framework,version,device,op_name,kernel_source,mla_dtype,kv_cache_dtype,num_heads,batch_size,isl,tp_size,step,latency
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,4,1,0,0.6579679995775223
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,8,1,0,0.6474039927124977
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,16,1,0,0.6502599939703941
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,32,1,0,0.6508039981126785
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,64,1,0,0.6507599875330925
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,128,1,0,0.6495960056781769
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,256,1,0,0.6469919979572296
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,512,1,0,0.7067440003156662
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,1024,1,0,0.8115079998970032
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,2048,1,0,1.2531799972057343
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,4096,1,0,2.5007160007953644
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,8192,1,0,6.667247951030731
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1,16384,1,0,19.972471952438354
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,4,1,0,0.6561799943447113
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,8,1,0,0.6479719951748848
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,16,1,0,0.6488279923796654
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,32,1,0,0.652659997344017
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,64,1,0,0.6521279960870743
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,128,1,0,0.662428006529808
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,256,1,0,0.6942199915647507
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,512,1,0,0.7818000018596649
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,1024,1,0,1.1298880130052567
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,2048,1,0,2.0697439908981323
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,4096,1,0,4.829212009906769
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,8192,1,0,13.765651941299438
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,2,16384,1,0,41.14753580093384
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,4,1,0,0.6545120105147362
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,8,1,0,0.6448640003800392
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,16,1,0,0.6543639823794365
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,32,1,0,0.652876004576683
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,64,1,0,0.6524240002036095
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,128,1,0,0.7029520124197006
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,256,1,0,0.7922719940543175
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,512,1,0,1.0992519855499268
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,1024,1,0,1.870967984199524
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,2048,1,0,4.001775950193405
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,4096,1,0,9.723688125610352
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,8192,1,0,26.911351919174194
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,4,16384,1,0,84.89535999298096
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,4,1,0,0.6576679944992065
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,8,1,0,0.673787996172905
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,16,1,0,0.6667800024151802
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,32,1,0,0.6635160073637962
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,64,1,0,0.7053960040211678
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,128,1,0,0.7822359949350357
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,256,1,0,1.1039920002222061
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,512,1,0,1.8199799954891205
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,1024,1,0,3.6780159771442413
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,2048,1,0,7.878339886665344
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,4096,1,0,19.71511220932007
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,8192,1,0,57.239179611206055
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,8,16384,1,0,176.68762588500977
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,4,1,0,0.6789240017533302
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,8,1,0,0.6815799996256828
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,16,1,0,0.6848039925098419
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,32,1,0,0.7155079990625381
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,64,1,0,0.8172000050544739
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,128,1,0,1.1195400059223175
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,256,1,0,1.831016018986702
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,512,1,0,3.4613920152187347
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,1024,1,0,7.106316030025482
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,2048,1,0,16.10554802417755
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,4096,1,0,40.51217222213745
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,16,8192,1,0,113.77320003509521
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,4,1,0,0.7518640086054802
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,8,1,0,0.7428280040621758
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,16,1,0,0.7852359861135483
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,32,1,0,0.893264003098011
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,64,1,0,1.1644560396671295
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,128,1,0,1.8261639922857285
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,256,1,0,3.415708005428314
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,512,1,0,6.754871964454651
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,1024,1,0,14.139584064483643
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,2048,1,0,32.0031955242157
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,32,4096,1,0,82.86181640625
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,4,1,0,0.8983399942517281
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,8,1,0,0.9396520107984543
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,16,1,0,1.0295120179653168
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,32,1,0,1.2831800132989883
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,64,1,0,1.9121240079402924
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,128,1,0,3.426100015640259
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,256,1,0,6.622595965862274
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,512,1,0,13.70530390739441
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,1024,1,0,28.43551206588745
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,64,2048,1,0,66.33419513702393
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,4,1,0,1.2379599958658218
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,8,1,0,1.3415880054235458
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,16,1,0,1.593543991446495
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,32,1,0,2.1721400022506714
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,64,1,0,3.6389279663562775
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,128,1,0,6.640751957893372
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,256,1,0,13.303707957267761
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,512,1,0,27.950216054916382
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,128,1024,1,0,58.97118043899536
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,256,4,1,0,1.9848520010709763
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,256,8,1,0,2.204375982284546
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,256,16,1,0,2.7629840075969696
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,256,32,1,0,4.11188805103302
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,256,64,1,0,7.13619589805603
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,256,128,1,0,13.278395891189575
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,256,256,1,0,27.181243896484375
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,256,512,1,0,57.416791915893555
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,512,4,1,0,3.4088279604911804
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,512,8,1,0,3.954908013343811
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,512,16,1,0,5.406175971031189
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,512,32,1,0,7.937807977199554
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,512,64,1,0,14.06741213798523
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,512,128,1,0,26.876219749450684
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,512,256,1,0,54.99360370635986
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1024,4,1,0,6.409856021404266
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1024,8,1,0,7.752968013286591
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1024,16,1,0,10.429880142211914
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1024,32,1,0,16.102724313735962
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1024,64,1,0,28.2853000164032
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,128,1024,128,1,0,54.89971590042114
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,4,1,0,0.6687640026211739
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,8,1,0,0.6749840006232262
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,16,1,0,0.6707399934530258
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,32,1,0,0.6848840042948723
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,64,1,0,0.6729400008916855
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,128,1,0,0.6690879985690117
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,256,1,0,0.6691479980945587
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,512,1,0,0.6910679861903191
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,1024,1,0,0.7284680008888245
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,2048,1,0,0.9053879976272583
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,4096,1,0,1.4036439955234528
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,8192,1,0,3.4408399760723114
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1,16384,1,0,10.602932095527649
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,4,1,0,0.6659479960799217
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,8,1,0,0.665680006146431
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,16,1,0,0.6626200005412102
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,32,1,0,0.6694519892334938
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,64,1,0,0.6770680025219917
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,128,1,0,0.6703519895672798
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,256,1,0,0.6898240074515343
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,512,1,0,0.7279160022735596
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,1024,1,0,0.8465959951281548
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,2048,1,0,1.1984359920024872
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,4096,1,0,2.5818519592285156
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,8192,1,0,6.816031992435455
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,2,16384,1,0,20.57333207130432
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,4,1,0,0.6724439933896065
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,8,1,0,0.6775039881467819
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,16,1,0,0.681319996714592
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,32,1,0,0.6728760004043579
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,64,1,0,0.6706719920039177
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,128,1,0,0.6892919987440109
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,256,1,0,0.7444159984588623
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,512,1,0,0.82778400182724
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,1024,1,0,1.102592021226883
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,2048,1,0,2.149507999420166
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,4096,1,0,5.010996043682098
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,8192,1,0,13.882007837295532
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,4,16384,1,0,41.439048290252686
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,4,1,0,0.6839120015501976
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,8,1,0,0.6938720047473907
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,16,1,0,0.6729760020971298
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,32,1,0,0.6819400042295456
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,64,1,0,0.7054160088300705
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,128,1,0,0.7582199946045876
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,256,1,0,0.8431359976530075
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,512,1,0,1.0757680088281631
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,1024,1,0,1.945464015007019
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,2048,1,0,3.993828058242798
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,4096,1,0,10.031848073005676
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,8,8192,1,0,27.144871950149536
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,4,1,0,0.7076120004057884
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,8,1,0,0.689079999923706
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,16,1,0,0.7019919976592064
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,32,1,0,0.7147399932146072
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,64,1,0,0.7662840113043785
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,128,1,0,0.8470999971032143
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,256,1,0,1.07921202480793
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,512,1,0,1.8862040042877197
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,1024,1,0,3.5776120126247406
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,2048,1,0,8.05674397945404
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,16,4096,1,0,20.010511875152588
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,4,1,0,0.7337520122528076
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,8,1,0,0.722307987511158
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,16,1,0,0.7345560044050217
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,32,1,0,0.7813959941267967
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,64,1,0,0.8736399933695793
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,128,1,0,1.0961199849843979
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,256,1,0,1.9199840128421783
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,512,1,0,3.487412005662918
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,1024,1,0,7.0728879570961
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,32,2048,1,0,16.18887186050415
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,4,1,0,0.7962080091238022
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,8,1,0,0.8334359899163246
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,16,1,0,0.8629840016365051
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,32,1,0,0.9767120033502579
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,64,1,0,1.1631319969892502
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,128,1,0,1.957299992442131
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,256,1,0,3.5103200376033783
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,512,1,0,6.9216678738594055
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,64,1024,1,0,14.526756048202515
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,4,1,0,1.0473840236663818
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,8,1,0,1.0888999998569489
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,16,1,0,1.1794680058956146
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,32,1,0,1.3856960088014603
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,64,1,0,2.103007972240448
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,128,1,0,3.5853559970855713
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,256,1,0,6.92768394947052
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,128,512,1,0,14.014976143836975
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,256,4,1,0,1.5019360035657883
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,256,8,1,0,1.6124320030212402
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,256,16,1,0,1.7544160187244415
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,256,32,1,0,2.4601919651031494
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,256,64,1,0,3.8683679699897766
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,256,128,1,0,6.970903992652893
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,256,256,1,0,13.972259998321533
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,512,4,1,0,2.419171988964081
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,512,8,1,0,2.5877639651298523
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,512,16,1,0,3.2420759797096252
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,512,32,1,0,4.5870519280433655
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,512,64,1,0,7.5049479603767395
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,512,128,1,0,13.874611735343933
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1024,4,1,0,4.287072002887726
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1024,8,1,0,4.875432074069977
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1024,16,1,0,6.237671971321106
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1024,32,1,0,8.998255848884583
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,64,1024,64,1,0,15.146972060203552
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,4,1,0,0.6589239984750748
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,8,1,0,0.6634960100054741
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,16,1,0,0.6604319959878922
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,32,1,0,0.670584000647068
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,64,1,0,0.664048008620739
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,128,1,0,0.6658120080828667
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,256,1,0,0.6778239980340004
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,512,1,0,0.6724560037255287
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,1024,1,0,0.7005759924650192
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,2048,1,0,0.7329519987106323
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,4096,1,0,0.9790639877319336
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,8192,1,0,1.8602359741926193
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1,16384,1,0,5.131760001182556
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,4,1,0,0.6625919938087463
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,8,1,0,0.662943996489048
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,16,1,0,0.649619996547699
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,32,1,0,0.667212001979351
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,64,1,0,0.6547000035643578
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,128,1,0,0.6556279957294464
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,256,1,0,0.6601880043745041
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,512,1,0,0.6784520000219345
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,1024,1,0,0.7349399998784065
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,2048,1,0,0.874328002333641
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,4096,1,0,1.4372840076684952
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,8192,1,0,3.961424022912979
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,2,16384,1,0,10.700363874435425
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,4,1,0,0.6645999923348427
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,8,1,0,0.654619999229908
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,16,1,0,0.6667359992861748
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,32,1,0,0.669844001531601
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,64,1,0,0.6559640020132065
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,128,1,0,0.6703080013394356
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,256,1,0,0.689740002155304
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,512,1,0,0.7298559993505478
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,1024,1,0,0.819180004298687
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,2048,1,0,1.22325599193573
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,4096,1,0,2.60589998960495
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,8192,1,0,6.912368059158325
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,4,16384,1,0,21.40235209465027
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,4,1,0,0.6718400046229362
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,8,1,0,0.6660280004143715
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,16,1,0,0.6641679927706718
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,32,1,0,0.6695479899644852
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,64,1,0,0.6627920046448708
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,128,1,0,0.6908520013093948
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,256,1,0,0.7374840080738068
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,512,1,0,0.7942960038781166
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,1024,1,0,1.1153719872236252
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,2048,1,0,2.1888840198516846
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,4096,1,0,5.062780022621155
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,8192,1,0,14.147340059280396
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,8,16384,1,0,42.03447198867798
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,4,1,0,0.6920119971036911
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,8,1,0,0.6837280094623566
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,16,1,0,0.6822079941630363
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,32,1,0,0.6880839988589287
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,64,1,0,0.7047279924154282
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,128,1,0,0.7496960014104843
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,256,1,0,0.8098639994859695
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,512,1,0,1.096180021762848
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,1024,1,0,2.2902559638023376
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,2048,1,0,4.134824097156525
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,4096,1,0,10.239527940750122
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,16,8192,1,0,27.862468004226685
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,4,1,0,0.72171600908041
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,8,1,0,0.7190760001540184
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,16,1,0,0.7141520082950592
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,32,1,0,0.7362519949674606
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,64,1,0,0.7828600034117699
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,128,1,0,0.8422079905867577
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,256,1,0,1.110252022743225
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,512,1,0,1.9517199993133545
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,1024,1,0,3.742668032646179
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,2048,1,0,8.43997585773468
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,32,4096,1,0,20.400219678878784
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,4,1,0,0.7762719914317131
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,8,1,0,0.7827480062842369
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,16,1,0,0.7971879914402962
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,32,1,0,0.8364320024847984
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,64,1,0,0.9057440012693405
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,128,1,0,1.1735079884529114
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,256,1,0,2.006643980741501
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,512,1,0,3.691884011030197
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,1024,1,0,7.526192009449005
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,64,2048,1,0,16.96139168739319
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,4,1,0,0.9035039991140366
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,8,1,0,0.9289639964699745
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,16,1,0,0.9752560034394264
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,32,1,0,1.047540009021759
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,64,1,0,1.295407995581627
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,128,1,0,2.100100040435791
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,256,1,0,3.7781120240688324
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,512,1,0,7.302255988121033
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,128,1024,1,0,15.099768161773682
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,256,4,1,0,1.2671719789505005
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,256,8,1,0,1.2978360056877136
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,256,16,1,0,1.3824719935655594
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,256,32,1,0,1.5840920060873032
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,256,64,1,0,2.346619963645935
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,256,128,1,0,3.9311919808387756
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,256,256,1,0,7.3802080154418945
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,256,512,1,0,14.655499815940857
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,512,4,1,0,1.9164360016584396
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,512,8,1,0,1.9742200076580048
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,512,16,1,0,2.205548018217087
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,512,32,1,0,3.030528038740158
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,512,64,1,0,4.411435961723328
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,512,128,1,0,7.695371925830841
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,512,256,1,0,14.791815996170044
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1024,4,1,0,3.2775039970874786
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1024,8,1,0,3.415116012096405
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1024,16,1,0,4.218092143535614
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1024,32,1,0,5.658640086650848
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1024,64,1,0,8.703155994415283
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,32,1024,128,1,0,15.353916049003601
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,4,1,0,0.7276800125837326
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,8,1,0,0.6655799895524979
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,16,1,0,0.686352014541626
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,32,1,0,0.700628012418747
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,64,1,0,0.6687920019030571
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,128,1,0,0.6841559931635857
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,256,1,0,0.6757560074329376
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,512,1,0,0.6651280000805855
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,1024,1,0,0.686208002269268
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,2048,1,0,0.7008039876818657
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,4096,1,0,0.7997720092535019
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,8192,1,0,1.2216320037841797
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1,16384,1,0,2.7114920020103455
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,4,1,0,0.6768599972128868
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,8,1,0,0.6672679930925369
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,16,1,0,0.6909879967570305
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,32,1,0,0.7221119925379753
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,64,1,0,0.6768200024962425
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,128,1,0,0.6669640019536018
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,256,1,0,0.6773560047149658
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,512,1,0,0.6679119989275932
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,1024,1,0,0.711743988096714
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,2048,1,0,0.7444920018315315
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,4096,1,0,1.022332027554512
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,8192,1,0,1.9430200010538101
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,2,16384,1,0,5.3853360414505005
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,4,1,0,0.6697519943118095
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,8,1,0,0.6815119981765747
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,16,1,0,0.6783920004963875
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,32,1,0,0.7186599969863892
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,64,1,0,0.7002639919519424
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,128,1,0,0.6664400100708008
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,256,1,0,0.6852080076932907
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,512,1,0,0.7108759880065918
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,1024,1,0,0.729155994951725
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,2048,1,0,0.8963720053434372
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,4096,1,0,1.51985602080822
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,8192,1,0,3.6440640091896057
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,4,16384,1,0,10.920707941055298
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,4,1,0,0.690035991370678
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,8,1,0,0.692115992307663
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,16,1,0,0.684748001396656
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,32,1,0,0.6851640045642853
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,64,1,0,0.6887519955635071
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,128,1,0,0.8174799904227257
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,256,1,0,0.7111760005354881
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,512,1,0,0.7560919970273972
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,1024,1,0,0.8506279960274696
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,2048,1,0,1.2967599928379059
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,4096,1,0,2.7837239801883698
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,8192,1,0,7.251039922237396
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,8,16384,1,0,21.689827919006348
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,4,1,0,0.6961679980158806
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,8,1,0,0.7017680108547211
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,16,1,0,0.7005520015954971
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,32,1,0,0.6945040002465248
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,64,1,0,0.7026440054178238
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,128,1,0,0.7230799943208694
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,256,1,0,0.7647239938378334
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,512,1,0,0.844311997294426
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,1024,1,0,1.205499991774559
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,2048,1,0,2.374659985303879
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,4096,1,0,5.430239975452423
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,16,8192,1,0,14.730124115943909
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,4,1,0,0.7301600053906441
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,8,1,0,0.7336239963769913
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,16,1,0,0.725588008761406
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,32,1,0,0.7413400039076805
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,64,1,0,0.752808004617691
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,128,1,0,0.7956480011343956
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,256,1,0,0.8745800033211708
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,512,1,0,1.1992439925670624
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,1024,1,0,2.2057160139083862
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,2048,1,0,4.635259926319122
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,32,4096,1,0,11.03769600391388
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,4,1,0,0.8453320041298866
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,8,1,0,0.8389919921755791
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,16,1,0,0.8250439986586571
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,32,1,0,0.8734080046415329
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,64,1,0,0.8924959972500801
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,128,1,0,1.043067991733551
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,256,1,0,1.2437919974327087
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,512,1,0,2.1834879517555237
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,1024,1,0,4.256219923496246
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,64,2048,1,0,9.22459602355957
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,4,1,0,0.892584003508091
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,8,1,0,0.8982480093836784
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,16,1,0,0.9269120022654533
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,32,1,0,0.9673600122332573
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,64,1,0,1.0327240079641342
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,128,1,0,1.3487320095300674
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,256,1,0,2.283783942461014
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,512,1,0,4.212288022041321
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,128,1024,1,0,8.322727918624878
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,256,4,1,0,1.1415720134973526
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,256,8,1,0,1.1946799904108047
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,256,16,1,0,1.2148639857769012
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,256,32,1,0,1.3331560045480728
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,256,64,1,0,1.5688599944114685
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,256,128,1,0,2.458571970462799
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,256,256,1,0,4.374235987663269
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,256,512,1,0,8.240892052650452
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,512,4,1,0,1.6986919939517975
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,512,8,1,0,1.7479079961776733
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,512,16,1,0,1.8492119759321213
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,512,32,1,0,2.0558639764785767
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,512,64,1,0,2.924796015024185
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,512,128,1,0,4.734455943107605
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,512,256,1,0,8.606815934181213
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1024,4,1,0,2.8285239934921265
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1024,8,1,0,2.9259400367736816
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1024,16,1,0,3.1551759839057922
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1024,32,1,0,3.977396011352539
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1024,64,1,0,5.691779971122742
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,flashinfer,fp8_block,fp8,16,1024,128,1,0,9.45986807346344
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,4,1,0,0.6750600039958954
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,8,1,0,0.6664000079035759
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,16,1,0,0.6539919972419739
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,32,1,0,0.6661440059542656
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,64,1,0,0.6651360020041466
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,128,1,0,0.665307991206646
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,256,1,0,0.6720720008015633
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,512,1,0,0.7132319957017899
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,1024,1,0,0.8245599940419197
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,2048,1,0,1.254148006439209
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,4096,1,0,2.5038479566574097
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,8192,1,0,6.6572840213775635
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1,16384,1,0,20.713072061538696
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,4,1,0,0.6654839962720871
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,8,1,0,0.6586640030145645
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,16,1,0,0.6599479988217354
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,32,1,0,0.6714479923248291
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,64,1,0,0.6716720014810562
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,128,1,0,0.6793320029973984
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,256,1,0,0.7145440056920052
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,512,1,0,0.7995719984173775
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,1024,1,0,1.1564359813928604
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,2048,1,0,2.0845880210399628
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,4096,1,0,4.882507920265198
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,8192,1,0,13.567631959915161
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,2,16384,1,0,41.32147550582886
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,4,1,0,0.6719920039176941
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,8,1,0,0.6601120010018349
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,16,1,0,0.6766119971871376
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,32,1,0,0.6725479960441589
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,64,1,0,0.6761040017008781
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,128,1,0,0.7167360037565231
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,256,1,0,0.8208919987082481
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,512,1,0,1.1343679875135422
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,1024,1,0,1.8982160091400146
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,2048,1,0,3.9844120144844055
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,4096,1,0,9.874911904335022
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,4,8192,1,0,26.736695766448975
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,4,1,0,0.668600007891655
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,8,1,0,0.6882879957556725
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,16,1,0,0.6763479933142662
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,32,1,0,0.6903759986162186
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,64,1,0,0.7237440124154091
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,128,1,0,0.8083239942789078
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,256,1,0,1.0921599864959717
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,512,1,0,1.7869639992713928
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,1024,1,0,3.555415987968445
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,2048,1,0,7.95168000459671
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,8,4096,1,0,20.06663203239441
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,4,1,0,0.6978479996323586
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,8,1,0,0.7031520009040833
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,16,1,0,0.7051279917359352
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,32,1,0,0.7453000023961067
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,64,1,0,0.839263990521431
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,128,1,0,1.1093159914016724
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,256,1,0,1.7820560038089752
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,512,1,0,3.37963604927063
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,1024,1,0,7.09442001581192
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,16,2048,1,0,16.01394808292389
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,4,1,0,0.765067994594574
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,8,1,0,0.7880760058760643
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,16,1,0,0.8154359981417656
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,32,1,0,0.9182679951190948
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,64,1,0,1.178623989224434
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,128,1,0,1.8207679837942123
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,256,1,0,3.367264002561569
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,512,1,0,6.679275989532471
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,32,1024,1,0,14.302255868911743
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,4,1,0,0.9453439936041832
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,8,1,0,0.9942000135779381
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,16,1,0,1.0892800092697144
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,32,1,0,1.3224360048770905
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,64,1,0,1.925379991531372
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,128,1,0,3.4031600058078766
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,256,1,0,6.501459956169128
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,64,512,1,0,13.27291202545166
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,128,4,1,0,1.3499919921159744
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,128,8,1,0,1.4601199924945831
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,128,16,1,0,1.6819920241832733
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,128,32,1,0,2.2540640234947205
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,128,64,1,0,3.6550879776477814
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,128,128,1,0,6.58405601978302
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,128,256,1,0,13.140619993209839
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,256,4,1,0,2.142752021551132
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,256,8,1,0,2.3732720017433167
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,256,16,1,0,2.9125479757785797
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,256,32,1,0,4.23335999250412
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,256,64,1,0,7.084867894649506
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,256,128,1,0,13.22681200504303
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,512,4,1,0,3.736692041158676
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,512,8,1,0,4.223927974700928
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,512,16,1,0,5.61846798658371
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,512,32,1,0,8.30460000038147
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,512,64,1,0,14.018108129501343
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1024,4,1,0,6.926519989967346
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1024,8,1,0,8.25210416316986
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1024,16,1,0,10.858232140541077
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,128,1024,32,1,0,16.717504024505615
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,4,1,0,0.6632080078125
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,8,1,0,0.6800040006637573
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,16,1,0,0.6682039871811867
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,32,1,0,0.6787960082292557
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,64,1,0,0.6822080016136169
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,128,1,0,0.6755479946732521
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,256,1,0,0.6702480018138885
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,512,1,0,0.6983160004019737
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,1024,1,0,0.7266639918088913
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,2048,1,0,0.901528000831604
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,4096,1,0,1.4081359803676605
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,8192,1,0,3.400076001882553
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1,16384,1,0,10.581215977668762
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,4,1,0,0.6749680116772652
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,8,1,0,0.6705799996852875
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,16,1,0,0.6765120103955269
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,32,1,0,0.6735439971089363
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,64,1,0,0.674347996711731
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,128,1,0,0.6738759949803352
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,256,1,0,0.7103800028562546
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,512,1,0,0.7387919872999191
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,1024,1,0,0.8618800044059753
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,2048,1,0,1.2069879919290543
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,4096,1,0,2.586872011423111
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,8192,1,0,6.778280079364777
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,2,16384,1,0,20.495155811309814
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,4,1,0,0.6718360111117363
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,8,1,0,0.6939360052347183
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,16,1,0,0.6864160001277924
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,32,1,0,0.6962799951434135
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,64,1,0,0.6857720017433167
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,128,1,0,0.7061600089073181
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,256,1,0,0.751224011182785
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,512,1,0,0.8326839953660965
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,1024,1,0,1.109415978193283
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,2048,1,0,2.1550920009613037
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,4096,1,0,4.864524006843567
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,8192,1,0,13.839855909347534
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,4,16384,1,0,41.62386417388916
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,4,1,0,0.6890399977564812
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,8,1,0,0.6934080049395561
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,16,1,0,0.68784399330616
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,32,1,0,0.6963199973106384
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,64,1,0,0.7048440054059029
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,128,1,0,0.8093199953436852
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,256,1,0,0.8568599969148636
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,512,1,0,1.06148000061512
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,1024,1,0,1.9551640003919601
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,2048,1,0,4.062511920928955
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,4096,1,0,10.193643927574158
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,8,8192,1,0,26.899675846099854
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,4,1,0,0.7011719942092896
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,8,1,0,0.7061400040984154
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,16,1,0,0.7049320042133331
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,32,1,0,0.7372400015592575
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,64,1,0,0.7811000049114227
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,128,1,0,0.8566759899258614
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,256,1,0,1.0739479959011078
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,512,1,0,1.8723839819431305
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,1024,1,0,3.648420065641403
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,2048,1,0,8.165519952774048
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,16,4096,1,0,20.307912349700928
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,4,1,0,0.7442359998822212
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,8,1,0,0.7350560054183006
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,16,1,0,0.7549720034003258
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,32,1,0,0.8079480081796646
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,64,1,0,0.9038559943437576
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,128,1,0,1.1027879863977432
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,256,1,0,1.8795360028743744
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,512,1,0,3.4500440061092377
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,1024,1,0,7.274387955665588
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,32,2048,1,0,16.47875165939331
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,4,1,0,0.8371519967913628
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,8,1,0,0.8551120012998581
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,16,1,0,0.9056880101561546
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,32,1,0,1.0031799972057343
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,64,1,0,1.1912559866905212
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,128,1,0,1.9444160014390945
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,256,1,0,3.4728200137615204
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,512,1,0,6.725624084472656
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,64,1024,1,0,14.689836025238037
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,4,1,0,1.0921999961137772
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,8,1,0,1.1529680043458939
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,16,1,0,1.231432005763054
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,32,1,0,1.4141919910907745
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,64,1,0,2.116371989250183
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,128,1,0,3.562291979789734
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,256,1,0,6.7596160769462585
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,128,512,1,0,13.779075860977173
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,256,4,1,0,1.5945679992437363
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,256,8,1,0,1.693908005952835
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,256,16,1,0,1.8472840040922165
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,256,32,1,0,2.5362280011177063
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,256,64,1,0,3.8841440081596375
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,256,128,1,0,6.946303963661194
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,256,256,1,0,13.507896184921265
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,512,4,1,0,2.5809760093688965
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,512,8,1,0,2.714228004217148
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,512,16,1,0,3.3680480122566223
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,512,32,1,0,4.730780005455017
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,512,64,1,0,7.5964959263801575
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,512,128,1,0,13.707328081130981
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1024,4,1,0,4.546931982040405
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1024,8,1,0,5.1749719977378845
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1024,16,1,0,6.58704799413681
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1024,32,1,0,9.277531862258911
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,64,1024,64,1,0,15.121359825134277
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,4,1,0,0.681879997253418
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,8,1,0,0.7271480038762093
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,16,1,0,0.7324160113930702
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,32,1,0,0.7002759948372841
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,64,1,0,0.6846119984984398
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,128,1,0,0.6888839900493622
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,256,1,0,0.7039759978652
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,512,1,0,0.7197200059890747
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,1024,1,0,0.7034480050206184
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,2048,1,0,0.7967760041356087
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,4096,1,0,1.0152640044689178
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,8192,1,0,1.8725640028715134
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1,16384,1,0,5.100776016712189
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,4,1,0,0.7040439993143082
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,8,1,0,0.6892480030655861
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,16,1,0,0.6888640001416206
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,32,1,0,0.7115919962525368
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,64,1,0,0.7238560020923615
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,128,1,0,0.7113479971885681
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,256,1,0,0.7024599984288216
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,512,1,0,0.7359599992632866
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,1024,1,0,0.7793120071291924
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,2048,1,0,0.91702800989151
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,4096,1,0,1.4693600088357925
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,8192,1,0,3.492524027824402
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,2,16384,1,0,10.889155864715576
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,4,1,0,0.6943119913339615
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,8,1,0,0.7036040052771568
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,16,1,0,0.7200839966535568
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,32,1,0,0.7139079943299294
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,64,1,0,0.7195360064506531
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,128,1,0,0.715503990650177
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,256,1,0,0.7228920012712479
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,512,1,0,0.8132079988718033
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,1024,1,0,0.8503839895129204
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,2048,1,0,1.2468760013580322
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,4096,1,0,2.638327956199646
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,8192,1,0,6.981092035770416
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,4,16384,1,0,21.18649983406067
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,4,1,0,0.7385959923267365
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,8,1,0,0.7323880046606064
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,16,1,0,0.7595439851284027
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,32,1,0,0.7217520028352737
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,64,1,0,0.7244519963860512
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,128,1,0,0.7245319932699203
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,256,1,0,0.7794039994478226
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,512,1,0,0.8418520018458366
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,1024,1,0,1.1515159904956818
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,2048,1,0,2.226287990808487
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,4096,1,0,5.06333601474762
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,8,8192,1,0,14.0109121799469
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,4,1,0,0.7137079909443855
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,8,1,0,0.707664005458355
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,16,1,0,0.7445000037550926
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,32,1,0,0.7475759983062744
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,64,1,0,0.7360880002379417
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,128,1,0,0.8093759939074516
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,256,1,0,0.8806559965014458
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,512,1,0,1.1090520024299622
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,1024,1,0,2.0327240228652954
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,2048,1,0,4.248500049114227
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,16,4096,1,0,10.37529194355011
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,4,1,0,0.7620360031723976
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,8,1,0,0.7695320025086403
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,16,1,0,0.7566679939627647
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,32,1,0,0.7655079960823059
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,64,1,0,0.821212001144886
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,128,1,0,0.8716680034995079
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,256,1,0,1.1276679933071136
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,512,1,0,1.961428016424179
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,1024,1,0,3.8135319650173187
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,32,2048,1,0,8.539223790168762
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,4,1,0,0.7999119982123375
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,8,1,0,0.8077000007033348
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,16,1,0,0.8304440006613731
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,32,1,0,0.9043200090527534
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,64,1,0,0.941708005964756
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,128,1,0,1.1921440064907074
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,256,1,0,1.998512014746666
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,512,1,0,3.669152021408081
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,64,1024,1,0,7.528532028198242
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,4,1,0,0.9550520032644272
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,8,1,0,0.9904520064592361
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,16,1,0,1.037435993552208
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,32,1,0,1.104300007224083
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,64,1,0,1.3196559846401215
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,128,1,0,2.096323996782303
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,256,1,0,3.729215979576111
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,128,512,1,0,7.252988040447235
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,256,4,1,0,1.3025240153074265
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,256,8,1,0,1.3492439985275269
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,256,16,1,0,1.4274080097675323
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,256,32,1,0,1.6301999986171722
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,256,64,1,0,2.3353200256824493
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,256,128,1,0,3.9048000276088715
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,256,256,1,0,7.289667963981628
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,512,4,1,0,2.051947981119156
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,512,8,1,0,2.070743978023529
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,512,16,1,0,2.3001720011234283
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,512,32,1,0,2.974572002887726
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,512,64,1,0,4.446424067020416
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,512,128,1,0,7.682016015052795
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1024,4,1,0,3.476572036743164
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1024,8,1,0,3.5547119975090027
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1024,16,1,0,4.310984015464783
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1024,32,1,0,5.751796007156372
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,32,1024,64,1,0,8.758228182792664
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,4,1,0,0.6622920036315918
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,8,1,0,0.7425199970602989
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,16,1,0,0.714651994407177
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,32,1,0,0.6815920025110245
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,64,1,0,0.6811359897255898
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,128,1,0,0.6746880039572716
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,256,1,0,0.6575240045785904
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,512,1,0,0.6814080029726028
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,1024,1,0,0.7575479969382286
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,2048,1,0,0.7559880018234253
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,4096,1,0,0.8073440045118332
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,8192,1,0,1.2301480025053024
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1,16384,1,0,2.7060279846191406
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,4,1,0,0.6869040057063103
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,8,1,0,0.6631679981946945
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,16,1,0,0.6722439974546432
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,32,1,0,0.6760639920830727
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,64,1,0,0.6716959998011589
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,128,1,0,0.7062840014696121
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,256,1,0,0.7030360028147697
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,512,1,0,0.7103919982910156
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,1024,1,0,0.7079439982771873
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,2048,1,0,0.7334839925169945
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,4096,1,0,1.0135080143809319
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,8192,1,0,1.942204013466835
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,2,16384,1,0,5.374475955963135
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,4,1,0,0.6981600075960159
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,8,1,0,0.6782119944691658
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,16,1,0,0.6747080013155937
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,32,1,0,0.6912320032715797
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,64,1,0,0.6781920120120049
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,128,1,0,0.6862759962677956
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,256,1,0,0.684079997241497
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,512,1,0,0.7289119884371758
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,1024,1,0,0.7421799972653389
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,2048,1,0,0.914620004594326
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,4096,1,0,1.5262559801340103
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,8192,1,0,3.7177399694919586
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,4,16384,1,0,10.801163911819458
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,4,1,0,0.7034440040588379
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,8,1,0,0.6880839988589287
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,16,1,0,0.6925720125436783
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,32,1,0,0.6733839958906174
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,64,1,0,0.7350960075855255
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,128,1,0,0.6929440125823021
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,256,1,0,0.7213119938969612
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,512,1,0,0.7419519945979118
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,1024,1,0,0.8706320002675056
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,2048,1,0,1.3077720254659653
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,4096,1,0,2.8150720596313477
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,8192,1,0,7.250572025775909
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,8,16384,1,0,21.79585576057434
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,4,1,0,0.709476001560688
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,8,1,0,0.7039160057902336
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,16,1,0,0.6986519917845726
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,32,1,0,0.7155599892139435
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,64,1,0,0.7084320113062859
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,128,1,0,0.7362559884786606
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,256,1,0,0.7689359933137894
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,512,1,0,0.8581439927220345
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,1024,1,0,1.207384005188942
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,2048,1,0,2.3864279985427856
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,4096,1,0,5.497235953807831
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,16,8192,1,0,14.901364207267761
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,4,1,0,0.7416080087423325
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,8,1,0,0.7411560043692589
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,16,1,0,0.7605879977345467
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,32,1,0,0.7549479976296425
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,64,1,0,0.779619999229908
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,128,1,0,0.7996960058808327
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,256,1,0,0.8763800039887428
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,512,1,0,1.1899639964103699
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,1024,1,0,2.231528013944626
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,2048,1,0,4.626976013183594
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,32,4096,1,0,11.095128059387207
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,4,1,0,0.7916239947080612
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,8,1,0,0.8123360127210617
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,16,1,0,0.8179040029644966
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,32,1,0,0.8407159969210625
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,64,1,0,0.8497200012207031
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,128,1,0,0.980768010020256
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,256,1,0,1.2493759840726852
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,512,1,0,2.174724042415619
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,1024,1,0,4.268839955329895
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,64,2048,1,0,9.247992038726807
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,4,1,0,0.9253840073943138
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,8,1,0,0.939580000936985
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,16,1,0,0.945468008518219
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,32,1,0,0.9640640094876289
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,64,1,0,1.0526600182056427
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,128,1,0,1.344196006655693
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,256,1,0,2.2678559720516205
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,512,1,0,4.144800007343292
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,128,1024,1,0,8.471260070800781
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,256,4,1,0,1.1850440055131912
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,256,8,1,0,1.1973600089550018
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,256,16,1,0,1.2534200102090836
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,256,32,1,0,1.335408017039299
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,256,64,1,0,1.5923800021409988
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,256,128,1,0,2.4411799609661102
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,256,256,1,0,4.285856068134308
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,256,512,1,0,8.143587827682495
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,512,4,1,0,1.7355359941720963
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,512,8,1,0,1.7889679968357086
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,512,16,1,0,1.8986479938030243
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,512,32,1,0,2.103060007095337
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,512,64,1,0,2.91935196518898
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,512,128,1,0,4.732387959957123
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,512,256,1,0,8.507767915725708
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1024,4,1,0,2.9265680015087128
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1024,8,1,0,2.986592024564743
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1024,16,1,0,3.23744398355484
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1024,32,1,0,3.988484025001526
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1024,64,1,0,5.681859970092773
SGLang,0.5.6.post2,NVIDIA H100 80GB HBM3,mla_context,fa3,fp8_block,fp8,16,1024,128,1,0,9.232435941696167
