123657227
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,1,power_law_1.01,0.30090431213378904
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,1,power_law_1.01,0.2454086494445801
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,1,power_law_1.01,0.19855039596557617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,1,power_law_1.01,0.34352191925048825
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,1,power_law_1.01,0.43346366882324217
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,1,power_law_1.01,0.6167961502075195
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.01,1.1570259094238282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,1,power_law_1.2,0.12192959785461426
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.01,1.5251341247558594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,1,power_law_1.2,0.1321555233001709
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.01,0.79500732421875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.01,2.2691571044921877
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,1,power_law_1.2,0.1420096015930176
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,1,power_law_1.2,0.1665235137939453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,1,power_law_1.2,0.34413055419921873
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,1,power_law_1.2,0.38302719116210937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,1,power_law_1.2,0.47064064025878904
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.01,4.441397705078125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.01,2.99342529296875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,1,power_law_1.2,0.5137011337280273
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,1,power_law_1.2,0.539914894104004
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,1,power_law_1.2,0.5677945709228516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,1,power_law_1.2,0.5923807907104492
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,1,power_law_1.2,0.6107468795776367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,1,power_law_1.2,0.6639673614501953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,1,power_law_1.2,0.22989311218261718
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,1,power_law_1.2,0.7162445068359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,1,power_law_1.2,0.761572494506836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,1,power_law_1.2,0.8285574340820313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,1,power_law_1.2,0.9216000366210938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,1,power_law_1.2,0.9949222564697265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,1,power_law_1.2,1.090167007446289
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,1,power_law_1.2,1.2295801544189453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,1,power_law_1.2,1.5017318725585938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,1,power_law_1.2,1.7934541320800783
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,1,power_law_1.2,2.290887603759766
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,1,power_law_1.2,0.14558400154113768
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,1,power_law_1.2,2.8317037963867184
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,1,power_law_1.2,0.15835007667541504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,1,power_law_1.2,0.16873472213745117
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,1,power_law_1.2,0.18696767807006837
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,1,power_law_1.2,0.25365760803222653
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,1,power_law_1.2,3.8549862670898434
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,1,power_law_1.2,0.29190399169921877
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,1,power_law_1.2,0.3286284637451172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,1,power_law_1.2,0.4017516708374023
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,1,power_law_1.2,0.42806144714355465
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,1,power_law_1.2,5.01808837890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,1,power_law_1.2,0.4434175872802735
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,1,power_law_1.2,0.4609081649780274
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,1,power_law_1.2,0.475030403137207
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,8,1,power_law_1.2,6.162962036132813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,1,power_law_1.2,0.48416767120361326
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,1,power_law_1.2,0.5033689498901367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,1,power_law_1.2,0.5193267059326172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,1,power_law_1.2,0.5377990341186523
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,1,power_law_1.2,0.6164377593994141
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,1,power_law_1.2,0.6637651062011718
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,1,power_law_1.2,0.7017740631103516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,1,power_law_1.2,0.7941139221191407
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,1,power_law_1.2,0.8645184326171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,1,power_law_1.2,1.0285453033447265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,1,power_law_1.2,1.1845536041259765
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,1,power_law_1.2,1.4958149719238283
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,1,power_law_1.2,1.7959341430664062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,1,power_law_1.2,0.07911295890808105
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,1,power_law_1.2,2.4243641662597657
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,1,power_law_1.2,0.07909120082855224
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,1,power_law_1.2,3.0424749755859373
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,8,1,power_law_1.2,3.669820861816406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,1,power_law_1.2,0.08126912117004395
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,1,power_law_1.2,0.08440640449523926
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,1,power_law_1.2,0.09115584373474121
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,1,power_law_1.2,0.10566592216491699
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,1,power_law_1.2,0.1004748821258545
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,1,power_law_1.2,0.11198016166687012
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,1,power_law_1.2,0.12053759574890137
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,1,power_law_1.2,0.13644607543945314
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,1,power_law_1.2,0.18232255935668945
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,1,power_law_1.2,0.20528448104858396
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,1,power_law_1.2,0.24727296829223633
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,1,power_law_1.2,0.2993811225891113
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,1,power_law_1.2,0.34433406829833985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.2,1.5277331542968748
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,1,power_law_1.2,0.08018176078796387
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,1,power_law_1.2,0.43459457397460943
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.2,0.7951251220703125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,1,power_law_1.2,0.6198073577880859
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.2,1.155667190551758
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.2,2.2620915222167968
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,1,power_law_1.01,0.09792192459106445
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,1,power_law_1.01,0.09729023933410644
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,1,power_law_1.01,0.09809791564941406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,1,power_law_1.01,0.09864831924438476
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,1,power_law_1.01,0.1004038429260254
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,1,power_law_1.01,0.10269375801086426
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,1,power_law_1.01,0.10439999580383301
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,1,power_law_1.01,0.10742079734802246
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,1,power_law_1.01,0.11095232009887694
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,1,power_law_1.01,0.1148755168914795
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,1,power_law_1.01,0.12279616355895998
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,1,power_law_1.01,0.12694656372070312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,1,power_law_1.01,0.1508236789703369
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,1,power_law_1.01,0.18230464935302734
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,1,power_law_1.01,0.23401664733886718
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,1,power_law_1.01,0.28264383316040037
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,1,power_law_1.01,0.3847910308837891
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,1,power_law_1.01,0.09249983787536621
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,1,power_law_1.01,0.4881286239624023
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,1,power_law_1.01,0.6912915039062499
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,1,power_law_1.01,0.9037427520751953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,1,power_law_1.01,1.3375404357910157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,1,power_law_1.01,1.7543743896484376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,1,power_law_1.01,0.03701247930526733
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,1,power_law_1.01,0.0414899206161499
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,1,power_law_1.01,0.04273087978363037
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,16,1,power_law_1.01,2.2104786682128905
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,1,power_law_1.01,0.043951997756958
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,1,power_law_1.01,0.04737152099609375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.2,2.9837728881835934
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.2,4.45716552734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,1,power_law_1.01,0.049823360443115236
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,1,power_law_1.01,0.04843647956848145
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,1,power_law_1.01,0.05317247867584228
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,1,power_law_1.01,0.05045631885528564
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,1,power_law_1.01,0.05308479785919189
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,1,power_law_1.01,0.05416512012481689
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,1,power_law_1.01,0.05187520027160645
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,1,power_law_1.01,0.06209152221679688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,1,power_law_1.01,0.0630022382736206
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,1,power_law_1.01,0.06858304023742676
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,1,power_law_1.01,0.06742144107818604
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,1,power_law_1.01,0.07533952236175537
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,1,power_law_1.01,0.09362496376037598
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,1,power_law_1.01,0.11900799751281739
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,1,power_law_1.01,0.13439231872558594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,1,power_law_1.01,0.04909440040588379
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,1,power_law_1.01,0.18087808609008788
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,1,power_law_1.01,0.233756160736084
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,1,power_law_1.01,0.3186790466308594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,1,power_law_1.01,0.41764225006103517
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,1,power_law_1.01,0.6079961776733398
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,1,power_law_1.01,0.7922207641601562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,16,1,power_law_1.01,0.9780223846435547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,1,power_law_1.01,0.01890944004058838
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,1,power_law_1.01,0.019392000436782835
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,1,power_law_1.01,0.02059648036956787
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,1,power_law_1.01,0.021353600025177003
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,1,power_law_1.01,0.021788160800933837
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,1,power_law_1.01,0.021809279918670654
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,1,power_law_1.01,0.02184767961502075
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,1,power_law_1.01,0.022195839881896974
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,1,power_law_1.01,0.022011520862579344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,1,power_law_1.01,0.022206079959869385
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,1,power_law_1.01,0.021436800956726072
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,1,power_law_1.01,0.025287039279937744
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,1,power_law_1.01,0.02399104118347168
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,1,power_law_1.01,0.027435519695281983
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,1,power_law_1.01,0.030033919811248776
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,1,power_law_1.01,0.036604158878326416
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.01,0.050716800689697264
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,1,power_law_1.01,0.033056640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,1,power_law_1.01,0.04327040195465088
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.01,0.06415616035461426
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.01,0.23640703201293944
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.01,0.34512512207031254
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,1,power_law_1.2,0.0514521598815918
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,1,power_law_1.2,0.05148352146148681
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,1,power_law_1.2,0.05618495941162109
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,1,power_law_1.2,0.06422848224639892
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,1,power_law_1.2,0.09562623977661133
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,1,power_law_1.2,0.09469440460205078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.01,0.4520742416381836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,1,power_law_1.2,0.09747455596923829
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,1,power_law_1.2,0.09649087905883788
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,1,power_law_1.2,0.09820416450500488
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,1,power_law_1.2,0.10120832443237304
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,1,power_law_1.2,0.10103103637695313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,1,power_law_1.2,0.10443903923034667
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,1,power_law_1.2,0.10819007873535155
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,1,power_law_1.2,0.11044672012329101
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,1,power_law_1.2,0.11534144401550292
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,1,power_law_1.2,0.12619711875915526
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,1,power_law_1.2,0.1264076805114746
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,1,power_law_1.2,0.14791040420532225
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,1,power_law_1.2,0.18660736083984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,1,power_law_1.2,0.2349440002441406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,1,power_law_1.2,0.28426111221313477
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,1,power_law_1.2,0.38740608215332034
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,1,power_law_1.2,0.48966529846191403
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,1,power_law_1.2,0.6952774047851562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.01,0.07549376010894775
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,1,power_law_1.2,0.9032640075683593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.01,0.1272064018249512
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.01,0.10159168243408204
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,1,power_law_1.2,0.03655807971954346
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,1,power_law_1.2,0.042271361351013184
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.01,0.18083520889282229
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,1,power_law_1.2,1.3238304138183594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,1,power_law_1.2,0.046938238143920896
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,1,power_law_1.2,0.04760064125061035
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,1,power_law_1.2,0.0488972806930542
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,1,power_law_1.2,0.04930047988891602
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,1,power_law_1.2,0.051695361137390136
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,1,power_law_1.2,0.051571841239929195
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,1,power_law_1.2,0.0525164794921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,1,power_law_1.2,1.7586976623535158
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,1,power_law_1.2,0.053825922012329094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,1,power_law_1.2,0.06170432090759277
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,1,power_law_1.2,0.06931392192840577
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,1,power_law_1.2,0.06888832092285156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,1,power_law_1.2,0.06932032108306885
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,1,power_law_1.2,0.07539519786834717
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,1,power_law_1.2,0.09262720108032227
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,1,power_law_1.2,0.11820544242858885
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,1,power_law_1.2,0.1333356761932373
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,1,power_law_1.2,0.18058111190795897
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,1,power_law_1.2,0.23425279617309572
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,1,power_law_1.2,0.3197804832458496
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,1,power_law_1.2,0.03877311944961548
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,1,power_law_1.2,0.037363839149475095
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,1,power_law_1.2,0.4210924911499023
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,1,power_law_1.2,0.019409919977188113
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,1,power_law_1.2,0.045576319694519044
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,1,power_law_1.2,0.6080908966064453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,1,power_law_1.2,0.01971968054771423
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,1,power_law_1.2,0.021843841075897215
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,1,power_law_1.2,0.021800320148468017
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,1,power_law_1.2,0.7925702667236327
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,1,power_law_1.2,0.02193471908569336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,1,power_law_1.2,0.02207871913909912
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,1,power_law_1.2,0.02237567901611328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,16,1,power_law_1.2,0.9681830596923827
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,1,power_law_1.2,0.022181119918823242
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,1,power_law_1.2,0.02220671892166138
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,1,power_law_1.2,0.02428031921386719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,1,power_law_1.2,0.030279040336608887
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,16,1,power_law_1.2,2.182380828857422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.01,0.5616236877441406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,1,power_law_1.2,0.018095359802246094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,1,power_law_1.2,0.020920960903167723
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,1,power_law_1.2,0.025354239940643307
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,1,power_law_1.2,0.034288640022277835
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,1,power_law_1.2,0.027606399059295655
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,1,power_law_1.2,0.04213759899139404
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,1,power_law_1.2,0.0367577600479126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.2,0.3577715301513672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.2,0.05159872055053712
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.2,0.5805625534057617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,1,power_law_1.01,0.0663155221939087
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,1,power_law_1.01,0.073438720703125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,1,power_law_1.01,0.07770880222320556
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,1,power_law_1.01,0.08985407829284668
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,1,power_law_1.01,0.1287071990966797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,1,power_law_1.01,0.13053183555603026
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,1,power_law_1.01,0.13155712127685545
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,1,power_law_1.01,0.13226816177368164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,1,power_law_1.01,0.13319359779357912
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,1,power_law_1.01,0.13505151748657226
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.2,0.10636927604675292
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,1,power_law_1.01,0.1435859203338623
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.2,0.1332268810272217
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.2,0.1890233612060547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,1,power_law_1.01,0.14739775657653809
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.2,0.24478336334228518
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,1,power_law_1.01,0.15235136032104493
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,1,power_law_1.01,0.15627327919006348
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,1,power_law_1.01,0.16081151962280274
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,1,power_law_1.01,0.16853567123413088
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.2,0.06529280185699463
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,1,power_law_1.01,0.1786796760559082
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,1,power_law_1.01,0.22606336593627932
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,1,power_law_1.01,0.2944870376586914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,1,power_law_1.01,0.3602751922607422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,1,power_law_1.01,0.44523967742919923
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.2,0.0791596794128418
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,1,power_law_1.01,0.6095571136474609
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,1,power_law_1.01,0.7736115264892578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,1,power_law_1.01,0.04384448051452637
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,1,power_law_1.01,0.050840959548950196
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,1,power_law_1.01,0.05109248161315918
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,1,power_law_1.01,1.1186354827880858
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,1,power_law_1.01,0.06382847785949707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,1,power_law_1.01,0.06905600070953369
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,1,power_law_1.01,0.07209727764129639
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,1,power_law_1.01,0.07231359958648682
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,1,power_law_1.01,1.4785958862304687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,1,power_law_1.01,0.07196479797363281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,1,power_law_1.01,0.07348480224609374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,1,power_law_1.01,0.07427199840545654
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,1,power_law_1.01,0.0768230390548706
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,1,power_law_1.01,0.07566976070404052
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,1,power_law_1.01,0.07889791965484619
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,1,power_law_1.01,0.08125696182250977
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,1,power_law_1.01,0.08528063774108888
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,1,power_law_1.01,0.08621439933776856
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.2,0.4702617645263672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,1,power_law_1.01,0.08960639953613281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,1,power_law_1.01,0.10044927597045898
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,1,power_law_1.01,0.13445183753967285
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,1,power_law_1.01,0.1553273582458496
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,1,power_law_1.01,2.174784698486328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,1,power_law_1.01,0.18677120208740233
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,1,power_law_1.01,0.252348804473877
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,1,power_law_1.01,0.3240409469604492
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,1,power_law_1.01,0.46299903869628906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,1,power_law_1.01,0.6024044799804688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,1,power_law_1.01,2.8974404907226563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,1,power_law_1.01,0.868326416015625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,1,power_law_1.01,1.1384896087646486
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,16,1,power_law_1.01,3.628113403320312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,16,1,power_law_1.01,1.4163993835449218
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,1,power_law_1.01,0.02473599910736084
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,1,power_law_1.01,0.03106688022613525
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,1,power_law_1.01,0.03139136075973511
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,1,power_law_1.01,0.03258624076843262
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,1,power_law_1.01,0.03135551929473877
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,1,power_law_1.01,0.03325695991516113
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,1,power_law_1.01,0.03148351907730103
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,1,power_law_1.01,0.021375999450683594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,1,power_law_1.01,0.031420800685882565
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,1,power_law_1.01,0.024234240055084226
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,1,power_law_1.01,0.029578878879547122
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,1,power_law_1.01,0.032868480682373045
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,1,power_law_1.01,0.03490880012512207
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,1,power_law_1.01,0.0439788818359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,1,power_law_1.01,0.04319488048553467
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,1,power_law_1.01,0.048900480270385745
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,1,power_law_1.01,0.0567136001586914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.01,0.0691161584854126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,1,power_law_1.01,0.04089344024658203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.01,0.08674943923950196
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.01,0.14080703735351563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.01,0.10255488395690918
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.01,0.1790675163269043
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,1,power_law_1.2,0.06915711879730224
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,1,power_law_1.2,0.06526016235351563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,1,power_law_1.2,0.07510144233703613
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,1,power_law_1.2,0.08812352180480956
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,1,power_law_1.2,0.12788479804992675
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,1,power_law_1.2,0.13076800346374512
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,1,power_law_1.2,0.13042240142822265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,1,power_law_1.2,0.13276927947998046
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.01,0.25340991973876953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,1,power_law_1.2,0.1341215991973877
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,1,power_law_1.2,0.13483648300170897
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.01,0.33081344604492186
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,1,power_law_1.2,0.1441907215118408
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,1,power_law_1.2,0.1488422393798828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,1,power_law_1.2,0.15161215782165527
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,1,power_law_1.2,0.1549638366699219
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,1,power_law_1.2,0.16033344268798827
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,1,power_law_1.2,0.17128639221191405
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,1,power_law_1.2,0.17869056701660155
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,1,power_law_1.2,0.22522432327270508
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,1,power_law_1.2,0.2986144065856934
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,1,power_law_1.2,0.3641292953491211
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,1,power_law_1.2,0.44742912292480475
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.01,0.48544574737548823
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,1,power_law_1.2,0.6148096084594726
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.01,0.6409209442138672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,1,power_law_1.2,0.774458236694336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,1,power_law_1.2,0.04411007881164551
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,1,power_law_1.2,0.04913087844848633
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,1,power_law_1.2,1.1272627258300782
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,1,power_law_1.2,0.050079998970031736
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,1,power_law_1.2,0.0640774393081665
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,1,power_law_1.2,0.06780159950256348
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,1,power_law_1.2,0.06984064102172852
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,1,power_law_1.2,0.07206655979156494
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,1,power_law_1.2,0.07172287940979004
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,1,power_law_1.2,1.4568946838378907
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,1,power_law_1.2,0.07184639930725098
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,1,power_law_1.2,0.07388095855712891
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,1,power_law_1.2,0.0748857593536377
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,1,power_law_1.2,0.07765247821807861
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,1,power_law_1.2,0.07750400066375732
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,1,power_law_1.2,0.081278076171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,1,power_law_1.2,0.0833676815032959
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,1,power_law_1.2,0.0884505558013916
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,1,power_law_1.2,0.09147520065307617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,1,power_law_1.2,0.10010304450988769
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,1,power_law_1.2,0.13456255912780762
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,1,power_law_1.2,2.184656066894531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,1,power_law_1.2,0.15554112434387207
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,1,power_law_1.2,0.18679744720458985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,1,power_law_1.2,0.2521414375305176
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,1,power_law_1.2,0.3234880065917969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,1,power_law_1.2,0.4636025619506836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,1,power_law_1.2,2.902193908691406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,1,power_law_1.2,0.6012128067016602
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,1,power_law_1.2,0.02100287914276123
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,1,power_law_1.2,0.02423935890197754
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,1,power_law_1.2,0.024671359062194826
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,1,power_law_1.2,0.02901504039764404
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,1,power_law_1.2,0.8667481231689453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,1,power_law_1.2,0.030369920730590822
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,1,power_law_1.2,0.03201152086257934
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,1,power_law_1.2,0.031186559200286866
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,1,power_law_1.2,0.03146239995956421
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,1,power_law_1.2,0.032629120349884036
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,1,power_law_1.2,0.0327455997467041
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,1,power_law_1.2,0.03246272087097168
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,1,power_law_1.2,1.1482080078124999
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,1,power_law_1.2,0.033393919467926025
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,16,1,power_law_1.2,3.679586486816406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,1,power_law_1.2,0.04596223831176758
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,16,1,power_law_1.2,1.4091270446777344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.01,0.804554214477539
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,1,power_law_1.2,0.04478464126586914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,1,power_law_1.2,0.03550400018692017
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,1,power_law_1.2,0.040871682167053225
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,1,power_law_1.2,0.057677440643310554
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,1,power_law_1.2,0.04969024181365967
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.2,0.06956031799316406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.2,0.08777088165283203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.2,0.18514623641967773
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.2,0.2633292770385742
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,1,power_law_1.01,0.09846719741821289
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,1,power_law_1.01,0.1047481632232666
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,1,power_law_1.01,0.06929279804229736
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,1,power_law_1.01,0.12431424140930177
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,1,power_law_1.01,0.1794790458679199
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,1,power_law_1.01,0.14854784011840821
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,1,power_law_1.01,0.2308876800537109
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,1,power_law_1.01,0.23997184753417972
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,1,power_law_1.01,0.24878271102905272
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,1,power_law_1.01,0.25529663085937504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,1,power_law_1.01,0.2681766319274902
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,1,power_law_1.01,0.2862784004211426
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,1,power_law_1.01,0.3200281524658203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,1,power_law_1.01,0.3423379135131836
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,1,power_law_1.01,0.37365375518798827
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,1,power_law_1.01,0.38437313079833985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,1,power_law_1.01,0.406192626953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,1,power_law_1.01,0.4558041763305664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,1,power_law_1.01,0.5510259246826171
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,1,power_law_1.01,0.6126144027709961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,1,power_law_1.01,0.6973305511474609
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,1,power_law_1.01,0.8774380493164063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,1,power_law_1.01,1.0406380462646485
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,1,power_law_1.01,1.4012684631347656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.2,0.10654080390930176
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,1,power_law_1.01,1.7652613830566406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,1,power_law_1.01,2.4916217041015623
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,1,power_law_1.01,0.10842623710632324
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,1,power_law_1.01,0.11310208320617676
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.2,0.34368511199951174
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.2,0.5028448104858398
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,1,power_law_1.01,3.2188385009765623
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.2,0.8356492614746094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,1,power_law_1.01,0.12342656135559082
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.2,0.14530303955078125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,1,power_law_1.01,0.2558572769165039
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,16,1,power_law_1.01,3.9511654663085936
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.2,0.6654137420654297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,1,power_law_1.01,0.14363391876220705
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,1,power_law_1.01,0.2622470474243164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,1,power_law_1.01,0.26877311706542967
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,1,power_law_1.01,0.27386367797851563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,1,power_law_1.01,0.2920172882080078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,1,power_law_1.01,0.2821574401855469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,1,power_law_1.01,0.29797311782836916
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,1,power_law_1.01,0.3077824020385742
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,1,power_law_1.01,0.3385855865478516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,1,power_law_1.01,0.34623294830322265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,1,power_law_1.01,0.35860992431640626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,1,power_law_1.01,0.38304576873779295
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,1,power_law_1.01,0.4092614364624024
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,1,power_law_1.01,0.4635987091064453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,1,power_law_1.01,0.5366860961914062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,1,power_law_1.01,0.6515097808837891
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,1,power_law_1.01,0.7587545776367188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,1,power_law_1.01,0.9986879730224609
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,1,power_law_1.01,1.251587219238281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,1,power_law_1.01,0.1990969657897949
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,1,power_law_1.01,1.7400460815429688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,1,power_law_1.01,0.16927871704101563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,1,power_law_1.01,2.233567352294922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,16,1,power_law_1.01,2.705333251953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,1,power_law_1.01,0.06476160049438476
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,1,power_law_1.01,0.07624383926391601
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,1,power_law_1.01,0.06731520175933839
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,1,power_law_1.01,0.09091456413269043
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,1,power_law_1.01,0.17998527526855468
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,1,power_law_1.01,0.11379008293151856
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,1,power_law_1.01,0.1342470359802246
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,1,power_law_1.01,0.18518463134765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,1,power_law_1.01,0.1930726432800293
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,1,power_law_1.01,0.16907264709472655
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,1,power_law_1.01,0.20061439514160156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,1,power_law_1.01,0.2068659210205078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,1,power_law_1.01,0.21726463317871092
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,1,power_law_1.01,0.22733184814453128
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,1,power_law_1.01,0.2379756736755371
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,1,power_law_1.01,0.25720767974853515
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,1,power_law_1.01,0.17499711990356445
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,1,power_law_1.01,0.29809343338012695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.01,0.33758590698242186
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.01,0.5188089752197266
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.01,0.42166400909423823
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.01,0.6970899200439453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.01,0.8963136291503906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,1,power_law_1.2,0.07300864219665527
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,1,power_law_1.2,0.09612544059753418
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,1,power_law_1.2,0.1040339183807373
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.01,1.319669189453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,1,power_law_1.2,0.12057279586791991
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,1,power_law_1.2,0.13844991683959962
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,1,power_law_1.2,0.16912128448486327
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,1,power_law_1.2,0.23654720306396487
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,1,power_law_1.2,0.2246335983276367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,1,power_law_1.2,0.24430143356323245
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,1,power_law_1.2,0.2559116744995117
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,1,power_law_1.2,0.2695871925354004
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,1,power_law_1.2,0.30483968734741207
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,1,power_law_1.2,0.3241759872436524
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,1,power_law_1.2,0.35473342895507814
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.01,1.7401068115234373
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,1,power_law_1.2,0.3795110321044922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,1,power_law_1.2,0.3933030319213867
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,1,power_law_1.2,0.4148569488525391
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,1,power_law_1.2,0.511973762512207
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,1,power_law_1.2,0.6238880157470703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,1,power_law_1.2,0.5560921478271484
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,1,power_law_1.2,0.7243846130371094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,1,power_law_1.2,0.893927001953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,1,power_law_1.2,1.063012466430664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,1,power_law_1.2,1.4138291931152343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,1,power_law_1.2,1.7760185241699218
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,1,power_law_1.2,0.10824959754943848
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,1,power_law_1.2,0.11358336448669433
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,1,power_law_1.2,2.511749725341797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,1,power_law_1.2,0.12201087951660157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,1,power_law_1.2,0.13990015983581544
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,1,power_law_1.2,0.18787328720092772
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,1,power_law_1.2,3.2374899291992185
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,1,power_law_1.2,0.2533011245727539
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,1,power_law_1.2,0.26093631744384765
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,1,power_law_1.2,0.2659513664245606
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,16,1,power_law_1.2,3.972835693359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,1,power_law_1.2,0.2727903938293457
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,1,power_law_1.2,0.28195007324218746
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,1,power_law_1.2,0.290634880065918
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,1,power_law_1.2,0.29881151199340816
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,1,power_law_1.2,0.31190464019775394
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,1,power_law_1.2,0.34408447265625003
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,1,power_law_1.2,0.3540172958374023
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,1,power_law_1.2,0.36834815979003904
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,1,power_law_1.2,0.4002412796020508
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,1,power_law_1.2,0.42845951080322264
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,1,power_law_1.2,0.5009241485595703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,1,power_law_1.2,0.5577081680297852
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,1,power_law_1.2,0.16126848220825196
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,1,power_law_1.2,0.6719725036621094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,1,power_law_1.2,0.8031238555908203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,1,power_law_1.2,1.0151622772216797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,1,power_law_1.2,1.2697638702392577
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,1,power_law_1.2,1.7577311706542968
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,1,power_law_1.2,0.06481728076934815
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,1,power_law_1.2,0.06810431957244874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,1,power_law_1.2,2.2512205505371092
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,1,power_law_1.2,0.07560959815979004
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,1,power_law_1.2,0.08868672370910644
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,16,1,power_law_1.2,2.723662719726563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,1,power_law_1.2,0.10424256324768066
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,1,power_law_1.2,0.12726783752441406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,1,power_law_1.2,0.16523199081420897
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,1,power_law_1.2,0.17398656845092772
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,1,power_law_1.2,0.1798931121826172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,1,power_law_1.2,0.18583232879638673
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,1,power_law_1.2,0.19494911193847658
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,1,power_law_1.2,0.20350080490112304
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,1,power_law_1.2,0.21220544815063475
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,1,power_law_1.2,0.22670528411865237
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,1,power_law_1.2,0.2413849639892578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,1,power_law_1.2,0.25365888595581054
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.01,2.5393708801269534
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.01,3.425883483886719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,16,1,power_law_1.01,4.294485778808594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,1,power_law_1.2,0.3457036972045898
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,1,power_law_1.2,0.28723968505859376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.2,0.40326656341552736
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.2,0.5478041458129883
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.2,0.6934272003173828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.2,1.0051538848876953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.2,1.3456959533691406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.2,2.038390350341797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,1,power_law_1.01,0.05298175811767578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,1,power_law_1.01,0.06157055854797363
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,1,power_law_1.01,0.09711935997009277
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,1,power_law_1.01,0.10037504196166994
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,1,power_law_1.01,0.0760640001296997
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,1,power_law_1.01,0.10463871955871581
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,1,power_law_1.01,0.10811903953552246
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,1,power_law_1.01,0.11419008255004884
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,1,power_law_1.01,0.14623040199279785
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,1,power_law_1.01,0.14869440078735352
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,1,power_law_1.01,0.15409791946411133
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,1,power_law_1.01,0.15737792015075686
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,1,power_law_1.01,0.16199808120727538
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,1,power_law_1.01,0.17051071166992188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,1,power_law_1.01,0.1916691207885742
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,1,power_law_1.01,0.21045696258544924
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,1,power_law_1.01,0.2506188774108887
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,1,power_law_1.01,0.3044998359680176
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,1,power_law_1.01,0.418163833618164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,1,power_law_1.01,0.5058214569091797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,1,power_law_1.01,0.06827839851379394
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,1,power_law_1.01,0.7103955078124999
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,1,power_law_1.01,0.9054822540283203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,1,power_law_1.01,1.2974470520019532
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,1,power_law_1.01,0.054290561676025396
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,1,power_law_1.01,0.05498239994049072
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,1,power_law_1.01,0.06506048202514648
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,1,power_law_1.01,1.6930854797363282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,1,power_law_1.01,0.07132287979125976
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,1,power_law_1.01,0.09419967651367187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,1,power_law_1.01,0.07759039878845216
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,16,1,power_law_1.01,2.0936793518066406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,1,power_law_1.01,0.1020576000213623
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,1,power_law_1.01,0.1045241641998291
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,1,power_law_1.01,0.10705599784851075
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,1,power_law_1.01,0.11022399902343749
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,1,power_law_1.01,0.12737088203430175
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,1,power_law_1.01,0.1289017581939697
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,1,power_law_1.01,0.1337222385406494
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,1,power_law_1.01,0.13493632316589355
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,1,power_law_1.01,0.1381926441192627
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,1,power_law_1.01,0.1484102439880371
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,1,power_law_1.01,0.16385023117065428
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,1,power_law_1.01,0.10899840354919435
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,1,power_law_1.01,0.19205951690673828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,1,power_law_1.01,0.22463680267333985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,1,power_law_1.01,0.26307584762573244
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,1,power_law_1.01,0.41448833465576174
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,1,power_law_1.01,0.567322883605957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,1,power_law_1.01,0.09890368461608887
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,1,power_law_1.01,0.7161888122558594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,1,power_law_1.01,1.011004180908203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,1,power_law_1.01,1.316012725830078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,1,power_law_1.01,0.3323737716674805
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,16,1,power_law_1.01,1.6234002685546876
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.2,2.760775146484375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,1,power_law_1.01,0.05054336071014405
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,1,power_law_1.01,0.04197504043579102
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,1,power_law_1.01,0.04585536003112793
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,1,power_law_1.01,0.08518400192260742
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,1,power_law_1.01,0.05600319862365723
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,1,power_law_1.01,0.08755840301513672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,1,power_law_1.01,0.08894847869873047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,1,power_law_1.01,0.09084351539611817
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,1,power_law_1.01,0.0944320011138916
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,1,power_law_1.01,0.09618304252624513
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,1,power_law_1.01,0.06265279769897461
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,1,power_law_1.01,0.10005824089050293
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,1,power_law_1.01,0.10624575614929199
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,1,power_law_1.01,0.08918208122253418
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,1,power_law_1.01,0.1361574363708496
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,1,power_law_1.01,0.11686911582946777
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,1,power_law_1.01,0.16108415603637696
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,1,power_law_1.01,0.19206079483032226
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,1,power_law_1.01,0.12239487648010254
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,1,power_law_1.01,0.2527654457092285
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,1,power_law_1.01,0.3151193618774414
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,1,power_law_1.01,0.44375167846679686
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,1,power_law_1.2,0.05420159816741943
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,1,power_law_1.01,0.5781760025024414
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,1,power_law_1.01,0.8509881591796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,1,power_law_1.2,0.06079232215881347
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,1,power_law_1.2,0.06559743881225585
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,1,power_law_1.2,0.09337663650512695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,1,power_law_1.2,0.09866175651550294
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,1,power_law_1.2,0.04108287811279297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,1,power_law_1.2,0.10235008239746093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,1,power_law_1.2,0.10686016082763672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,1,power_law_1.2,0.10828864097595214
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,1,power_law_1.2,0.11645567893981934
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,1,power_law_1.2,0.14689215660095215
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,1,power_law_1.2,0.14902400016784667
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,1,power_law_1.2,0.1557587242126465
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,1,power_law_1.2,0.1581612777709961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,1,power_law_1.2,0.16181888580322265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,1,power_law_1.01,1.1257913970947266
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,1,power_law_1.2,0.17195775985717773
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,1,power_law_1.2,0.19221120834350586
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,1,power_law_1.2,0.21039424896240236
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,1,power_law_1.2,0.2523820877075195
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,1,power_law_1.2,0.30748992919921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,1,power_law_1.2,0.4180108642578125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,1,power_law_1.2,0.5077433776855469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,1,power_law_1.2,0.712823715209961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,1,power_law_1.01,1.6846476745605468
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,1,power_law_1.2,0.0746553611755371
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,1,power_law_1.2,0.9082649230957032
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,1,power_law_1.2,0.05576255798339844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,1,power_law_1.2,0.06235392093658447
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,1,power_law_1.2,1.3022227478027344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,1,power_law_1.2,0.06968768119812012
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,1,power_law_1.2,0.0567846393585205
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,1,power_law_1.2,0.07569087982177734
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,1,power_law_1.2,0.09940608024597168
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,1,power_law_1.2,1.6998028564453125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,1,power_law_1.2,0.10305472373962403
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,1,power_law_1.2,0.1064038372039795
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,1,power_law_1.2,0.10797375679016112
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,1,power_law_1.2,0.10887295722961425
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,1,power_law_1.2,0.1112287998199463
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,1,power_law_1.2,0.1286303997039795
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,1,power_law_1.2,0.13119680404663087
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,1,power_law_1.2,0.1359615993499756
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,16,1,power_law_1.2,2.1014968872070314
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,1,power_law_1.2,0.13759360313415528
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,1,power_law_1.2,0.14160896301269532
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,1,power_law_1.2,0.15004672050476073
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,1,power_law_1.2,0.17452735900878907
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,1,power_law_1.2,0.19508224487304687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,1,power_law_1.2,0.23177087783813474
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,1,power_law_1.2,0.26562431335449216
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,1,power_law_1.2,0.33383552551269535
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,1,power_law_1.2,0.4189497756958008
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,1,power_law_1.2,0.571138572692871
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,1,power_law_1.2,0.7226182556152343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,1,power_law_1.2,0.0430188798904419
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,1,power_law_1.2,0.04629695892333984
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,1,power_law_1.2,1.0127308654785157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,1,power_law_1.2,0.050067200660705566
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,1,power_law_1.2,0.0545081615447998
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,1,power_law_1.2,1.3198060607910156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,1,power_law_1.2,0.06025023937225342
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,1,power_law_1.2,0.08144895553588867
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,1,power_law_1.2,0.08600704193115234
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,16,1,power_law_1.2,1.6255673217773439
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,1,power_law_1.2,0.08837568283081054
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,1,power_law_1.2,0.08964351654052735
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,1,power_law_1.2,0.09190272331237793
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,1,power_law_1.2,0.09650495529174805
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,1,power_law_1.2,0.1009068775177002
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,1,power_law_1.2,0.10447039604187011
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,1,power_law_1.01,2.27088134765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,16,1,power_law_1.01,2.8977951049804687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,1,power_law_1.2,0.11657216072082519
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,1,power_law_1.2,0.13166272163391113
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,1,power_law_1.2,0.1232703971862793
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,1,power_law_1.2,0.22951488494873046
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,1,power_law_1.2,0.14928256034851076
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,1,power_law_1.2,0.1874553680419922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,1,power_law_1.2,0.3121279907226563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,1,power_law_1.2,0.40458240509033205
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,1,power_law_1.01,0.06063360214233399
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,1,power_law_1.01,0.08444352149963379
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,1,power_law_1.2,0.7791027069091797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,1,power_law_1.01,0.09383359909057618
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,1,power_law_1.2,1.146650848388672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,1,power_law_1.01,0.13018815994262695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,1,power_law_1.01,0.11586624145507814
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,1,power_law_1.2,1.5163258361816407
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,1,power_law_1.2,0.5890892791748047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,1,power_law_1.01,0.18044351577758788
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,1,power_law_1.01,0.2007686424255371
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,1,power_law_1.01,0.19373184204101562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,1,power_law_1.01,0.20722047805786134
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,1,power_law_1.01,0.21813823699951174
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,1,power_law_1.01,0.23109247207641603
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,1,power_law_1.01,0.2581299209594726
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,1,power_law_1.01,0.26689088821411133
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,1,power_law_1.01,0.2732467269897461
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,1,power_law_1.01,0.28136383056640624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,1,power_law_1.01,0.30169536590576174
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,1,power_law_1.01,0.3561363220214844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,1,power_law_1.01,0.4353631973266602
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,1,power_law_1.01,0.2605049514770508
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,1,power_law_1.01,0.5000384140014649
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,1,power_law_1.01,0.5831321716308594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,1,power_law_1.01,0.7726573181152344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,1,power_law_1.01,0.9430054473876954
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,1,power_law_1.2,3.05210693359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,1,power_law_1.01,1.3077958679199218
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,1,power_law_1.01,0.08164992332458496
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,1,power_law_1.01,1.6746099853515624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,1,power_law_1.01,0.08694784164428711
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,1,power_law_1.01,0.09827584266662598
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,1,power_law_1.01,0.11558015823364258
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,1,power_law_1.01,2.400524139404297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,1,power_law_1.01,0.1325107192993164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,1,power_law_1.01,0.18552576065063478
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,1,power_law_1.01,0.19351936340332032
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,1,power_law_1.01,0.20046079635620115
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,1,power_law_1.01,0.2039743995666504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,1,power_law_1.01,0.2076915168762207
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,1,power_law_1.01,3.1331890869140624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,1,power_law_1.01,0.21234495162963868
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,1,power_law_1.01,0.27523967742919925
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,1,power_law_1.01,0.21815231323242185
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,1,power_law_1.01,0.2555539131164551
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,1,power_law_1.01,0.2628761672973633
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,1,power_law_1.01,0.2610521507263184
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,16,1,power_law_1.01,3.8660491943359374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,1,power_law_1.01,0.2951328086853027
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,1,power_law_1.01,0.3223040008544922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,1,power_law_1.01,0.34584705352783207
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,1,power_law_1.01,0.40187328338623046
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,1,power_law_1.01,0.483678092956543
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,1,power_law_1.01,0.6175641632080078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,1,power_law_1.01,0.7366643524169921
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,1,power_law_1.01,1.0042304229736327
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,1,power_law_1.01,1.2655903625488283
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,1,power_law_1.01,1.785356750488281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,16,1,power_law_1.2,3.836098022460938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,1,power_law_1.01,2.331005401611328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,16,1,power_law_1.01,2.8549185180664063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,1,power_law_1.01,0.060189437866210935
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,1,power_law_1.01,0.1402841567993164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,1,power_law_1.01,0.05478528022766114
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,1,power_law_1.01,0.09039487838745117
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,1,power_law_1.01,0.1229478359222412
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,1,power_law_1.01,0.14459391593933107
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,1,power_law_1.01,0.07361023902893067
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,1,power_law_1.01,0.1468569564819336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,1,power_law_1.01,0.04845823764801026
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,1,power_law_1.01,0.13204928398132326
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,1,power_law_1.01,0.1564729595184326
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,1,power_law_1.01,0.15376959800720216
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,1,power_law_1.01,0.1636742401123047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,1,power_law_1.01,0.16993280410766604
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,1,power_law_1.01,0.17894912719726563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,1,power_law_1.01,0.1897100830078125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,1,power_law_1.01,0.20893632888793948
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,1,power_law_1.01,0.24948415756225586
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,1,power_law_1.01,0.2903980827331543
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,1,power_law_1.01,0.38041919708251953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,1,power_law_1.01,0.47263103485107416
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,1,power_law_1.01,0.6610118103027344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,1,power_law_1.2,0.06665535926818847
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,1,power_law_1.01,0.8537593841552734
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,1,power_law_1.2,0.08494527816772461
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,1,power_law_1.2,0.09342399597167969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,1,power_law_1.2,0.1119718360900879
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,1,power_law_1.2,0.12242495536804199
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,1,power_law_1.2,0.19000896453857422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,1,power_law_1.2,0.19536640167236327
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,1,power_law_1.2,0.20542335510253906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,1,power_law_1.2,0.21812671661376953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,1,power_law_1.2,0.22489343643188478
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,1,power_law_1.2,0.2400262451171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,1,power_law_1.2,0.27139455795288087
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,1,power_law_1.2,0.26586559295654294
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,1,power_law_1.2,0.2724115180969238
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,1,power_law_1.2,0.2784351921081543
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,1,power_law_1.2,0.28672767639160157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,1,power_law_1.2,0.3102495956420898
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,1,power_law_1.2,0.364851188659668
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,1,power_law_1.2,0.4373247909545898
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,1,power_law_1.2,0.5048339080810547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,1,power_law_1.2,0.5895999908447266
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,1,power_law_1.2,0.7750969696044921
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,1,power_law_1.2,0.9507878112792969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,1,power_law_1.2,1.3146035766601563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,1,power_law_1.2,1.6813267517089845
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,1,power_law_1.2,2.4118829345703126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,1,power_law_1.2,3.1482470703125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,16,1,power_law_1.2,3.8807788085937505
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,1,power_law_1.2,0.0906713581085205
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,1,power_law_1.2,0.09924287796020508
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,1,power_law_1.2,0.08132927894592286
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,1,power_law_1.2,0.13079744338989258
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,1,power_law_1.01,1.2561753845214843
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,1,power_law_1.2,0.18528383255004882
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,1,power_law_1.01,1.6640205383300781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,1,power_law_1.2,0.19389503479003906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,1,power_law_1.2,0.20731775283813478
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,1,power_law_1.2,0.20156671524047853
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,1,power_law_1.2,0.2105331230163574
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,1,power_law_1.2,0.2151321601867676
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,1,power_law_1.01,2.520272064208984
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,1,power_law_1.2,0.22085952758789062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,1,power_law_1.2,0.27429632186889646
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,1,power_law_1.2,0.27858816146850585
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,1,power_law_1.2,0.28361215591430666
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,1,power_law_1.01,3.3929824829101562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,1,power_law_1.2,0.30464319229125975
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,1,power_law_1.2,0.32667774200439453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,1,power_law_1.2,0.353875846862793
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,1,power_law_1.2,0.42710014343261715
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,1,power_law_1.2,0.48853374481201167
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,1,power_law_1.2,0.626135025024414
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,16,1,power_law_1.01,4.280698852539063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,1,power_law_1.2,0.7547526550292969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,1,power_law_1.2,1.0095468902587892
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,1,power_law_1.2,0.29074111938476566
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,1,power_law_1.2,1.2721478271484377
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,1,power_law_1.2,0.04839039802551269
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,1,power_law_1.2,0.05419199943542481
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,1,power_law_1.2,1.796669464111328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,1,power_law_1.2,0.07047359943389893
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,1,power_law_1.2,0.08431872367858886
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,1,power_law_1.2,0.12457728385925293
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,1,power_law_1.2,2.346001892089844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,1,power_law_1.2,0.1344870376586914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,1,power_law_1.2,0.14160767555236817
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,1,power_law_1.2,0.14688192367553712
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,1,power_law_1.2,0.1529593563079834
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,16,1,power_law_1.2,2.8688531494140626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,1,power_law_1.2,0.15765695571899413
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,1,power_law_1.2,0.16647104263305662
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,1,power_law_1.2,0.17296831130981444
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,1,power_law_1.2,0.18267328262329102
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,1,power_law_1.2,0.10453184127807617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,1,power_law_1.2,0.20931776046752928
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,1,power_law_1.2,0.34970367431640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,1,power_law_1.2,0.23728000640869143
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,1,power_law_1.2,0.19717056274414063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,1,power_law_1.2,0.4850764846801757
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,1,power_law_1.2,0.6237977600097656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,1,power_law_1.2,0.2894035148620605
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,1,power_law_1.2,1.2537779235839843
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,1,power_law_1.2,1.9670956420898438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,1,power_law_1.2,0.9326630401611329
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,1,power_law_1.01,0.11770879745483398
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,1,power_law_1.01,0.1315244770050049
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,1,power_law_1.01,0.13847167968750002
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,1,power_law_1.01,0.18622400283813476
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,1,power_law_1.2,2.690118408203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,1,power_law_1.01,0.22610111236572267
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,1,power_law_1.01,0.24753152847290041
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,1,power_law_1.01,0.3239775848388672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,1,power_law_1.01,0.34618431091308594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,1,power_law_1.01,0.36556671142578123
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,1,power_law_1.01,0.3788288116455078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,1,power_law_1.01,0.3853343963623047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,1,power_law_1.01,0.40250240325927733
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,1,power_law_1.01,0.41657470703125005
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,1,power_law_1.01,0.43875583648681643
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,1,power_law_1.01,0.462913932800293
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,1,power_law_1.01,0.6094188690185547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,1,power_law_1.2,4.175040588378907
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,1,power_law_1.01,0.6407673645019532
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,1,power_law_1.01,0.6807174682617188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,1,power_law_1.01,0.8165599822998046
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,1,power_law_1.01,0.8903775787353515
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,1,power_law_1.01,1.0411302185058593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,1,power_law_1.01,1.2331858825683593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,1,power_law_1.01,1.6337805175781248
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,1,power_law_1.2,5.563746337890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,1,power_law_1.01,2.0160140991210938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,1,power_law_1.01,0.1450214385986328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,1,power_law_1.01,0.15232255935668945
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,1,power_law_1.01,2.789039306640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,1,power_law_1.01,0.16328704833984375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,1,power_law_1.01,0.17887359619140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,1,power_law_1.01,0.2561484718322754
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,1,power_law_1.01,0.2934828758239746
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,1,power_law_1.01,3.5629574584960935
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,1,power_law_1.01,0.33528766632080076
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,1,power_law_1.01,0.21675071716308594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,16,1,power_law_1.01,4.359823303222656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,1,power_law_1.01,0.3475884628295899
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,1,power_law_1.01,0.35585025787353514
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,1,power_law_1.01,0.36819072723388674
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,1,power_law_1.01,0.3790131378173828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,1,power_law_1.01,0.3870336151123047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,1,power_law_1.01,0.41776256561279296
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,1,power_law_1.01,0.40422977447509767
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,1,power_law_1.01,0.4290752029418945
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,1,power_law_1.01,0.4901785659790039
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,1,power_law_1.01,0.5189670562744141
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,1,power_law_1.01,0.6113932800292969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,1,power_law_1.01,0.5451820755004884
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,1,power_law_1.01,0.6734214019775391
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,1,power_law_1.01,0.8178643035888673
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,1,power_law_1.01,0.9423993682861328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,1,power_law_1.01,1.215893096923828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,1,power_law_1.01,1.4863314819335938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,1,power_law_1.01,2.018765411376953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,1,power_law_1.01,2.5718841552734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,16,1,power_law_1.01,3.1117144775390626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,1,power_law_1.01,0.08365632057189941
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,1,power_law_1.01,0.08366847991943359
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,1,power_law_1.01,0.08617664337158203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,1,power_law_1.01,0.11011199951171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,1,power_law_1.01,0.0893881607055664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,1,power_law_1.01,0.09628288269042969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,1,power_law_1.01,0.08519488334655761
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,1,power_law_1.01,0.11680576324462891
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,1,power_law_1.01,0.12348480224609375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,1,power_law_1.01,0.13719552040100097
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,1,power_law_1.01,0.10265536308288574
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,1,power_law_1.01,0.17892543792724608
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,1,power_law_1.01,0.18794431686401367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,1,power_law_1.01,0.2340230369567871
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,1,power_law_1.01,0.2956236839294434
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,1,power_law_1.01,0.42347839355468747
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,1,power_law_1.01,0.33590721130371093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.01,0.7711277008056641
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,1,power_law_1.01,0.5960934448242188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.01,1.1211654663085937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.01,2.178572235107422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.01,1.4750349426269531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.01,2.893779296875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,1,power_law_1.2,0.12626303672790526
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,1,power_law_1.2,0.14350079536437987
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,1,power_law_1.2,0.16747392654418944
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,1,power_law_1.2,0.23148927688598633
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,1,power_law_1.2,0.29007936477661134
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,1,power_law_1.2,0.3276268768310547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,1,power_law_1.2,0.21935359954833986
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,1,power_law_1.2,0.34893375396728515
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,1,power_law_1.2,0.3684864044189453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.01,4.288713073730468
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,1,power_law_1.2,0.3832320022583008
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,1,power_law_1.2,0.39429630279541017
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,1,power_law_1.2,0.415109748840332
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,1,power_law_1.2,0.4426412963867188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,1,power_law_1.2,0.6079532623291015
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,1,power_law_1.2,0.47012416839599613
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,1,power_law_1.2,0.6427129364013672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,1,power_law_1.2,0.69927490234375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,1,power_law_1.2,0.8283999633789062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,1,power_law_1.2,0.8995641326904297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,1,power_law_1.2,1.0572876739501953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,1,power_law_1.2,1.2558470153808594
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,1,power_law_1.2,1.6468966674804686
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,1,power_law_1.2,2.035960998535156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,1,power_law_1.2,2.8203442382812502
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,1,power_law_1.2,0.15224127769470214
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,1,power_law_1.2,0.16240896224975584
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,1,power_law_1.2,3.5968856811523438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,1,power_law_1.2,0.17935359954833985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,16,1,power_law_1.2,4.381401672363281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,1,power_law_1.2,0.20720319747924804
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,1,power_law_1.2,0.24345151901245118
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,1,power_law_1.2,0.2743206405639648
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,1,power_law_1.2,0.3266329574584961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,1,power_law_1.2,0.3376947021484375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,1,power_law_1.2,0.34771457672119144
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,1,power_law_1.2,0.36270591735839847
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,1,power_law_1.2,0.1454681587219238
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,1,power_law_1.2,0.3812179183959961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,1,power_law_1.2,0.3974028778076172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,1,power_law_1.2,0.4121971130371094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,1,power_law_1.2,0.37351871490478517
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,1,power_law_1.2,0.42792831420898436
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,1,power_law_1.2,0.4878246307373047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,1,power_law_1.2,0.5293791961669922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,1,power_law_1.2,0.562848014831543
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,1,power_law_1.2,0.6313721466064453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,1,power_law_1.2,0.698463363647461
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,1,power_law_1.2,0.8519161224365235
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,1,power_law_1.2,0.9786847686767578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,1,power_law_1.2,1.2506137847900392
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,1,power_law_1.2,1.522158660888672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,1,power_law_1.2,0.08370240211486817
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,1,power_law_1.2,2.070897979736328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,1,power_law_1.2,2.6052539062499998
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,1,power_law_1.2,0.08360320091247558
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,16,1,power_law_1.2,3.1546841430664063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,1,power_law_1.2,0.08483648300170898
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,1,power_law_1.2,0.08572928428649902
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,1,power_law_1.2,0.08970047950744628
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,1,power_law_1.2,0.11673407554626465
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,1,power_law_1.2,0.10371199607849121
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,1,power_law_1.2,0.09629695892333985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,1,power_law_1.2,0.11434816360473632
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,1,power_law_1.2,0.1244371223449707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,1,power_law_1.2,0.1744076728820801
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,1,power_law_1.2,0.1372812843322754
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,1,power_law_1.2,0.19360639572143554
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,1,power_law_1.2,0.42330944061279296
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,1,power_law_1.2,0.23271743774414064
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,1,power_law_1.2,0.29532800674438475
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,1,power_law_1.2,0.3381792068481445
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,1,power_law_1.2,0.5966758346557617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.2,0.7724678039550781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.2,1.47542724609375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.2,2.1801222229003905
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.2,4.286597900390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,1,power_law_1.01,0.03883968114852905
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,1,power_law_1.01,0.03903359889984131
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,1,power_law_1.01,0.044949760437011724
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,1,power_law_1.01,0.049085440635681155
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,1,power_law_1.01,0.07395840167999268
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,1,power_law_1.01,0.07586431980133057
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,1,power_law_1.01,0.07709824085235596
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,1,power_law_1.01,0.08283647537231445
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,1,power_law_1.01,0.08666879653930663
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,1,power_law_1.01,0.0894489574432373
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,1,power_law_1.01,0.09134976387023926
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,1,power_law_1.01,0.09694848060607911
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,1,power_law_1.01,0.10019200325012208
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,1,power_law_1.01,0.10095935821533204
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,1,power_law_1.01,0.11349120140075683
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,1,power_law_1.01,0.12323904037475586
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,1,power_law_1.01,0.15066431999206542
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,1,power_law_1.01,0.1809222412109375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,1,power_law_1.01,0.24856319427490234
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,1,power_law_1.01,0.3070636749267578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,1,power_law_1.01,0.4317574310302734
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,1,power_law_1.01,0.5545280075073242
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,1,power_law_1.01,0.07847167968749999
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,1,power_law_1.01,0.0739686393737793
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,1,power_law_1.01,0.08027839660644531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,1,power_law_1.01,0.032052481174468996
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,1,power_law_1.01,0.03365056037902832
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,1,power_law_1.01,0.03615744113922119
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,1,power_law_1.01,0.8020569610595704
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,1,power_law_1.01,0.038192000389099125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,1,power_law_1.01,0.044037761688232424
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,1,power_law_1.01,1.046216354370117
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,1,power_law_1.01,0.04436160087585449
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,1,power_law_1.01,0.04561855792999268
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,32,1,power_law_1.01,1.2920121765136718
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,1,power_law_1.01,0.04699135780334472
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,1,power_law_1.01,0.04673535823822021
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,1,power_law_1.01,0.04476928234100342
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,1,power_law_1.01,0.04673920154571533
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,1,power_law_1.01,0.048749442100524905
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,1,power_law_1.01,0.0502348804473877
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,1,power_law_1.01,0.056254081726074226
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,1,power_law_1.01,0.059502081871032717
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,1,power_law_1.01,0.060316162109375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,1,power_law_1.01,0.05866496086120605
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,1,power_law_1.01,0.06535744190216064
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,1,power_law_1.01,0.07682559967041017
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,1,power_law_1.01,0.09210111618041991
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,1,power_law_1.01,0.1477126407623291
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,1,power_law_1.01,0.10867391586303712
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,1,power_law_1.01,0.17984128952026368
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,1,power_law_1.01,0.2572691154479981
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,1,power_law_1.01,0.3355718231201172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,1,power_law_1.01,0.48679744720458984
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.2,1.1289798736572265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,1,power_law_1.01,0.6293062210083008
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,32,1,power_law_1.01,0.774181137084961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.2,2.875727233886719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,1,power_law_1.01,0.019360640048980714
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,1,power_law_1.01,0.02099071979522705
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,1,power_law_1.01,0.0205785608291626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,1,power_law_1.01,0.021000320911407473
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,1,power_law_1.01,0.018139519691467286
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,1,power_law_1.01,0.02099839925765991
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,1,power_law_1.01,0.021395840644836427
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,1,power_law_1.01,0.018933119773864745
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,1,power_law_1.01,0.020535039901733398
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,1,power_law_1.01,0.020553600788116456
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,1,power_law_1.01,0.020974080562591552
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,1,power_law_1.01,0.022024960517883302
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,1,power_law_1.01,0.02301055908203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,1,power_law_1.01,0.021825919151306154
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,1,power_law_1.01,0.02423487901687622
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,1,power_law_1.01,0.026352639198303225
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,1,power_law_1.01,0.02841023921966553
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,1,power_law_1.01,0.03431999921798706
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.01,0.040330238342285156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.01,0.06150847911834717
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.01,0.05199488162994385
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.01,0.08329471588134765
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,1,power_law_1.2,0.03578432083129883
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,1,power_law_1.2,0.04679168224334716
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,1,power_law_1.2,0.047331199645996094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.01,0.10346943855285644
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,1,power_law_1.2,0.06446080207824707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,1,power_law_1.2,0.07722815990447998
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,1,power_law_1.2,0.04777152061462402
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,1,power_law_1.2,0.08019840240478515
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,1,power_law_1.2,0.07742208003997803
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,1,power_law_1.2,0.07961408138275147
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,1,power_law_1.2,0.08262720108032226
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,1,power_law_1.2,0.08616512298583985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,1,power_law_1.2,0.08853631973266601
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,1,power_law_1.2,0.08806976318359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,1,power_law_1.2,0.09185728073120117
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,1,power_law_1.2,0.09662015914916992
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,1,power_law_1.2,0.10185983657836914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,1,power_law_1.2,0.10165760040283203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,1,power_law_1.2,0.11408639907836915
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,1,power_law_1.2,0.12351615905761719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,1,power_law_1.2,0.1516140842437744
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,1,power_law_1.2,0.1818547248840332
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,1,power_law_1.2,0.25022975921630863
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,1,power_law_1.2,0.30815935134887695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,1,power_law_1.2,0.43265472412109374
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,1,power_law_1.2,0.557858543395996
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,1,power_law_1.2,0.8031942749023437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,1,power_law_1.2,0.03532351970672608
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,1,power_law_1.2,0.03661632061004639
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,1,power_law_1.2,0.036962559223175054
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,1,power_law_1.2,0.03896575927734375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,32,1,power_law_1.2,1.2935615539550782
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,1,power_law_1.2,0.04419839859008789
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,1,power_law_1.2,0.045564799308776854
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,1,power_law_1.2,0.04707712173461914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,1,power_law_1.2,0.04760447978973389
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,1,power_law_1.2,0.047680001258850094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,1,power_law_1.2,0.04921919822692871
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,1,power_law_1.2,0.05005951881408691
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,1,power_law_1.2,0.05130623817443848
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,1,power_law_1.2,0.05640704154968261
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,1,power_law_1.2,0.060242562294006353
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.01,0.14043840408325195
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,1,power_law_1.2,0.06221439838409424
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,1,power_law_1.2,0.06560192108154297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,1,power_law_1.2,0.07479616165161132
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,1,power_law_1.2,0.09159168243408203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,1,power_law_1.2,0.044997758865356445
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,1,power_law_1.2,0.1083897590637207
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,1,power_law_1.2,0.14743552207946778
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,1,power_law_1.2,0.18012800216674804
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,1,power_law_1.2,0.25717119216918943
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,1,power_law_1.2,0.3352576065063476
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,1,power_law_1.2,1.0450579071044923
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,1,power_law_1.2,0.4869177627563476
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,1,power_law_1.2,0.05745984077453613
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,1,power_law_1.2,0.018506879806518557
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,1,power_law_1.2,0.6333779144287109
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,1,power_law_1.2,0.018952319622039797
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,1,power_law_1.2,0.020570240020751952
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,1,power_law_1.2,0.02055743932723999
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,1,power_law_1.2,0.020944640636444092
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,1,power_law_1.2,0.020742399692535402
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,1,power_law_1.2,0.0209932804107666
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,32,1,power_law_1.2,0.768334732055664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,1,power_law_1.2,0.02055999994277954
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,1,power_law_1.2,0.020992639064788817
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.01,0.26692352294921873
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,1,power_law_1.2,0.023034880161285402
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.01,0.35020606994628906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,1,power_law_1.2,0.026044158935546874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.01,0.18431999206542968
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,32,1,power_law_1.01,0.4357247924804687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,1,power_law_1.2,0.03479935884475708
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,1,power_law_1.2,0.017724159955978393
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,1,power_law_1.2,0.022401919364929197
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,1,power_law_1.2,0.021862399578094483
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.2,0.08592639923095703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.2,0.10692159652709961
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,1,power_law_1.2,0.020288000106811525
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.2,0.05255680084228516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.2,0.04075007915496826
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,1,power_law_1.2,0.024491519927978517
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,1,power_law_1.2,0.02878783941268921
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,1,power_law_1.01,0.05744319915771484
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,1,power_law_1.01,0.06248640060424805
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,1,power_law_1.01,0.06331071853637696
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,32,1,power_law_1.2,0.4519935989379883
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,1,power_law_1.01,0.07247360229492188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,1,power_law_1.01,0.11455167770385741
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,1,power_law_1.01,0.11595711708068848
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,1,power_law_1.01,0.11664447784423829
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,1,power_law_1.01,0.11721728324890138
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,1,power_law_1.01,0.11868160247802735
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,1,power_law_1.01,0.12004416465759278
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,1,power_law_1.01,0.12465791702270508
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,1,power_law_1.01,0.12805695533752443
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,1,power_law_1.01,0.12973440170288086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,1,power_law_1.01,0.1331494426727295
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,1,power_law_1.01,0.1367296028137207
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,1,power_law_1.01,0.14095359802246094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,1,power_law_1.01,0.14467967987060545
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,1,power_law_1.01,0.162159366607666
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,1,power_law_1.01,0.1776755142211914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,1,power_law_1.01,0.229549446105957
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,1,power_law_1.01,0.2771718406677246
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,1,power_law_1.01,0.38102718353271486
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,1,power_law_1.01,0.4765254211425781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,1,power_law_1.01,0.6718994903564453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,1,power_law_1.01,0.8658131408691407
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,1,power_law_1.01,1.2593766021728516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,1,power_law_1.01,0.03901439905166626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,1,power_law_1.01,0.04146815776824951
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,1,power_law_1.01,0.0436627197265625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,1,power_law_1.01,1.6720518493652343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,1,power_law_1.01,0.04608831882476806
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,1,power_law_1.01,0.055326719284057614
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,1,power_law_1.01,0.055000958442687986
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,1,power_law_1.01,0.05681920051574707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,1,power_law_1.01,0.05700352191925049
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,1,power_law_1.01,0.05816319942474365
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,32,1,power_law_1.01,2.045948791503906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,1,power_law_1.01,0.05743616104125977
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,1,power_law_1.01,0.05984576225280762
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,1,power_law_1.01,0.06063871860504151
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,1,power_law_1.01,0.06382847785949707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,1,power_law_1.01,0.06694015979766846
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,1,power_law_1.01,0.06952767848968507
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,1,power_law_1.01,0.0734438419342041
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,1,power_law_1.01,0.07686592102050781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,1,power_law_1.01,0.08658687591552734
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,1,power_law_1.01,0.09840576171875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,1,power_law_1.01,0.1173516845703125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,1,power_law_1.01,0.14215488433837892
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,1,power_law_1.01,0.20100416183471678
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,1,power_law_1.01,0.24561088562011718
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,1,power_law_1.01,0.3501644897460937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,1,power_law_1.01,0.45313152313232424
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,1,power_law_1.01,0.6611417388916015
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,1,power_law_1.01,0.8552787017822265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,32,1,power_law_1.01,1.0606880187988281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.2,0.06300032138824463
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.2,0.14769791603088378
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.2,0.1921900749206543
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.2,0.27861824035644533
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.2,0.36503681182861325
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,1,power_law_1.01,0.020184319019317627
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,1,power_law_1.01,0.023382399082183838
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,1,power_law_1.01,0.023427200317382813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,1,power_law_1.01,0.02099967956542969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,1,power_law_1.01,0.023824639320373535
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,1,power_law_1.01,0.01936959981918335
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,1,power_law_1.01,0.022890241146087648
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,1,power_law_1.01,0.02345472097396851
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,1,power_law_1.01,0.023447680473327636
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,1,power_law_1.01,0.023016319274902344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,1,power_law_1.01,0.023173758983612062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,1,power_law_1.01,0.025122559070587157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,1,power_law_1.01,0.026299519538879396
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,1,power_law_1.01,0.029667840003967282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,1,power_law_1.01,0.030213758945465085
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,1,power_law_1.01,0.0361631989479065
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,1,power_law_1.01,0.044745597839355465
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,1,power_law_1.01,0.03209919929504394
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.01,0.05640192031860351
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.01,0.0812435245513916
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.01,0.06887104034423827
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,1,power_law_1.2,0.058860797882080075
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,1,power_law_1.2,0.061598081588745114
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,1,power_law_1.2,0.06312895774841308
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,1,power_law_1.2,0.0727513599395752
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,1,power_law_1.2,0.11462464332580566
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.01,0.10930944442749024
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,1,power_law_1.2,0.1153651237487793
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,1,power_law_1.2,0.11724543571472168
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,1,power_law_1.2,0.11742464065551758
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,1,power_law_1.2,0.11912768363952637
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,1,power_law_1.2,0.11991616249084473
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.01,0.13792896270751953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,1,power_law_1.2,0.1252723217010498
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,1,power_law_1.2,0.12959360122680663
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,1,power_law_1.2,0.12816320419311525
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,1,power_law_1.2,0.13350144386291504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,1,power_law_1.2,0.1379641628265381
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,1,power_law_1.2,0.14317567825317384
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,1,power_law_1.2,0.14545023918151856
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,1,power_law_1.2,0.1623334312438965
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,1,power_law_1.2,0.1793600082397461
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,1,power_law_1.2,0.2339641571044922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,1,power_law_1.2,0.2806559944152832
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,1,power_law_1.2,0.38130687713623046
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.01,0.1974617576599121
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,1,power_law_1.2,0.4784422302246094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,1,power_law_1.2,0.6726342773437499
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,1,power_law_1.2,0.8666604614257812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.01,0.2564211273193359
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,1,power_law_1.2,0.037781760692596436
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,1,power_law_1.2,0.04172800064086914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,1,power_law_1.2,0.04649407863616943
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,1,power_law_1.2,1.2599097442626952
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,1,power_law_1.2,0.05454592227935791
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,1,power_law_1.2,0.056290559768676765
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,1,power_law_1.2,0.05555903911590576
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,1,power_law_1.2,1.6639199829101563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,1,power_law_1.2,0.05825344085693359
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,1,power_law_1.2,0.05719359874725342
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,1,power_law_1.2,0.060001921653747556
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,1,power_law_1.2,0.060622720718383795
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.01,0.37182590484619144
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,1,power_law_1.2,0.061229438781738275
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,1,power_law_1.2,0.0649561595916748
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,1,power_law_1.2,0.06710464000701905
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,1,power_law_1.2,0.06972224235534667
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,1,power_law_1.2,0.07401855945587157
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,32,1,power_law_1.2,2.063020782470703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,1,power_law_1.2,0.07791103839874267
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,1,power_law_1.2,0.08751744270324707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,1,power_law_1.2,0.09763072013854981
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,1,power_law_1.2,0.11626496315002441
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,1,power_law_1.2,0.142795524597168
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,1,power_law_1.2,0.20066495895385744
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,1,power_law_1.2,0.24556352615356447
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,1,power_law_1.2,0.3499135971069336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,1,power_law_1.2,0.4529100799560547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,1,power_law_1.2,0.0418720006942749
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,1,power_law_1.2,0.019368319511413573
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,1,power_law_1.2,0.019751039743423463
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,1,power_law_1.2,0.021069440841674805
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,1,power_law_1.2,0.6623961639404297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,1,power_law_1.2,0.023058559894561768
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,1,power_law_1.2,0.02262847900390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,1,power_law_1.2,0.023016319274902344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,1,power_law_1.2,0.023400321006774902
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,1,power_law_1.2,0.02349567890167236
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,1,power_law_1.2,0.8623468780517578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,1,power_law_1.2,0.023402879238128664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,1,power_law_1.2,0.023840000629425047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,1,power_law_1.2,0.024315519332885744
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,32,1,power_law_1.2,1.0584275054931642
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,1,power_law_1.2,0.026328320503234866
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,1,power_law_1.2,0.028794240951538087
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,1,power_law_1.2,0.03254271984100342
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,1,power_law_1.2,0.036590080261230465
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.01,0.48936321258544924
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.01,0.6105420684814453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,1,power_law_1.2,0.02509439945220947
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.2,0.057230720520019526
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,1,power_law_1.2,0.030713601112365724
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.2,0.07061439990997315
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.2,0.08434880256652832
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,1,power_law_1.2,0.045173120498657224
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.2,0.11367168426513671
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,1,power_law_1.01,0.06931263923645019
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,1,power_law_1.01,0.10078144073486328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,1,power_law_1.01,0.1141766357421875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,1,power_law_1.01,0.13425855636596679
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,1,power_law_1.01,0.15703488349914552
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,1,power_law_1.01,0.17051136016845703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,1,power_law_1.01,0.0775167989730835
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,1,power_law_1.01,0.17452415466308593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,1,power_law_1.01,0.1807417678833008
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,1,power_law_1.01,0.18444927215576173
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,1,power_law_1.01,0.19516351699829101
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,1,power_law_1.01,0.20563776016235352
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,1,power_law_1.01,0.2183283233642578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,1,power_law_1.01,0.241976318359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,1,power_law_1.01,0.32126014709472656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,1,power_law_1.01,0.3303488159179687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,1,power_law_1.01,0.3459360122680664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.2,0.1436575984954834
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,1,power_law_1.01,0.37639808654785156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,1,power_law_1.01,0.4054272079467774
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,1,power_law_1.01,0.4675443267822265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,1,power_law_1.01,0.527715835571289
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,1,power_law_1.01,0.691968002319336
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,1,power_law_1.01,0.8595980834960937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,1,power_law_1.01,1.1332742309570312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,1,power_law_1.01,1.438110656738281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,1,power_law_1.01,2.0177650451660156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,1,power_law_1.01,2.6103469848632814
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.2,0.20594303131103517
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,1,power_law_1.01,0.11231552124023439
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.2,0.26743871688842774
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,32,1,power_law_1.01,3.2163488769531248
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.2,0.5107884979248046
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,1,power_law_1.01,0.14057279586791993
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,1,power_law_1.01,0.18419519424438477
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,1,power_law_1.01,0.21776256561279297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,1,power_law_1.01,0.1079199981689453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,1,power_law_1.01,0.22290367126464844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.2,0.3878803253173828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,1,power_law_1.01,0.2282943916320801
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,1,power_law_1.01,0.23448511123657229
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,1,power_law_1.01,0.16012096405029297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.2,0.6351891326904296
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,1,power_law_1.01,0.2419635200500488
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,1,power_law_1.01,0.2659097671508789
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,1,power_law_1.01,0.2561619186401367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,1,power_law_1.01,0.2507769584655762
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,1,power_law_1.01,0.3174201583862305
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,1,power_law_1.01,0.32404735565185544
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,1,power_law_1.01,0.33567806243896486
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,1,power_law_1.01,0.35910209655761716
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,1,power_law_1.01,0.38716991424560543
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,1,power_law_1.01,0.4415008163452148
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,1,power_law_1.01,0.5164761734008789
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,1,power_law_1.01,0.6317139053344727
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,1,power_law_1.01,0.7402342224121095
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,1,power_law_1.01,0.9766764831542968
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,1,power_law_1.01,1.2288735961914061
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,1,power_law_1.01,1.7169331359863282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,1,power_law_1.01,2.2059039306640624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,1,power_law_1.01,0.11934975624084472
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,32,1,power_law_1.01,2.678752136230469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,1,power_law_1.01,0.07150144100189208
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,1,power_law_1.01,0.08238719940185547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,1,power_law_1.01,0.07345471858978272
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,1,power_law_1.01,0.09432576179504396
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,1,power_law_1.01,0.1711897659301758
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,1,power_law_1.01,0.17525440216064453
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,1,power_law_1.01,0.18348159790039062
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,1,power_law_1.01,0.13052351951599123
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,1,power_law_1.01,0.1906675148010254
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,1,power_law_1.01,0.19667327880859375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,1,power_law_1.01,0.1612006378173828
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,1,power_law_1.01,0.2058259201049805
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,1,power_law_1.01,0.21714239120483397
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,1,power_law_1.01,0.11247615814208985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,1,power_law_1.01,0.22660160064697266
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,1,power_law_1.01,0.16631168365478516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,1,power_law_1.01,0.24343168258666995
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.01,0.3306988906860352
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.01,0.40536064147949225
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.01,0.5099929428100586
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,1,power_law_1.01,0.2840492820739746
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.01,0.682757797241211
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.01,0.875066909790039
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,1,power_law_1.2,0.07254848003387451
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.01,1.2844461059570311
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,1,power_law_1.2,0.0989087963104248
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.01,1.6961900329589845
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,1,power_law_1.2,0.10879296302795409
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,1,power_law_1.2,0.07258048057556152
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,1,power_law_1.2,0.1283942413330078
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,1,power_law_1.2,0.13887999534606935
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,1,power_law_1.2,0.17382976531982422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,1,power_law_1.2,0.16847488403320313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,1,power_law_1.2,0.17983423233032228
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,1,power_law_1.2,0.18517824172973635
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,1,power_law_1.2,0.19549951553344727
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,1,power_law_1.2,0.21035839080810548
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,1,power_law_1.2,0.2217203140258789
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,1,power_law_1.2,0.2468288040161133
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,1,power_law_1.2,0.32417152404785154
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,1,power_law_1.2,0.33479873657226566
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,1,power_law_1.2,0.3488102340698242
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,1,power_law_1.2,0.38127296447753906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,1,power_law_1.2,0.41131774902343754
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,1,power_law_1.2,0.47268798828124997
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,1,power_law_1.2,0.5389247894287109
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,1,power_law_1.2,0.69683837890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,1,power_law_1.2,0.8637216186523438
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,1,power_law_1.2,1.1408979034423827
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,1,power_law_1.2,1.4441267395019532
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,1,power_law_1.2,0.10903871536254883
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,1,power_law_1.2,2.0310771179199216
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,1,power_law_1.2,0.11211392402648926
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,1,power_law_1.2,2.635347900390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,1,power_law_1.2,0.1346009635925293
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,1,power_law_1.2,0.11922623634338378
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,1,power_law_1.2,0.157706880569458
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,1,power_law_1.2,0.17508672714233398
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,1,power_law_1.2,0.21363840103149415
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,1,power_law_1.2,0.2228486442565918
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,1,power_law_1.2,0.22699199676513673
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,32,1,power_law_1.2,3.2364569091796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,1,power_law_1.2,0.23209600448608397
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,1,power_law_1.2,0.24130495071411132
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,1,power_law_1.2,0.25086015701293946
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,1,power_law_1.2,0.2587295913696289
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,1,power_law_1.2,0.2701657676696777
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,1,power_law_1.2,0.32158081054687504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,1,power_law_1.2,0.3318662261962891
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,1,power_law_1.2,0.3444217681884766
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,1,power_law_1.2,0.37622528076171874
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,1,power_law_1.2,0.404692497253418
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,1,power_law_1.2,0.47916927337646487
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,1,power_law_1.2,0.5356768035888673
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,1,power_law_1.2,0.6494681549072265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,1,power_law_1.2,0.7835430145263672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,1,power_law_1.2,0.9926809692382813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,1,power_law_1.2,1.2473197174072266
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,1,power_law_1.2,1.7335244750976564
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,1,power_law_1.2,2.222861480712891
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,1,power_law_1.2,0.07131968021392823
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,1,power_law_1.2,0.07464064121246337
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,1,power_law_1.2,0.08196352005004884
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,1,power_law_1.2,0.09238783836364746
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,1,power_law_1.2,0.10759167671203614
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,1,power_law_1.2,0.12383999824523925
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,32,1,power_law_1.2,2.6981887817382812
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,1,power_law_1.2,0.15896063804626465
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,1,power_law_1.2,0.16608064651489257
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,1,power_law_1.2,0.17137344360351564
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,1,power_law_1.2,0.17615232467651368
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,1,power_law_1.2,0.18534656524658205
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,1,power_law_1.2,0.19350719451904297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,1,power_law_1.2,0.20242944717407227
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,1,power_law_1.2,0.21585151672363284
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,1,power_law_1.2,0.2292595291137695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,1,power_law_1.2,0.241265926361084
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,32,1,power_law_1.01,2.4874169921875002
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,32,1,power_law_1.01,3.449693298339844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,32,1,power_law_1.01,4.204913330078125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,1,power_law_1.2,0.2715641593933106
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,1,power_law_1.2,0.33396224975585936
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.2,0.39123710632324216
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.2,0.5316531372070312
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.2,0.6686195373535156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.2,0.9822515106201172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,1,power_law_1.01,0.03820991992950439
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.2,1.2990931701660156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,1,power_law_1.01,0.05049024105072022
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,1,power_law_1.01,0.08681920051574707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,1,power_law_1.01,0.06627711772918701
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,1,power_law_1.01,0.09017855644226073
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,1,power_law_1.01,0.09136639595031738
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,1,power_law_1.01,0.0971660804748535
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,1,power_law_1.01,0.13699007987976075
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,1,power_law_1.01,0.1396979236602783
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,1,power_law_1.01,0.1456390380859375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,1,power_law_1.01,0.14959424018859863
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.2,2.018415985107422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,1,power_law_1.01,0.153821439743042
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,1,power_law_1.01,0.16330175399780272
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,1,power_law_1.01,0.20070655822753905
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,1,power_law_1.01,0.23950464248657227
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,1,power_law_1.01,0.27811840057373044
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,1,power_law_1.01,0.36772350311279295
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,1,power_law_1.01,0.4734476852416992
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,1,power_law_1.01,0.18243328094482422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,1,power_law_1.01,0.08308223724365235
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,1,power_law_1.01,0.637066879272461
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,1,power_law_1.01,0.8159986877441406
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,1,power_law_1.01,1.1703103637695311
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,32,1,power_law_1.01,1.8902951049804688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,1,power_law_1.01,0.05459839820861816
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,1,power_law_1.01,0.06030911922454833
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,1,power_law_1.01,1.5271781921386718
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,1,power_law_1.01,0.06484416007995605
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,1,power_law_1.01,0.07200704097747804
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,1,power_law_1.01,0.05314815998077392
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,1,power_law_1.01,0.09005120277404785
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,1,power_law_1.01,0.08889023780822754
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,1,power_law_1.01,0.08755711555480958
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,1,power_law_1.01,0.095797119140625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,1,power_law_1.01,0.09738112449645996
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,1,power_law_1.01,0.09187583923339844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,1,power_law_1.01,0.1157363224029541
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,1,power_law_1.01,0.11680319786071777
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,1,power_law_1.01,0.12356672286987305
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,1,power_law_1.01,0.12667200088500977
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,1,power_law_1.01,0.13046208381652832
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,1,power_law_1.01,0.14147135734558106
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,1,power_law_1.01,0.15820799827575685
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,1,power_law_1.01,0.18603519439697264
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,1,power_law_1.01,0.22009727478027347
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,1,power_law_1.01,0.2600351905822754
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,1,power_law_1.01,0.33032257080078126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,1,power_law_1.01,0.4144998550415039
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,1,power_law_1.01,0.5690227127075195
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,1,power_law_1.01,0.7197132873535156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,1,power_law_1.01,1.0160473632812501
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.2,2.7829931640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,1,power_law_1.01,1.3243052673339843
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,32,1,power_law_1.01,1.6293017578125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,1,power_law_1.2,0.03861311912536621
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,1,power_law_1.2,0.04658815860748291
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,1,power_law_1.2,0.0576793622970581
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,1,power_law_1.2,0.0621452808380127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,1,power_law_1.2,0.06568575859069824
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,1,power_law_1.2,0.0809331226348877
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,1,power_law_1.2,0.08580544471740723
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,1,power_law_1.2,0.08808768272399903
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,1,power_law_1.2,0.09220864295959472
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,1,power_law_1.2,0.0936780834197998
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,1,power_law_1.2,0.09788991928100585
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,1,power_law_1.2,0.13768128395080564
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,1,power_law_1.2,0.14046143531799316
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,1,power_law_1.2,0.14665151596069337
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,1,power_law_1.2,0.15024703979492188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,1,power_law_1.2,0.154266881942749
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,1,power_law_1.2,0.16426944732666016
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,1,power_law_1.2,0.18373823165893555
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,1,power_law_1.2,0.24045888900756837
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,1,power_law_1.2,0.20202239990234375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,1,power_law_1.2,0.2800985527038574
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,1,power_law_1.2,0.36775360107421873
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,1,power_law_1.2,0.47617473602294924
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,1,power_law_1.2,0.6389632034301758
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,1,power_law_1.2,0.8218784332275391
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,1,power_law_1.2,0.055027837753295894
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,1,power_law_1.2,0.05322559833526611
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,1,power_law_1.2,0.059159040451049805
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,1,power_law_1.2,1.53472900390625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,1,power_law_1.2,0.06394624233245849
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,1,power_law_1.2,0.07116799831390382
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,1,power_law_1.2,0.08931391716003419
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,1,power_law_1.2,0.0884921646118164
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,32,1,power_law_1.2,1.8967193603515624
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,1,power_law_1.2,0.09155776023864745
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,1,power_law_1.2,0.09570303916931153
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,1,power_law_1.2,0.09400192260742188
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,1,power_law_1.2,1.1737977600097655
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,1,power_law_1.2,0.0990783977508545
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,1,power_law_1.2,0.11858559608459474
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,1,power_law_1.2,0.11900416374206543
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,1,power_law_1.2,0.12560959815979006
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,1,power_law_1.2,0.13012928009033203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,1,power_law_1.2,0.13356032371520996
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,1,power_law_1.2,0.1435264015197754
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,1,power_law_1.2,0.16849727630615235
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,1,power_law_1.2,0.18885183334350586
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,1,power_law_1.2,0.22687295913696287
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,1,power_law_1.2,0.2628716850280762
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,1,power_law_1.2,0.3321318435668945
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,1,power_law_1.2,0.41974720001220706
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,1,power_law_1.2,0.5726764678955079
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,1,power_law_1.2,0.7242098999023437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,1,power_law_1.2,1.0175711822509765
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,1,power_law_1.2,1.3255520629882813
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,32,1,power_law_1.2,1.6327116394042966
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,1,power_law_1.01,0.08921600341796875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,1,power_law_1.01,0.10197440147399903
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,1,power_law_1.01,0.05827455997467042
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,1,power_law_1.01,0.1108255958557129
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,1,power_law_1.01,0.13755135536193847
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,1,power_law_1.01,0.14354047775268555
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,1,power_law_1.01,0.1497875213623047
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,1,power_law_1.01,0.07760447978973388
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,1,power_law_1.01,0.14968128204345704
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,1,power_law_1.01,0.1590841579437256
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,1,power_law_1.01,0.1713670349121094
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,1,power_law_1.01,0.18233280181884765
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,1,power_law_1.01,0.24282304763793947
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,1,power_law_1.01,0.26581951141357424
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,1,power_law_1.01,0.29709312438964847
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,1,power_law_1.01,0.24907840728759764
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,1,power_law_1.01,0.32679870605468747
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,1,power_law_1.01,0.3973644638061523
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,1,power_law_1.01,0.4899628829956055
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,1,power_law_1.01,0.6511052703857422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,1,power_law_1.01,0.22603839874267578
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,1,power_law_1.01,0.23348672866821288
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,1,power_law_1.01,0.7980697631835938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,1,power_law_1.01,1.0992249298095704
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,1,power_law_1.01,0.08444800376892089
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,1,power_law_1.01,0.09641152381896972
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,1,power_law_1.01,1.4092941284179688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,1,power_law_1.01,0.10746239662170409
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,1,power_law_1.01,0.12458751678466798
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,1,power_law_1.01,0.15123519897460938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,1,power_law_1.01,0.1575712013244629
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,1,power_law_1.01,2.018901824951172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,1,power_law_1.01,0.1678528022766113
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,1,power_law_1.01,0.17334655761718748
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,1,power_law_1.01,0.1769811248779297
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,1,power_law_1.01,0.1856051254272461
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,1,power_law_1.01,2.6389242553710934
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,1,power_law_1.01,0.21106367111206054
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,1,power_law_1.01,0.21635135650634765
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,1,power_law_1.01,0.21762752532958984
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,1,power_law_1.01,0.22851583480834964
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,32,1,power_law_1.01,3.261426696777344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,1,power_law_1.01,0.2488486480712891
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,1,power_law_1.01,0.2737484741210937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,1,power_law_1.01,0.29825855255126954
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,1,power_law_1.01,0.35142398834228517
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,1,power_law_1.01,0.41100799560546875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,1,power_law_1.01,0.5390969467163086
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,1,power_law_1.01,0.6473133087158203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,1,power_law_1.01,0.15874752044677734
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,1,power_law_1.01,0.888625946044922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,1,power_law_1.01,0.07917312145233155
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,1,power_law_1.01,1.1210739135742187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,1,power_law_1.01,1.5788275146484376
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,1,power_law_1.01,2.0590675354003904
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,32,1,power_law_1.01,2.541845703125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,1,power_law_1.2,0.08004287719726563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,1,power_law_1.2,0.060325760841369626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,1,power_law_1.2,0.0867750358581543
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,1,power_law_1.2,0.13565695762634278
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,1,power_law_1.2,0.10727104187011718
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,1,power_law_1.2,0.14386176109313964
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,1,power_law_1.2,0.0982636833190918
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,1,power_law_1.2,0.151145601272583
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,1,power_law_1.2,0.15685248374938965
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,1,power_law_1.2,0.1636396789550781
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,1,power_law_1.2,0.17256895065307618
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,1,power_law_1.2,0.1873523139953613
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,1,power_law_1.2,0.2334752082824707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,1,power_law_1.2,0.23831872940063475
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,1,power_law_1.2,0.24523071289062498
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,1,power_law_1.2,0.25238336563110353
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,1,power_law_1.2,0.26654399871826173
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,1,power_law_1.2,0.2993811225891113
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,1,power_law_1.2,0.3290591812133789
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,1,power_law_1.2,0.40888000488281245
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,1,power_law_1.2,0.4916025543212891
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,1,power_law_1.2,0.6547763061523437
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,1,power_law_1.2,0.8048480224609375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,1,power_law_1.2,1.1038406372070313
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,1,power_law_1.2,1.4146566772460938
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,1,power_law_1.2,2.0287628173828125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,1,power_law_1.2,0.08082048416137695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,1,power_law_1.2,0.07964352130889893
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,1,power_law_1.2,0.09333888053894043
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,1,power_law_1.2,0.11403327941894532
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,1,power_law_1.2,2.6488223266601563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,1,power_law_1.2,0.10463616371154785
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,1,power_law_1.2,0.15774208068847656
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,1,power_law_1.2,0.1509964847564697
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,1,power_law_1.2,0.1672640037536621
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,1,power_law_1.2,0.17215679168701173
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,1,power_law_1.2,0.17574207305908202
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,1,power_law_1.2,0.18307775497436524
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,32,1,power_law_1.2,3.273321533203125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,1,power_law_1.2,0.19586175918579102
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,1,power_law_1.2,0.2270182418823242
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,1,power_law_1.2,0.23190080642700198
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,1,power_law_1.2,0.2354431915283203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,1,power_law_1.2,0.2562950325012207
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,1,power_law_1.2,0.23841663360595705
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,1,power_law_1.2,0.3053977584838867
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,1,power_law_1.2,0.278482551574707
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,1,power_law_1.2,0.35974014282226563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,1,power_law_1.2,0.41605182647705075
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,1,power_law_1.2,0.5476780700683593
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,1,power_law_1.2,0.6532422637939452
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,1,power_law_1.2,0.8884857940673829
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,1,power_law_1.2,1.1252403259277344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,1,power_law_1.2,1.5850912475585939
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,1,power_law_1.2,2.0665939331054686
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,32,1,power_law_1.2,2.5499526977539064
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,1,power_law_1.01,0.09347328186035156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,1,power_law_1.01,0.1157913589477539
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,1,power_law_1.01,0.15549440383911134
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,1,power_law_1.01,0.19024959564208985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,1,power_law_1.01,0.23065727233886718
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,1,power_law_1.01,0.22256256103515626
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,1,power_law_1.01,0.23839487075805663
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,1,power_law_1.01,0.18955968856811523
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,1,power_law_1.01,0.24565055847167966
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,1,power_law_1.01,0.25650367736816404
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,1,power_law_1.01,0.27128000259399415
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,1,power_law_1.01,0.2953228759765625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,1,power_law_1.01,0.323612174987793
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,1,power_law_1.01,0.34163265228271483
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,1,power_law_1.01,0.4388761520385742
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,1,power_law_1.01,0.47367679595947265
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,1,power_law_1.01,0.5078144073486328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,1,power_law_1.01,0.6587264251708984
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,1,power_law_1.01,0.5768467330932617
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,1,power_law_1.01,0.8675615692138672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,1,power_law_1.01,1.0043449401855469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,1,power_law_1.01,1.3206124877929688
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,1,power_law_1.01,1.643396453857422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,1,power_law_1.01,0.14460415840148927
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,1,power_law_1.01,2.294145965576172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,1,power_law_1.01,0.14808447837829591
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,1,power_law_1.01,0.15776000022888184
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,1,power_law_1.01,2.944646301269531
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,1,power_law_1.01,0.1749817657470703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,1,power_law_1.01,0.20449920654296877
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,1,power_law_1.01,0.23654912948608398
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,32,1,power_law_1.01,3.6010772705078127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,1,power_law_1.01,0.2541228866577149
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,1,power_law_1.01,0.2933107185363769
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,1,power_law_1.01,0.30346431732177737
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,1,power_law_1.01,0.3116511917114258
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,1,power_law_1.01,0.3235398483276367
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,1,power_law_1.01,0.3310111999511719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,1,power_law_1.01,0.3404166412353516
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,1,power_law_1.01,0.3550726318359375
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,1,power_law_1.01,0.3793638229370117
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,1,power_law_1.01,0.3682988739013672
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,1,power_law_1.01,0.46001342773437504
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,1,power_law_1.01,0.48743423461914065
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,1,power_law_1.01,0.5793849563598632
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,1,power_law_1.01,0.7885427093505859
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,1,power_law_1.01,0.5147430419921875
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,1,power_law_1.01,0.6412806701660156
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,1,power_law_1.01,0.9126790618896484
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,1,power_law_1.01,1.187100830078125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,1,power_law_1.01,1.4566470336914061
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,1,power_law_1.01,1.9868301391601562
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,1,power_law_1.01,2.535181427001953
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,32,1,power_law_1.01,3.072931823730469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,1,power_law_1.01,0.09341631889343262
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,1,power_law_1.01,0.093569917678833
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,1,power_law_1.01,0.09481856346130371
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,1,power_law_1.01,0.11231936454772948
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,1,power_law_1.01,0.09237824440002441
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,1,power_law_1.01,0.09806719779968262
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,1,power_law_1.01,0.11848575592041015
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,1,power_law_1.01,0.12555583953857422
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,1,power_law_1.01,0.13288895606994627
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,1,power_law_1.01,0.14629247665405273
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,1,power_law_1.01,0.10441408157348633
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,1,power_law_1.01,0.1830463981628418
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,1,power_law_1.01,0.2999744033813477
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,1,power_law_1.01,0.22546688079833985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,1,power_law_1.01,0.18884031295776366
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,1,power_law_1.01,0.3407219314575195
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,1,power_law_1.01,0.4306579208374023
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,1,power_law_1.01,0.6014156723022461
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,1,power_law_1.01,1.1101484680175782
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,1,power_law_1.01,1.4539981079101563
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,1,power_law_1.01,0.7740096282958985
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,1,power_law_1.2,0.08649279594421386
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,1,power_law_1.01,2.144042205810547
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,1,power_law_1.2,0.09711999893188476
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,1,power_law_1.2,0.12801664352416992
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,1,power_law_1.2,0.14813183784484862
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,1,power_law_1.2,0.1832012748718262
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,1,power_law_1.2,0.18013376235961914
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,1,power_law_1.2,0.2130668830871582
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,1,power_law_1.01,2.8310336303710937
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,1,power_law_1.2,0.2264192008972168
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,1,power_law_1.2,0.2324281692504883
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,1,power_law_1.2,0.2438822364807129
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,1,power_law_1.01,4.235944213867187
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,1,power_law_1.2,0.25924671173095704
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,1,power_law_1.2,0.27
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,1,power_law_1.2,0.2978732872009277
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,1,power_law_1.2,0.3263078308105469
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,1,power_law_1.2,0.3416755294799805
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,1,power_law_1.2,0.4390764617919922
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,1,power_law_1.2,0.4792915344238281
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,1,power_law_1.2,0.5122918319702149
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,1,power_law_1.2,0.5929721450805664
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,1,power_law_1.2,0.11818367958068847
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,1,power_law_1.2,0.6837235260009765
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,1,power_law_1.2,0.8725785827636718
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,1,power_law_1.2,1.0115916442871093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,1,power_law_1.2,1.332641296386719
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,1,power_law_1.2,1.6502207946777343
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,1,power_law_1.2,0.1483027172088623
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,1,power_law_1.2,0.149487361907959
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,1,power_law_1.2,0.1567302417755127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,1,power_law_1.2,2.299255065917969
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,1,power_law_1.2,0.1714579200744629
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,1,power_law_1.2,0.20307199478149415
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,1,power_law_1.2,2.9660250854492185
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,1,power_law_1.2,0.22653184890747072
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,1,power_law_1.2,0.2427302360534668
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,1,power_law_1.2,0.28430015563964844
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,1,power_law_1.2,0.29587135314941404
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,32,1,power_law_1.2,3.624820556640625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,1,power_law_1.2,0.3050022315979004
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,1,power_law_1.2,0.3179417610168457
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,1,power_law_1.2,0.3275993728637695
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,1,power_law_1.2,0.3351795196533203
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,1,power_law_1.2,0.3488185501098633
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,1,power_law_1.2,0.36225601196289064
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,1,power_law_1.2,0.4566368103027344
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,1,power_law_1.2,0.37868671417236327
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,1,power_law_1.2,0.4974873733520508
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,1,power_law_1.2,0.5314988708496093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,1,power_law_1.2,0.5986975860595704
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,1,power_law_1.2,0.6644159698486328
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,1,power_law_1.2,0.8200294494628906
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,1,power_law_1.2,0.9469529724121093
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,1,power_law_1.2,1.2193702697753905
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,1,power_law_1.2,1.49223876953125
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,1,power_law_1.2,2.0406048583984377
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,1,power_law_1.2,0.09285951614379882
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,1,power_law_1.2,0.09424703598022462
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,1,power_law_1.2,2.5659194946289063
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,32,1,power_law_1.2,3.1164120483398436
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,1,power_law_1.2,0.09836352348327637
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,1,power_law_1.2,0.10545599937438965
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,1,power_law_1.2,0.0949503993988037
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,1,power_law_1.2,0.11890239715576172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,1,power_law_1.2,0.11189375877380373
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,1,power_law_1.2,0.12578559875488282
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,1,power_law_1.2,0.13306048393249512
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,1,power_law_1.2,0.1459436798095703
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,1,power_law_1.2,0.19340480804443358
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,1,power_law_1.2,0.23045440673828127
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,1,power_law_1.2,0.30281471252441405
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,1,power_law_1.2,0.09279744148254394
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,1,power_law_1.2,0.17475072860717772
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,1,power_law_1.2,0.3378265762329101
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,1,power_law_1.2,0.431256332397461
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,1,power_law_1.2,0.7737798309326172
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,1,power_law_1.2,0.6000716781616211
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,1,power_law_1.2,1.1159552001953126
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,1,power_law_1.2,1.4573458862304687
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,1,power_law_1.2,2.1440512084960934
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,1,power_law_1.2,2.8291558837890625
SGLang,0.5.6.post2,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,1,power_law_1.2,4.2261022949218745
